diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/README.md b/REGEN-main/cosmos_policy/_src/imaginaire/attention/README.md new file mode 100644 index 0000000000000000000000000000000000000000..81fa3324614c26d7190dac2fab683a3d516f2b45 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/README.md @@ -0,0 +1,47 @@ +# Imaginaire Attention Subpackage + +A subpackage within cosmos_policy._src.imaginaire that integrates only the best and most reliable +solutions, and provides simple APIs to end-users. + +For more information, please refer to the [docs](docs/). + +## Basic API + +```python +from cosmos_policy._src.imaginaire.attention import attention + +output = attention( + query=query, + key=key, + value=value, +) +``` + +* **Optional** `scale`: attention (softmax/dot product) scale. Defaults to `head_dim ** -0.5`. +* **Optional** `return_lse`: returns logsumexp if `True` +* **Optional** `backend`: explicitly set backend instead of automatically selecting the best compatible + +## Tensor layouts + +Imaginaire Attention only supports one tensor memory layout: +heads-last torch contiguous (`torch.contiguous_format`). + +With this layout, input tensors `query`, `key`, and `value` are represented as rank-4 tensors, with +dimension 0 representing batch, dimension 1 representing sequence length, dimension 2 representing +attention heads, and dimension 3 representing head dimension. +This layout is also consistent with the `contiguous_format` memory layout in PyTorch, meaning the +right-most dimension (head dimension) is the major dimension (has stride 1), and tokens from +different heads are interleaved. + +```python +def verify_heads_last_contig_tensor(x: Tensor): + assert x.shape[0] == batch + assert x.shape[1] == seqlen + assert x.shape[2] == heads + assert x.shape[3] == head_dim + + assert x.stride(3) == 1 + assert x.stride(2) == head_dim + assert x.stride(1) == heads * head_dim + assert x.stride(0) == heads * head_dim * seqlen +``` diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..c2893be29c398a1df8c326bd3f9428f36ab11ff8 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/__init__.py @@ -0,0 +1,29 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + + +""" + +from cosmos_policy._src.imaginaire.attention.frontend import ( + attention, + multi_dimensional_attention, + spatio_temporal_attention, +) + +__all__ = ["attention", "multi_dimensional_attention", "spatio_temporal_attention"] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/backends.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/backends.py new file mode 100644 index 0000000000000000000000000000000000000000..aeb42a1141f2aa10fcbc5814fe35b15861b52071 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/backends.py @@ -0,0 +1,348 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Frontend APIs +""" + +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.cudnn.checks import cudnn_attention_check +from cosmos_policy._src.imaginaire.attention.flash2.checks import flash2_attention_check +from cosmos_policy._src.imaginaire.attention.flash3.checks import flash3_attention_check +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.natten.checks import ( + natten_attention_check, + natten_multi_dim_attention_check, +) +from cosmos_policy._src.imaginaire.attention.utils import get_arch_tag +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +BACKEND_CHECK_MAP = { + "cudnn": cudnn_attention_check, + "natten": natten_attention_check, + "flash2": flash2_attention_check, + "flash3": flash3_attention_check, +} + +BACKEND_MULTI_DIM_CHECK_MAP = { + "natten": natten_multi_dim_attention_check, +} + + +def is_backend_compatible( + backend: str, + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool, + causal_type: CausalType | None, + is_varlen: bool, + raise_error: bool = False, +) -> bool: + """ + Input validation function a specified backend. + Runs the common and backend-specific checks. Returns False if any checks fail, otherwise True. + + Parameters: + backend (str): selected backend. + + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`). + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`). + + is_causal (bool): whether or not causal masking is enabled. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`. Required when `is_causal = True`. + + is_varlen (bool): whether or not a variable length (varlen) use case. Must be inferred + beforehand based on arguments such as seqlens_{Q,KV} or cumulative_seqlen_{Q,KV} being + passed. + + raise_error (bool): whether to raise an error if any checks fail or no backend is selected, + instead of just returning False. Default is False. + + Returns: + success (bool): whether use case is compatible with the backend. + + """ + if backend is None: + raise ValueError("Cannot pass None backend to is_backend_compatible.") + + if backend not in BACKEND_CHECK_MAP: + raise ValueError(f"Unrecognized backend name {backend}.") + + return BACKEND_CHECK_MAP[backend]( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + is_varlen=is_varlen, + raise_error=raise_error, + ) + + +def get_backend_list(arch_tag: int) -> list[str]: + """ + Returns list of supported backends according to arch tag (attention.utils.get_arch_tag). + Backends are ordered based on their known performance levels, so that the best-performing + compatible backend is selected. + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + backend_list (list[str]): a list of backend names (string). Empty if device is not supported. + + """ + + if arch_tag < 75: + log.debug(f"Minimum architecture supported for Attention is 75, got {arch_tag=}.") + return [] + + if arch_tag == 90: + return [ + "flash3", + "cudnn", + "natten", + "flash2", + ] + + if arch_tag in [100, 103]: + return [ + # "flash4", + "cudnn", + "natten", + "flash2", + ] + + if arch_tag >= 80: + return [ + "flash2", + "cudnn", + "natten", + ] + + return ["natten"] + + +def choose_backend( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool, + causal_type: CausalType | None, + is_varlen: bool, + backend: str | None = None, + raise_error: bool = True, +) -> str | None: + """ + Selects a compatible backend, unless one is already selected, which runs its corresponding + checks. + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`). + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`). + + is_causal (bool): whether or not causal masking is enabled. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`. Required when `is_causal = True`. + + is_varlen (bool): whether or not a variable length (varlen) use case. Must be inferred + beforehand based on arguments such as seqlens_{Q,KV} or cumulative_seqlen_{Q,KV} being + passed. + + backend (str | None): selected backend, if any. + + raise_error (bool): whether to raise an error if any checks fail or no backend is selected, + instead of just returning False. Default is **True**. + + Returns: + backend (str | None): selected backend, or None if no backends are compatible. + + """ + if backend is not None: + if is_backend_compatible( + backend=backend, + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + is_varlen=is_varlen, + raise_error=raise_error, + ): + return backend + return None + + arch_tag = get_arch_tag(query.device) + backend_list = get_backend_list(arch_tag) + for backend in backend_list: + if is_backend_compatible( + backend=backend, + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + is_varlen=is_varlen, + raise_error=False, + ): + return backend + + if not raise_error: + return None + + raise ValueError( + "Could not find a compatible Attention backend for this use case / device. " + "Try running with debug logs to find out why." + ) + + +def is_multi_dim_backend_compatible( + backend: str, + query: Tensor, + key: Tensor, + value: Tensor, + raise_error: bool = False, +) -> bool: + """ + Input validation function a specified multi-dimensional backend. + Runs the common and backend-specific checks. Returns False if any checks fail, otherwise True. + + Parameters: + backend (str): selected backend. + + query (Tensor): 4-D, 5-D, or 6-D query tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads, head_dim]`). + + key (Tensor): 4-D, 5-D, or 6-D key tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim]`). + + value (Tensor): 4-D, 5-D, or 6-D value tensor, with heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim_v]`). + + raise_error (bool): whether to raise an error if any checks fail or no backend is selected, + instead of just returning False. Default is False. + + Returns: + success (bool): whether use case is compatible with the backend. + + """ + if backend is None: + raise ValueError("Cannot pass None backend to is_backend_compatible.") + + if backend not in BACKEND_MULTI_DIM_CHECK_MAP: + raise ValueError(f"Unrecognized backend name {backend}.") + + return BACKEND_MULTI_DIM_CHECK_MAP[backend]( + query=query, + key=key, + value=value, + raise_error=raise_error, + ) + + +def get_multi_dim_backend_list(arch_tag: int) -> list[str]: + """ + Returns list of supported multi-dimensional backends according to arch tag (attention.utils.get_arch_tag). + Backends are ordered based on their known performance levels, so that the best-performing + compatible backend is selected. + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + backend_list (list[str]): a list of backend names (string). Empty if device is not supported. + + """ + + if arch_tag < 75: + log.debug(f"Minimum architecture supported for Multi-Dimensional Attention is 75, got {arch_tag=}.") + return [] + + # NATTEN is the only supported backend for now + return ["natten"] + + +def choose_multi_dim_backend( + query: Tensor, + key: Tensor, + value: Tensor, + backend: str | None = None, +) -> str: + """ + Selects a compatible multi-dimensional backend, unless one is already selected, which runs its + corresponding checks. + + Parameters: + query (Tensor): 4-D, 5-D, or 6-D query tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads, head_dim]`). + + key (Tensor): 4-D, 5-D, or 6-D key tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim]`). + + value (Tensor): 4-D, 5-D, or 6-D value tensor, with heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim_v]`). + + backend (str | None): selected backend, if any. + + Returns: + backend (str): selected backend. + + """ + if backend is not None: + assert is_multi_dim_backend_compatible( + backend=backend, + query=query, + key=key, + value=value, + raise_error=True, + ) + return backend + + arch_tag = get_arch_tag(query.device) + backend_list = get_multi_dim_backend_list(arch_tag) + for backend in backend_list: + if is_multi_dim_backend_compatible( + backend=backend, + query=query, + key=key, + value=value, + raise_error=False, + ): + return backend + + raise ValueError( + "Could not find a compatible Multi-Dimensional Attention backend for this use case / device. " + "Try running with debug logs to find out why." + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/checks.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/checks.py new file mode 100644 index 0000000000000000000000000000000000000000..21561ff8eb43386e28a5b60a2c6cbc75e9de2465 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/checks.py @@ -0,0 +1,500 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Common, op-specific, and backend-specific checks +""" + +from collections.abc import Sequence +from functools import partial +from typing import Any + +import torch +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.utils import log_or_raise_error +from cosmos_policy._src.imaginaire.attention.varlen import generate_varlen_parameters + + +def _universal_tensor_checks(query: Tensor, key: Tensor, value: Tensor, raise_error: bool = True) -> bool: + target_fn = partial(log_or_raise_error, raise_error=raise_error) + + if query.is_sparse or key.is_sparse or value.is_sparse: + target_fn("This operation does not support sparse tensors.", exception=NotImplementedError) + return False + + if query.is_nested or key.is_nested or value.is_nested: + target_fn("This operation does not support nested tensors.", exception=NotImplementedError) + return False + + if query.device != key.device or query.device != value.device: + target_fn( + f"Query, key, and value must be on the same device, got {query.device=}, {key.device=}, {value.device=}.", + exception=ValueError, + ) + return False + + if query.dtype != key.dtype or query.dtype != value.dtype: + target_fn( + f"Query, key, and value must assume the same data type, got {query.dtype=}, {key.dtype=}, {value.dtype=}.", + exception=ValueError, + ) + return False + + return True + + +def _universal_attention_checks( + query: Tensor, + key: Tensor, + value: Tensor, + supported_dtypes_forward: list[torch.dtype] | None = None, + supported_dtypes_backward: list[torch.dtype] | None = None, + supports_mla: bool = True, + supports_gqa_mqa: bool = True, + raise_error: bool = True, + backend_name: str | None = None, +) -> bool: + backend_name = backend_name or "Attention" + if not _universal_tensor_checks(query, key, value, raise_error=raise_error): + return False + + target_fn = partial(log_or_raise_error, raise_error=raise_error) + + if query.dim() != key.dim() or query.dim() != value.dim(): + target_fn( + f"Q, K, and V must have the same rank, got {query.dim()=}, {key.dim()=}, {value.dim()=}.", + exception=ValueError, + ) + return False + + if query.shape[0] != key.shape[0] or query.shape[0] != value.shape[0]: + target_fn( + f"Q, K, and V must match in batch size, got {query.shape[0]=}, {key.shape[0]=}, {value.shape[0]=}.", + exception=ValueError, + ) + return False + + if query.shape[-1] != key.shape[-1]: + target_fn( + f"Q and K head dims must match, got {query.shape[-1]=}, {key.shape[-1]=}.", + exception=ValueError, + ) + return False + + if key.shape[-2] != value.shape[-2]: + target_fn( + f"K and V must always have the same number of heads, got {key.shape[2]=}, {value.shape[2]=}.", + exception=ValueError, + ) + return False + + if not supports_mla and query.shape[-1] != value.shape[-1]: + target_fn( + f"{backend_name} does not support different head dims for QK and V, got " + f"{query.shape[-1]=}, {value.shape[-1]=}.", + exception=ValueError, + ) + return False + + if not supports_gqa_mqa and (query.shape[-2] != key.shape[-2] or query.shape[-2] != value.shape[-2]): + target_fn( + f"{backend_name} does not support GQA/MQA, therefore number of heads in Q, K, and V " + f"must match, got {query.shape[-2]=}, {key.shape[-2]=}, {value.shape[-2]=}.", + exception=ValueError, + ) + return False + + if supports_gqa_mqa: + heads_q = query.shape[-2] + heads_kv = key.shape[-2] + + if heads_q < heads_kv or heads_q % heads_kv != 0: + target_fn( + f"KV heads must evenly divide Q heads, got {heads_q=}, {heads_kv=}.", + exception=ValueError, + ) + return False + + # _universal_tensor_checks guarantees query.dtype == key.dtype == value.dtype + if supported_dtypes_forward is not None and query.dtype not in supported_dtypes_forward: + target_fn( + f"{backend_name} does not support forward pass (inference) with data type {query.dtype}; " + f"supported dtypes: {supported_dtypes_forward}.", + exception=ValueError, + ) + return False + + if supported_dtypes_backward is not None and query.requires_grad and query.dtype not in supported_dtypes_backward: + target_fn( + f"{backend_name} does not support backward pass (training) with data type {query.dtype}; " + f"supported dtypes: {supported_dtypes_backward}.", + exception=ValueError, + ) + return False + + return True + + +def attention_tensor_checks( + query: Tensor, + key: Tensor, + value: Tensor, + supported_dtypes_forward: list[torch.dtype] | None = None, + supported_dtypes_backward: list[torch.dtype] | None = None, + supports_mla: bool = True, + supports_gqa_mqa: bool = True, + raise_error: bool = True, + backend_name: str | None = None, +) -> bool: + backend_name = backend_name or "Attention" + if not _universal_tensor_checks(query, key, value, raise_error=raise_error): + return False + + if not _universal_attention_checks( + query=query, + key=key, + value=value, + supported_dtypes_forward=supported_dtypes_forward, + supported_dtypes_backward=supported_dtypes_backward, + supports_mla=supports_mla, + supports_gqa_mqa=supports_gqa_mqa, + raise_error=raise_error, + backend_name=backend_name, + ): + return False + + target_fn = partial(log_or_raise_error, raise_error=raise_error) + + if query.dim() != 4: + target_fn( + f"Attention expects 4-D tensors as inputs, got {query.dim()=}.", + exception=ValueError, + ) + return False + + if key.shape[1] != value.shape[1]: + target_fn( + f"K and V must match in sequence length, got {key.shape[1]=}, {value.shape[1]=}.", + exception=ValueError, + ) + return False + + return True + + +def varlen_tensor_checks( + query: Tensor, + key: Tensor, + value: Tensor, + seqlens_Q: Tensor | None = None, + seqlens_KV: Tensor | None = None, + cumulative_seqlen_Q: Tensor | None = None, + cumulative_seqlen_KV: Tensor | None = None, + max_seqlen_Q: int | None = None, + max_seqlen_KV: int | None = None, +) -> tuple[None, None, int, int] | tuple[Tensor, Tensor, int, int]: + if query.shape[0] != key.shape[0] or query.shape[0] != value.shape[0]: + raise ValueError( + f"Q, K, and V must match in batch size, got {query.shape[0]=}, {key.shape[0]=}, {value.shape[0]=}." + ) + + if all( + x is None + for x in [ + seqlens_Q, + seqlens_KV, + cumulative_seqlen_Q, + cumulative_seqlen_KV, + ] + ) and all( + x is None or x == 0 + for x in [ + max_seqlen_Q, + max_seqlen_KV, + ] + ): + # Not varlen + return None, None, 0, 0 + + if seqlens_Q is not None or seqlens_KV is not None: + # Generate cumulative_seqlen_{Q,KV}, max_seqlen_{Q,KV}, total_seqlen_{Q,KV} + # based on user input + return generate_varlen_parameters( + query=query, + key=key, + value=value, + seqlens_Q=seqlens_Q, + seqlens_KV=seqlens_KV, + ) + + # Validate user-input cumulative_seqlen_{Q,KV}, max_seqlen_{Q,KV}, total_seqlen_{Q,KV} + if any( + x is None + for x in [ + cumulative_seqlen_Q, + cumulative_seqlen_KV, + max_seqlen_Q, + max_seqlen_KV, + ] + ) or any( + x == 0 + for x in [ + max_seqlen_Q, + max_seqlen_KV, + ] + ): + raise ValueError( + "Variable length Attention requires all 6 of " + "cumulative_seqlen_{Q,KV}, max_seqlen_{Q,KV}, total_seqlen_{Q,KV} to be set." + ) + + if query.shape[0] != 1: + raise ValueError( + f"Variable length Attention only supports sequence-packed memory layout (batch = 1), got {query.shape[0]=}." + ) + + assert cumulative_seqlen_Q is not None + assert cumulative_seqlen_KV is not None + assert max_seqlen_Q is not None + assert max_seqlen_KV is not None + + if not isinstance(max_seqlen_Q, int) or not isinstance(max_seqlen_KV, int): + raise ValueError( + f"max_seqlen_Q and max_seqlen_KV must be ints, got {type(max_seqlen_Q)=}, {type(max_seqlen_KV)=}." + ) + + total_seqlen_Q = query.shape[1] + total_seqlen_KV = key.shape[1] + if max_seqlen_Q > total_seqlen_Q: + raise ValueError(f"Maximum sequence length cannot exceed total, got {max_seqlen_Q=}, {total_seqlen_Q=}.") + + if max_seqlen_KV > total_seqlen_KV: + raise ValueError(f"Maximum sequence length cannot exceed total, got {max_seqlen_KV=}, {total_seqlen_KV=}.") + + if max_seqlen_Q < 1 or max_seqlen_KV < 1: + raise ValueError(f"Maximum sequence length cannot be less than 1, got {max_seqlen_Q=}, {max_seqlen_KV=}.") + + if not isinstance(cumulative_seqlen_Q, Tensor) or not isinstance(cumulative_seqlen_KV, Tensor): + raise ValueError("cumulative_seqlen_Q and cumulative_seqlen_KV must both be tensors.") + + if cumulative_seqlen_Q.device != query.device or cumulative_seqlen_KV.device != query.device: + raise ValueError( + "cumulative_seqlen_Q and cumulative_seqlen_KV must be on the same device as QKV, but " + f"{cumulative_seqlen_Q.device=}, {cumulative_seqlen_KV.device=}, {query.device=}." + ) + + if cumulative_seqlen_Q.dtype != torch.int32 or cumulative_seqlen_KV.dtype != torch.int32: + raise ValueError( + "cumulative_seqlen_Q and cumulative_seqlen_KV must both be torch.int32 tensors, got " + f"{cumulative_seqlen_Q.dtype=}, {cumulative_seqlen_KV.dtype=}." + ) + + if cumulative_seqlen_Q.dim() != 1 or cumulative_seqlen_KV.dim() != 1: + raise ValueError( + "cumulative_seqlen_Q and cumulative_seqlen_KV must both be 1-D tensors, got " + f"{cumulative_seqlen_Q.dim()=}, {cumulative_seqlen_KV.dim()=}." + ) + + if cumulative_seqlen_Q.shape[0] != cumulative_seqlen_KV.shape[0]: + raise ValueError( + "cumulative_seqlen_Q and cumulative_seqlen_KV must match in size, got " + f"{cumulative_seqlen_Q.shape=}, {cumulative_seqlen_KV.shape=}." + ) + + if cumulative_seqlen_Q.shape[0] < 2: + raise ValueError( + "cumulative_seqlen_Q and cumulative_seqlen_KV must contain at least 2 elements, got " + f"{cumulative_seqlen_Q.shape=}, {cumulative_seqlen_KV.shape=}." + ) + + return ( + cumulative_seqlen_Q, + cumulative_seqlen_KV, + max_seqlen_Q, + max_seqlen_KV, + ) + + +def attention_param_checks( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool, + causal_type: CausalType, +): + if is_causal and (causal_type is None or not isinstance(causal_type, CausalType)): + raise ValueError( + f"Argument causal_type must be specified as an enum instance of CausalType when is_causal=True, got {causal_type=}." + ) + + assert query.dim() == key.dim() == value.dim() == 4 + assert key.shape[1] == value.shape[1] + if is_causal and causal_type == CausalType.DontCare and query.shape[1] != key.shape[1]: + raise ValueError( + "Causal mask type DontCare is only valid when seqlen_q == seqlen_kv, got " + f"{query.shape[1]=}, {key.shape[1]=}." + ) + + +def multi_dim_attention_tensor_checks( + query: Tensor, + key: Tensor, + value: Tensor, + supported_dtypes_forward: list[torch.dtype] | None = None, + supported_dtypes_backward: list[torch.dtype] | None = None, + supports_mla: bool = True, + supports_gqa_mqa: bool = True, + raise_error: bool = True, + backend_name: str | None = None, +) -> bool: + backend_name = backend_name or "Multi-Dimensional Attention" + if not _universal_tensor_checks(query, key, value, raise_error=raise_error): + return False + + if not _universal_attention_checks( + query=query, + key=key, + value=value, + supported_dtypes_forward=supported_dtypes_forward, + supported_dtypes_backward=supported_dtypes_backward, + supports_mla=supports_mla, + supports_gqa_mqa=supports_gqa_mqa, + raise_error=raise_error, + backend_name=backend_name, + ): + return False + + target_fn = partial(log_or_raise_error, raise_error=raise_error) + + if query.dim() not in [4, 5, 6]: + target_fn( + f"Multi-Dimensional Attention supports 4-D, 5-D, or 6-D tensors as inputs, got {query.dim()=}.", + exception=ValueError, + ) + return False + + num_dims = query.dim() - 3 # minus batch, heads, head_dim + + q_token_layout_shape = query.shape[1 : 1 + num_dims] + k_token_layout_shape = key.shape[1 : 1 + num_dims] + v_token_layout_shape = value.shape[1 : 1 + num_dims] + + if q_token_layout_shape != k_token_layout_shape or q_token_layout_shape != v_token_layout_shape: + target_fn( + "Q, K and V must match in their token layout shapes in multi-dimensional attention, " + f"got {q_token_layout_shape=}, {k_token_layout_shape=}, {v_token_layout_shape=}.", + exception=ValueError, + ) + return False + + return True + + +def check_valid_tuple_or_element(param: Any, num_dims: int, typename: type) -> tuple | None: + if isinstance(param, typename): + return tuple(param for _ in range(num_dims)) + + if isinstance(param, Sequence) and len(param) == num_dims and all(isinstance(x, typename) for x in param): + return param + + return None + + +def multi_dim_attention_param_filter( + query: Tensor, + window_size: tuple | int = -1, + stride: tuple | int = 1, + dilation: tuple | int = 1, + is_causal: tuple | bool = False, +) -> tuple[tuple, tuple, tuple, tuple, tuple, tuple]: + """ + Converts all multi-dimensional parameters to standard types. + """ + assert query.dim() in [4, 5, 6] + num_dims = query.dim() - 3 + + token_layout_shape = tuple(s for s in query.shape[1 : 1 + num_dims]) + + window_size_ = check_valid_tuple_or_element(window_size, num_dims, int) + if window_size_ is None: + raise ValueError( + f"Parameter 'window_size' must be either an int or tuple of {num_dims} ints, got {window_size=}." + ) + + stride_ = check_valid_tuple_or_element(stride, num_dims, int) + if stride_ is None: + raise ValueError(f"Parameter 'stride' must be either an int or tuple of {num_dims} ints, got {stride=}.") + + dilation_ = check_valid_tuple_or_element(dilation, num_dims, int) + if dilation_ is None: + raise ValueError(f"Parameter 'dilation' must be either an int or tuple of {num_dims} ints, got {dilation=}.") + + is_causal_ = check_valid_tuple_or_element(is_causal, num_dims, bool) + if is_causal_ is None: + raise ValueError( + f"Parameter 'is_causal' must be either a boolean or tuple of {num_dims} booleans, got {is_causal=}." + ) + + # Map -1 windows to corresponding size in token layout + window_size_ = tuple(w if w != -1 else x for x, w in zip(token_layout_shape, window_size_)) + + return token_layout_shape, window_size_, stride_, dilation_, is_causal_ + + +def multi_dim_attention_param_checks( + query: Tensor, + window_size: tuple, + stride: tuple, + dilation: tuple, + is_causal: tuple, +): + """ + Validates multi-dimensional parameters. + """ + assert query.dim() in [4, 5, 6] + num_dims = query.dim() - 3 + + token_layout_shape = tuple(s for s in query.shape[1 : 1 + num_dims]) + + if any(x <= 1 for x in token_layout_shape): + raise ValueError(f"Token layout dimensions must all be >= 2, got {token_layout_shape=} ({query.shape=}).") + + if any(w <= 1 for w in window_size): + raise ValueError( + "Parameter 'window_size' must be either -1 (no sparsity) or >= 2 along every dimension, " + f"got {window_size=}." + ) + + if any(w * d > x for x, w, d in zip(token_layout_shape, window_size, dilation)): + raise ValueError( + "The product of 'window_size' and 'dilation' cannot be greater than the input " + f"(token layout shape), got {window_size=}, {dilation=}, {token_layout_shape=} ({query.shape=})." + ) + + if any(s < 1 for s in stride): + raise ValueError(f"Parameter 'stride' allows positive integers only, got {stride=}.") + + if any(s > w for w, s in zip(window_size, stride)): + raise ValueError( + f"Parameter 'stride' cannot be greater than window size along any dimension, got {window_size=}, {stride=}." + ) + + if any(d < 1 for d in dilation): + raise ValueError(f"Parameter 'dilation' allows positive integers only, got {dilation=}.") diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..6fd1cb89c5208c74106c824659dd6e3588f65450 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/__init__.py @@ -0,0 +1,97 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +cuDNN Backend +""" + +import torch + +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +# (ahassani) [11-20-2025] Banning cuDNN until reliability issues are resolved. +# Versions checked: 91300, 91400, 91500 +# (ahassani) [12-01-2025] +# 91500 ran on both GB200 and H100 SXM. +CUDNN_DISALLOWED = True + +CUDNN_MIN_BACKEND_VERSION = 91300 +CUDNN_MIN_FRONTEND_VERSION = [1, 14, 0] + + +def cudnn_supported() -> bool: + """ + Returns whether cuDNN Attention is supported in this environment. + Requirements are: + * Presence of CUDA Runtime (via PyTorch) + * Presence of cuDNN and its Python frontend, meeting minimum version requirements + + This check guards imports / dependencies on the cuDNN package. + """ + if not torch.cuda.is_available(): + log.debug("cuDNN Attention is not supported because PyTorch did not detect CUDA runtime.") + return False + + try: + import cudnn + + except ImportError: + log.debug("cuDNN Attention is not supported because the frontend Python package was not found.") + return False + except Exception as e: + log.debug(f"cuDNN Attention is not supported because importing the frontend Python package failed: {e}") + return False + + if cudnn.backend_version() < CUDNN_MIN_BACKEND_VERSION: + log.debug( + "cuDNN Attention is not supported due to insufficient cuDNN backend version " + f"{cudnn.backend_version()=}, expected at least {CUDNN_MIN_BACKEND_VERSION=}." + ) + return False + + cudnn_frontend_version_split = cudnn.__version__.split(".") + if len(cudnn_frontend_version_split) != 3: + log.debug(f"Unable to parse cuDNN frontend version {cudnn.__version__}.") + return False + + try: + cudnn_frontend_version = [int(x) for x in cudnn_frontend_version_split] + except ValueError: + log.debug(f"Unable to parse cuDNN frontend version as an int list: {cudnn.__version__}.") + return False + + if cudnn_frontend_version < CUDNN_MIN_FRONTEND_VERSION: + log.debug( + "cuDNN Attention is not supported due to insufficient cuDNN frontend version " + f"{cudnn_frontend_version=}, expected at least {CUDNN_MIN_FRONTEND_VERSION=}." + ) + return False + + return True + + +CUDNN_SUPPORTED = cudnn_supported() + + +if CUDNN_SUPPORTED: + from cosmos_policy._src.imaginaire.attention.cudnn.functions import cudnn_attention + +else: + from cosmos_policy._src.imaginaire.attention.cudnn.stubs import cudnn_attention + +__all__ = ["cudnn_attention", "CUDNN_SUPPORTED"] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/checks.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/checks.py new file mode 100644 index 0000000000000000000000000000000000000000..ec3dd5fa5e2333251b3733eedf5d8b0e072d437e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/checks.py @@ -0,0 +1,128 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +cudNN backend checks +""" + +from functools import partial + +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.checks import attention_param_checks, attention_tensor_checks +from cosmos_policy._src.imaginaire.attention.cudnn import CUDNN_DISALLOWED, CUDNN_SUPPORTED +from cosmos_policy._src.imaginaire.attention.cudnn.meta import get_bwd_dtypes, get_fwd_dtypes +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.utils import get_arch_tag, is_torch_compiling, log_or_raise_error + + +def cudnn_attention_check( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool, + causal_type: CausalType, + is_varlen: bool, + raise_error: bool = False, +) -> bool: + """ + Input validation function for the cuDNN backend. + Runs the common and cuDNN-specific checks. Returns False if any checks fail, otherwise True. + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`). + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`). + + is_causal (bool): whether or not causal masking is enabled. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`. Required when `is_causal = True`. + + is_varlen (bool): whether or not a variable length (varlen) use case. Must be inferred + beforehand based on arguments such as seqlens_{Q,KV} or cumulative_seqlen_{Q,KV} being + passed. + + raise_error (bool): whether to raise an error if any checks fail or no backend is selected, + instead of just returning False. Default is False. + + Returns: + success (bool): whether use case is compatible with cuDNN backend. + + """ + target_fn = partial(log_or_raise_error, raise_error=raise_error) + + if not CUDNN_SUPPORTED: + target_fn( + "cuDNN is not supported in this environment. Run with debug logs to find out why, or choose another backend.", + exception=RuntimeError, + ) + return False + + if CUDNN_DISALLOWED: + target_fn("cuDNN backend is disabled. Please choose another backend.", exception=RuntimeError) + return False + + if is_torch_compiling(): + target_fn( + "cuDNN backend does not support torch.compile yet.", + exception=RuntimeError, + ) + return False + + arch_tag = get_arch_tag(query.device) + fwd_dtypes = get_fwd_dtypes(arch_tag) + bwd_dtypes = get_bwd_dtypes(arch_tag) + if not attention_tensor_checks( + query=query, + key=key, + value=value, + supported_dtypes_forward=fwd_dtypes, + supported_dtypes_backward=bwd_dtypes, + supports_mla=False, + supports_gqa_mqa=False, + raise_error=raise_error, + backend_name="cuDNN Attention", + ): + target_fn("cuDNN does not support the given inputs.", exception=RuntimeError) + return False + + if is_varlen: + target_fn("Varlen for cuDNN Attention is not integrated yet.", exception=RuntimeError) + return False + + # Verifies causal_type is a CausalType instance when is_causal + # Verifies DontCare is not used unless seqlen_q == seqlen_kv + attention_param_checks( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + ) + + if is_causal and causal_type not in [CausalType.TopLeft, CausalType.DontCare]: + target_fn("cuDNN Attention only supports top-left causal masking for now.", exception=RuntimeError) + return False + + return True diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/cudnn_forward.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/cudnn_forward.py new file mode 100644 index 0000000000000000000000000000000000000000..5b727d29ec220a901bbe5f7dba920046fab42d1f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/cudnn_forward.py @@ -0,0 +1,411 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +cuDNN Backend: intermediate APIs +Only safe to import when CUDNN_SUPPORTED is True. +""" + +from functools import lru_cache +from typing import Callable + +import cudnn +import torch +from torch import Size, Tensor + +from cosmos_policy._src.imaginaire.attention.utils import get_arch_tag +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +# Force using padded mask as a potential workaround for failing use cases +FORCE_PADDED_MASK = False + +CUDNN_GRAPH_CACHE_SIZE = 64 + +log.debug(f"cuDNN Attention graphs are cached using an LRU cache with capacity {CUDNN_GRAPH_CACHE_SIZE}.") +log.debug(f"cuDNN Attention {FORCE_PADDED_MASK=}.") + + +def get_dtype_choices(arch_tag: int) -> dict: + """ + Returns data type choices according to arch tag (attention.utils.get_arch_tag). + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + data_type_choices (dict): a map from PyTorch data types to cuDNN data types. Empty if device + is not supported. + + """ + + if arch_tag < 80: + log.debug("cuDNN Attention is not supported because compute capability is below the minimum (8.0).") + return {} + + ## NOTE (ahassani): As of version 91400 FP8 inference via the python frontend does + ## not seem to work. + # if arch_tag in [90, 100]: + # log.debug(f"cuDNN Attention supports FP8 for {arch_tag=}.") + # return { + # torch.float16: cudnn.data_type.HALF, + # torch.bfloat16: cudnn.data_type.BFLOAT16, + # torch.float8_e4m3fn: cudnn.data_type.FP8_E4M3, + # torch.float8_e5m2: cudnn.data_type.FP8_E5M2, + # } + + log.debug(f"cuDNN Attention only supports FP16 and BF16 for {arch_tag=}.") + return { + torch.float16: cudnn.data_type.HALF, + torch.bfloat16: cudnn.data_type.BFLOAT16, + } + + +def cudnn_sdpa_fwd_generate_operands( + q: Tensor, k: Tensor, v: Tensor, num_heads: int, return_lse: bool = False +) -> tuple[Tensor, Tensor, Tensor, Tensor, Tensor | None]: + """ + Takes torch input operands (Q, K, V), validates them, and returns views compatible with cuDNN + APIs ("strided" view with heads-first logical layout but heads-last physical layout. + + NOTE: this operation tries to specifically avoid memory copies and express everything as tensor + views, therefore it is crucial to not manipulate the outputs in __any way__ after this point, + and directly call cuDNN SDPA operations on it. + This is also what makes this operation very efficient and low in overhead, as there are no + device/CUDA operations or barriers with host/CPU. + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`) + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`) + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`) + + num_heads (int): Number of attention heads. Used for layout validation. + + Other Parameters: + return_lse (bool): Whether to store and return the logsumexp values. Default is False. + + Returns: + query_cudnn_layout (Tensor): 4-D query tensor, with the cuDNN strided layout + (`[batch, heads, seqlen, head_dim]`). + + key_cudnn_layout (Tensor): 4-D key tensor, with the cuDNN strided layout + (`[batch, heads_kv, seqlen_kv, head_dim]`). + + value_cudnn_layout (Tensor): 4-D output tensor, with the cuDNN strided layout + (`[batch, heads_kv, seqlen_kv, head_dim_v]`). + + output_cudnn_layout (Tensor): 4-D output tensor, with the cuDNN strided layout + (`[batch, heads, seqlen, head_dim_v]`). + + logsumexp_cudnn_layout (Tensor | None): only returned when return_lse is True. logsumexp + tensor, with the cuDNN strided layout (`[batch, heads, seqlen, 1]`). + """ + + if q.shape[0] != k.shape[0] or q.shape[0] != v.shape[0]: + raise ValueError( + f"All attention operands must match in batch size, got {q.shape[0]=}, {k.shape[0]=}, {v.shape[0]=}." + ) + + if q.shape[-1] != k.shape[-1]: + raise ValueError(f"Query and key must match in head dim, got {q.shape[-1]=}, {k.shape[-1]=}.") + + if q.shape[-2] != num_heads: + raise ValueError( + f"The heads-last layout considers q.shape[-2] as number of heads, got {q.shape[-2]=} but {num_heads=}." + ) + + if k.shape[-2] != num_heads: + raise ValueError( + f"The heads-last layout considers k.shape[-2] as number of heads, got {k.shape[-2]=} but {num_heads=}." + ) + + if v.shape[-2] != num_heads: + raise ValueError( + f"The heads-last layout considers v.shape[-2] as number of heads, got {v.shape[-2]=} but {num_heads=}." + ) + + if not q.is_contiguous() or not k.is_contiguous() or not v.is_contiguous(): + raise ValueError( + "All attention operands must be contiguous, got " + f"{q.is_contiguous()=}, {k.is_contiguous()=}, {v.is_contiguous()=}." + ) + + if q.dtype != k.dtype or q.dtype != v.dtype: + raise ValueError(f"All attention operands must match in dtype, got {q.dtype=}, {k.dtype=}, {v.dtype=}.") + + if q.device != k.device or q.device != v.device: + raise ValueError( + f"All attention operands must be on the same device, got {q.device=}, {k.device=}, {v.device=}." + ) + + dtype = q.dtype + device = q.device + arch_tag = get_arch_tag(device) + dtype_choices = get_dtype_choices(arch_tag) + + if dtype not in dtype_choices: + raise ValueError(f"Data type {dtype} is not supported; choices are: {dtype_choices.keys()}.") + + if arch_tag < 80: + raise NotImplementedError(f"cuDNN Attention only supports SM80 and later, but {device=} is SM{arch_tag}.") + + batch, seqlen_q, _, head_dim_qk = q.shape + _, _, _, head_dim_v = v.shape + + output = torch.empty([batch, seqlen_q, num_heads, head_dim_v], dtype=dtype, device=device) + lse = None + if return_lse: + lse = torch.empty([batch, seqlen_q, num_heads, 1], dtype=dtype, device=device) + + q_cudnn_layout = q.permute(0, 2, 1, 3) + k_cudnn_layout = k.permute(0, 2, 1, 3) + v_cudnn_layout = v.permute(0, 2, 1, 3) + output_cudnn_layout = output.permute(0, 2, 1, 3) + lse_cudnn_layout = None + assert q_cudnn_layout.data_ptr() == q.data_ptr() + assert k_cudnn_layout.data_ptr() == k.data_ptr() + assert v_cudnn_layout.data_ptr() == v.data_ptr() + assert output_cudnn_layout.data_ptr() == output.data_ptr() + + if return_lse: + lse_cudnn_layout = lse.permute(0, 2, 1, 3) + assert lse_cudnn_layout.data_ptr() == lse.data_ptr() + + return q_cudnn_layout, k_cudnn_layout, v_cudnn_layout, output_cudnn_layout, lse_cudnn_layout + + +@lru_cache(maxsize=CUDNN_GRAPH_CACHE_SIZE) +def cudnn_sdpa_fwd_generate_op( + dtype: torch.dtype, + device: torch.device, + q_shape: Size, + q_stride: Size, + k_shape: Size, + k_stride: Size, + v_shape: Size, + v_stride: Size, + output_shape: Size, + output_stride: Size, + lse_shape: Size | None = None, + lse_stride: Size | None = None, + is_causal: bool = False, + attn_scale: float | None = None, + seqlen_Q: int | None = None, + seqlen_KV: int | None = None, +) -> Callable: + """ + Takes use case metadata that has been validated and generated by cudnn_sdpa_fwd_generate_operands + and returns a callable cuDNN SDPA forward operation. + This function does NOT perform attention and rather prepares and builds the cuDNN graph + responsible for doing so. + + The final callable that it returns will take any q, k, v, output and (optionally) lse tensors + matching the same attributes (shape, device, dtype, etc) and call cuDNN SDPA forward on them. + + Parameters: + dtype (torch.dtype): Tensor data type for Q, K, V, and output. + + device (torch.device): Torch (CUDA) device where tensors are and where Attention will run. + + q_shape (Size): The shape of the 4-D query tensor with the cuDNN strided layout + (`[batch, heads, seqlen, head_dim]`). + + q_stride (Size): The stride of the 4-D query tensor with the cuDNN strided layout. + + k_shape (Size): The shape of the 4-D key tensor with the cuDNN strided layout + (`[batch, heads_kv, seqlen_kv, head_dim]`). + + k_stride (Size): The stride of the 4-D key tensor with the cuDNN strided layout. + + v_shape (Size): The shape of the 4-D value tensor with the cuDNN strided layout + (`[batch, heads_kv, seqlen_kv, head_dim_v]`). + + v_stride (Size): The stride of the 4-D value tensor with the cuDNN strided layout. + + output_shape (Size): The shape of the 4-D output tensor with the cuDNN strided layout + (`[batch, heads, seqlen, head_dim_v]`). + + output_stride (Size): The stride of the 4-D output tensor with the cuDNN strided layout. + + lse_shape (Size | None): The shape of the 4-D logsumexp tensor with the cuDNN strided + layout (`[batch, heads, seqlen, 1]`). + + lse_stride (Size | None): The stride of the 4-D logsumexp tensor with the cuDNN strided + layout. + + Other Parameters: + is_causal (bool): whether or not causal masking is enabled. Default is False. + + attn_scale (float | None): Dot product scale (attention scale). Defaults to + head_dim ** -0.5. + + Returns: + cudnn_sdpa_forward_exec (Callable): Function executing the cuDNN graph with the SDPA + forward operation. Function signature: + + query_cudnn_layout (Tensor): 4-D query tensor, with the cuDNN strided layout + (`[batch, heads, seqlen, head_dim]`). + + key_cudnn_layout (Tensor): 4-D key tensor, with the cuDNN strided layout + (`[batch, heads_kv, seqlen_kv, head_dim]`). + + value_cudnn_layout (Tensor): 4-D output tensor, with the cuDNN strided layout + (`[batch, heads_kv, seqlen_kv, head_dim_v]`). + + output_cudnn_layout (Tensor): 4-D output tensor, with the cuDNN strided layout + (`[batch, heads, seqlen, head_dim_v]`). + + logsumexp_cudnn_layout (Tensor | None): Optional logsumexp tensor, with the + cuDNN strided layout (`[batch, heads, seqlen, 1]`). + """ + + attn_scale = attn_scale if attn_scale is not None else q_shape[-1] ** -0.5 + + arch_tag = get_arch_tag(device) + dtype_choices = get_dtype_choices(arch_tag) + + assert dtype in dtype_choices + cudnn_dtype = dtype_choices[dtype] + + graph = cudnn.pygraph( + io_data_type=cudnn_dtype, + intermediate_data_type=cudnn.data_type.FLOAT, + compute_data_type=cudnn.data_type.FLOAT, + ) + + q_cudnn = graph.tensor(dim=q_shape, stride=q_stride, data_type=cudnn_dtype) + k_cudnn = graph.tensor(dim=k_shape, stride=k_stride, data_type=cudnn_dtype) + v_cudnn = graph.tensor(dim=v_shape, stride=v_stride, data_type=cudnn_dtype) + + assert (lse_shape is None and lse_stride is None) or (lse_shape is not None and lse_stride is not None) + generate_stats = lse_shape is not None + + seqlen_q_cudnn = None + seqlen_kv_cudnn = None + use_padding_mask = FORCE_PADDED_MASK or seqlen_Q is not None or seqlen_KV is not None + if use_padding_mask: + seqlen_Q = seqlen_Q if seqlen_Q is not None else q_shape[2] + seqlen_KV = seqlen_KV if seqlen_KV is not None else k_shape[2] + + seqlen_q_cudnn = graph.tensor(dim=[q_shape[0], 1, 1, 1], stride=[1, 1, 1, 1], data_type=cudnn.data_type.INT32) + seqlen_kv_cudnn = graph.tensor(dim=[k_shape[0], 1, 1, 1], stride=[1, 1, 1, 1], data_type=cudnn.data_type.INT32) + + o_cudnn, lse_cudnn = graph.sdpa( + q=q_cudnn, + k=k_cudnn, + v=v_cudnn, + generate_stats=generate_stats, + attn_scale=attn_scale, + use_causal_mask=is_causal, + use_padding_mask=use_padding_mask, + seq_len_q=seqlen_q_cudnn, + seq_len_kv=seqlen_kv_cudnn, + ) + + o_cudnn.set_output(True).set_data_type(cudnn_dtype).set_dim(output_shape).set_stride(output_stride) + if generate_stats: + lse_cudnn.set_output(True).set_dim(lse_shape).set_stride(lse_stride) + + graph.validate() + graph.build_operation_graph() + graph.create_execution_plans([cudnn.heur_mode.A, cudnn.heur_mode.FALLBACK]) + graph.check_support() + graph.build_plans() + + workspace_size_bytes = graph.get_workspace_size() + log.debug(f"Generated cuDNN Attention graph. Scratch space required: {workspace_size_bytes} bytes.") + + handle = cudnn.create_handle() + + def cudnn_operation(q: Tensor, k: Tensor, v: Tensor, output: Tensor, lse: Tensor | None = None): + # NOTE: This is INCREDIBLY important to do -- this is what wasted days of my time + # with random NaNs and illegal memory accesses and things of that nature. + stream = torch.cuda.current_stream(q.device) + cudnn.set_stream(handle=handle, stream=stream.cuda_stream) + + # caching allocator plays nicely with the LRU cache over this, but for now let's avoid + # premature optimization. + workspace = torch.zeros(workspace_size_bytes, device=device, dtype=torch.uint8) + + variant_pack = { + q_cudnn: q, + k_cudnn: k, + v_cudnn: v, + o_cudnn: output, + } + + if use_padding_mask: + batch = k.shape[0] + seqlen_q_cu = torch.tensor([seqlen_Q for _ in range(batch)]).to(device).reshape(batch, 1, 1, 1) + seqlen_kv_cu = torch.tensor([seqlen_KV for _ in range(batch)]).to(device).reshape(batch, 1, 1, 1) + log.debug(f"{q.shape=}, {k.shape=}, {seqlen_q_cu=}, {seqlen_kv_cu=}") + variant_pack[seqlen_q_cudnn] = seqlen_q_cu + variant_pack[seqlen_kv_cudnn] = seqlen_kv_cu + + if generate_stats: + assert lse is not None + assert lse_cudnn is not None + variant_pack[lse_cudnn] = lse + else: + assert lse is None and lse_cudnn is None + + log.debug("Generated cuDNN Attention graph executed") + return graph.execute(variant_pack, workspace, handle=handle) + + return cudnn_operation + + +def cudnn_sdpa_fwd_post_process( + output_cudnn_layout: Tensor, + lse_cudnn_layout: Tensor | None = None, +) -> tuple[Tensor, Tensor | None]: + """ + Takes torch tensor views validated and generated by cudnn_sdpa_fwd_generate_operands and + maps back to torch contiguous layout (heads-last both logical and physical). + It should be called after the cuDNN operation. + + Like cudnn_sdpa_fwd_generate_operands, this function is expected to be minimal overhead. + + Parameters: + output_cudnn_layout (Tensor): 4-D output tensor, with the cuDNN strided layout + (`[batch, heads, seqlen, head_dim_v]`). + + logsumexp_cudnn_layout (Tensor | None): Optional logsumexp tensor, with the cuDNN + strided layout (`[batch, heads, seqlen, 1]`). + + Returns: + output (Tensor): 4-D output tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim_v]`). + + logsumexp (Tensor | None): Optional logsumexp tensor, with the heads-last contiguous + layout (`[batch, seqlen, heads, 1]`). + """ + + output = output_cudnn_layout.permute(0, 2, 1, 3) + lse = None + assert output.data_ptr() == output_cudnn_layout.data_ptr() + + if lse_cudnn_layout is not None: + lse = lse_cudnn_layout.permute(0, 2, 1, 3) + assert lse.data_ptr() == lse_cudnn_layout.data_ptr() + + return output, lse diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/functions.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/functions.py new file mode 100644 index 0000000000000000000000000000000000000000..297862a322ebdb8f483fa6f46abd4027c4b474df --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/functions.py @@ -0,0 +1,262 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +cuDNN Backend: intermediate APIs +Only safe to import when CUDNN_SUPPORTED is True. +""" + +import time +from functools import partial + +import torch +from torch import Tensor +from torch.amp import custom_bwd, custom_fwd +from torch.autograd import Function + +from cosmos_policy._src.imaginaire.attention.cudnn.checks import cudnn_attention_check +from cosmos_policy._src.imaginaire.attention.cudnn.cudnn_forward import ( + cudnn_sdpa_fwd_generate_op, + cudnn_sdpa_fwd_generate_operands, + cudnn_sdpa_fwd_post_process, +) +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +amp_fwd = partial(custom_fwd, device_type="cuda") +amp_bwd = partial(custom_bwd, device_type="cuda") + + +CUDNN_PADDING_REQUIRED = False + + +class CudnnAttentionAutogradFn(Function): + @staticmethod + @amp_fwd + def forward( + ctx, + query: Tensor, + key: Tensor, + value: Tensor, + num_heads: int, + is_causal: bool, + scale: float, + ) -> tuple[Tensor, Tensor]: + query = query.contiguous() + key = key.contiguous() + value = value.contiguous() + + seqlen_Q = None + seqlen_KV = None + padding_Q = 0 + padding_KV = 0 + + # NOTE (ahassani): this may resolve some of the bugs caused by weird seqlens, + # but as of 11/12/2025 does not seem to fix any issues. Keeping here in case + # it ever comes back. + if CUDNN_PADDING_REQUIRED: + Q_multiplier = 256 + KV_multiplier = 256 + + if query.shape[1] % Q_multiplier != 0: + seqlen_Q = query.shape[1] + padding_Q = Q_multiplier - (seqlen_Q % Q_multiplier) + + old_shape = query.shape + query = torch.nn.functional.pad(query, (0, 0, 0, 0, 0, padding_Q), "constant", 0) + log.debug(f"cuDNN Attention: padded query from {old_shape} to {query.shape}.") + + if key.shape[1] % KV_multiplier != 0: + seqlen_KV = key.shape[1] + padding_KV = KV_multiplier - (seqlen_KV % KV_multiplier) + + old_shape = key.shape + key = torch.nn.functional.pad(key, (0, 0, 0, 0, 0, padding_KV), "constant", 0) + value = torch.nn.functional.pad(value, (0, 0, 0, 0, 0, padding_KV), "constant", 0) + log.debug(f"cuDNN Attention: padded KV from {old_shape} to {key.shape}.") + + # Transform operands to cuDNN-compatible layouts, make output tensors + (q_cudnn_layout, k_cudnn_layout, v_cudnn_layout, output_cudnn_layout, lse_cudnn_layout) = ( + cudnn_sdpa_fwd_generate_operands(q=query, k=key, v=value, num_heads=num_heads, return_lse=True) + ) + + # Construct graph + assert q_cudnn_layout.device == k_cudnn_layout.device == v_cudnn_layout.device == output_cudnn_layout.device + assert q_cudnn_layout.dtype == k_cudnn_layout.dtype == v_cudnn_layout.dtype == output_cudnn_layout.dtype + cudnn_graph_gen_start = time.time() * 1e3 + cudnn_sdpa = cudnn_sdpa_fwd_generate_op( + dtype=q_cudnn_layout.dtype, + device=q_cudnn_layout.device, + q_shape=q_cudnn_layout.shape, + q_stride=q_cudnn_layout.stride(), + k_shape=k_cudnn_layout.shape, + k_stride=k_cudnn_layout.stride(), + v_shape=v_cudnn_layout.shape, + v_stride=v_cudnn_layout.stride(), + output_shape=output_cudnn_layout.shape, + output_stride=output_cudnn_layout.stride(), + lse_shape=None if lse_cudnn_layout is None else lse_cudnn_layout.shape, + lse_stride=None if lse_cudnn_layout is None else lse_cudnn_layout.stride(), + is_causal=is_causal, + attn_scale=scale, + seqlen_Q=seqlen_Q, + seqlen_KV=seqlen_KV, + ) + cudnn_graph_gen_time = time.time() * 1e3 - cudnn_graph_gen_start + log.debug(f"cuDNN Attention forward graph generation took {cudnn_graph_gen_time:.1f} ms.") + + # Execute graph + cudnn_sdpa( + q=q_cudnn_layout, + k=k_cudnn_layout, + v=v_cudnn_layout, + output=output_cudnn_layout, + lse=lse_cudnn_layout, + ) + + # Transform outputs back to torch contiguous layouts + output, logsumexp = cudnn_sdpa_fwd_post_process( + output_cudnn_layout=output_cudnn_layout, + lse_cudnn_layout=lse_cudnn_layout, + ) + + ctx.save_for_backward(q_cudnn_layout, k_cudnn_layout, v_cudnn_layout, lse_cudnn_layout, output_cudnn_layout) + ctx.num_heads = num_heads + ctx.scale = scale + + if padding_Q > 0: + old_shape = output.shape + output = output[:, :seqlen_Q, :, :] + logsumexp = logsumexp[:, :seqlen_Q, :, :] + assert output.shape[1] == seqlen_Q + assert logsumexp.shape[1] == seqlen_Q + log.debug(f"cuDNN Attention: unpadded output from {old_shape} to {output.shape}.") + + return output, logsumexp + + @staticmethod + @amp_bwd + def backward( + ctx, grad_out: Tensor, grad_lse: Tensor + ) -> tuple[ + Tensor, + Tensor, + Tensor, + None, + None, + None, + ]: + raise NotImplementedError() + + +def cudnn_attention( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool = False, + causal_type: CausalType | None = None, + scale: float | None = None, + cumulative_seqlen_Q: Tensor | None = None, + cumulative_seqlen_KV: Tensor | None = None, + max_seqlen_Q: int | None = None, + max_seqlen_KV: int | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + """ + Runs cuDNN Attention on given operands (Q, K, V) with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`) + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`) + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`) + + is_causal (bool): whether or not causal masking is enabled. Default is False. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`. Required when `is_causal = True`. + + scale (float | None): Dot product scale (attention scale). Defaults to head_dim ** -0.5. + + cumulative_seqlen_Q (Tensor | None): (varlen) Optional 1-D tensor with size `batch + 1` + indicating the cumulative sum of number of query tokens in each batch, with an + additional 0 element in the beginning. Must be passed together with + `cumulative_seqlen_KV` and `max_seqlen_{Q,KV}`. + + cumulative_seqlen_KV (Tensor | None): (varlen) Optional 1-D tensor with size `batch + 1` + indicating the cumulative sum of number of key/value tokens in each batch, with an + additional 0 element in the beginning. Must be passed together with + `cumulative_seqlen_Q` and `max_seqlen_{Q,KV}`. + + max_seqlen_Q (int | None): (varlen) Optional integer indicating the maximum query + sequence length in all batches. Must be passed together with `cumulative_seqlen_{Q,KV}` + and `max_seqlen_KV`. + + max_seqlen_KV (int | None): (varlen) Optional integer indicating the maximum key/value + sequence length in all batches. Must be passed together with `cumulative_seqlen_{Q,KV}` + and `max_seqlen_Q`. + + Other Parameters: + return_lse (bool): Whether to return the logsumexp values. Default is False. + + backend_kwargs (dict | None): Key-value pair for passing arguments specific to cuDNN's + attention operator, if any. + + Returns: + output (Tensor): 4-D output tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim_v]`). + + logsumexp (Tensor): logsumexp tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, 1]`). Only returned when return_lse is True. + """ + + is_varlen = cumulative_seqlen_Q is not None + assert cudnn_attention_check( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + is_varlen=is_varlen, + raise_error=True, + ) + + assert not is_varlen # cudnn_attention_check should prevent this assertion failing + + num_heads = query.shape[-2] + scale = scale if scale is not None else query.shape[-1] ** -0.5 + + output, lse = CudnnAttentionAutogradFn.apply( + query, + key, + value, + num_heads, + is_causal, + scale, + ) + + if return_lse: + return output, lse + + return output diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/meta.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/meta.py new file mode 100644 index 0000000000000000000000000000000000000000..7e966e13be22b738d9e9c455385f1e8c4f255aa0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/meta.py @@ -0,0 +1,63 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +cuDNN Backend: metadata +Always safe to import (as long as torch is available.) +""" + +import torch + +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + + +def get_fwd_dtypes(arch_tag: int) -> list[torch.dtype]: + """ + Returns data type choices for forward pass according to arch tag (attention.utils.get_arch_tag). + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + data_type_choices (list): a list of PyTorch data types. Empty if device is not supported. + + """ + + if arch_tag < 80: + log.debug("cuDNN Attention is not supported because compute capability is below the minimum (8.0).") + return [] + + ## NOTE (ahassani): As of version 91400 FP8 inference via the python frontend does + ## not seem to work. + log.debug(f"cuDNN Attention only supports FP16 and BF16 for {arch_tag=}.") + return [torch.float16, torch.bfloat16] + + +def get_bwd_dtypes(arch_tag: int) -> list[torch.dtype]: + """ + Returns data type choices for backward pass according to arch tag (attention.utils.get_arch_tag). + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + data_type_choices (list): a list of PyTorch data types. Empty if device is not supported. + + """ + + return [] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/stubs.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/stubs.py new file mode 100644 index 0000000000000000000000000000000000000000..835a8f8736c3a651b83831bd65210031d67c847a --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/cudnn/stubs.py @@ -0,0 +1,46 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +cuDNN Backend: intermediate API stubs +Always safe to import (as long as torch is available.) +""" + +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.masks import CausalType + + +def cudnn_attention( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool = False, + causal_type: CausalType | None = None, + scale: float | None = None, + cumulative_seqlen_Q: Tensor | None = None, + cumulative_seqlen_KV: Tensor | None = None, + max_seqlen_Q: int | None = None, + max_seqlen_KV: int | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + raise RuntimeError( + "Tried to run cuDNN attention, but it is not supported / available. " + "Try running with debug logs enabled to see why." + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/README.md b/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/README.md new file mode 100644 index 0000000000000000000000000000000000000000..5e176b9a93209cb166e1f0b4b431f58d81983247 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/README.md @@ -0,0 +1,10 @@ +# Imaginaire Attention Subpackage Docs + +* [Basic API & Intro](../README.md) +* Docs (you are here) + * [Backends](backends.md) + * Features + * [Basic features](features.md) + * [Multi-dimensional Attention](multi-dim.md) + * [Spatio-Temporal Attention](multi-dim.md#spatio-temporal-attention) + * [APIs](apis.md) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/apis.md b/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/apis.md new file mode 100644 index 0000000000000000000000000000000000000000..35ed69710af105bd9dbd832ad1544c956e5b95f6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/apis.md @@ -0,0 +1,28 @@ +# Imaginaire Attention Subpackage Docs > APIs + +## Attention + +::: cosmos_policy._src.imaginaire.attention + options: + heading_level: 3 + show_object_full_path: true + members: + - attention + +## Multi-Dimensional Attention + +::: cosmos_policy._src.imaginaire.attention + options: + heading_level: 3 + show_object_full_path: true + members: + - multi_dimensional_attention + +### Spatio-Temporal Attention + +::: cosmos_policy._src.imaginaire.attention + options: + heading_level: 3 + show_object_full_path: true + members: + - spatio_temporal_attention diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/backends.md b/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/backends.md new file mode 100644 index 0000000000000000000000000000000000000000..659c77de23311355b0dd7c5c9203ef33b0f1784d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/backends.md @@ -0,0 +1,73 @@ +# Imaginaire Attention Subpackage Docs > Backends + +The goal is to support as many stable and reliable backends as possible, both for feature coverage, +and for delivering the best performance. + +## NATTEN +[NATTEN](https://natten.org) ships standard Attention kernels in addition to sparse / +multi-dimensional kernels. + +Minimum version required: `0.21.5.dev3`. + +### Feature coverage + +| Feat/Backend | Ampere/RTX | Hopper | Blackwell | +|--------------|--------------------|--------|--------------------| +| Causal mask | :white_check_mark: | | :white_check_mark: | +| Varlen | :white_check_mark: | | :white_check_mark: | +| GQA/MQA | | | :white_check_mark: | +| MLA | :white_check_mark: | | | + +This backend supports torch compile. + +## Flash Attention v2 + +Flash Attention v2 (original C++ kernels) are available under the `flash2` backend. +Requires the `flash_attn` package. + +Minimum version required: `2.7.0`. +Maximum version supported: `2.7.4`. + +This backend supports torch compile. + +### Feature coverage + +| Feat/Backend | Ampere/RTX | +|--------------|--------------------| +| Causal mask | :white_check_mark: | +| Varlen | :white_check_mark: | +| GQA/MQA | :white_check_mark: | +| MLA | | + +## Flash Attention v3 + +Flash Attention v3 (original C++ kernels) are available under the `flash3` backend. +Requires the `flash_attn_3` package. + +Version required: `3.0.0.b*`. + +### Feature coverage + +| Feat/Backend | Ampere/RTX | +|--------------|--------------------| +| Causal mask | :white_check_mark: | +| Varlen | :white_check_mark: | +| GQA/MQA | :white_check_mark: | +| MLA | | + +MLA is technically supported, but disabled due to an API bug in the backward pass. + +Torch compile is NOT yet supported for this backend. + +## cuDNN + +**NOTE**: due to numerical instability on Blackwell, this backend is not yet fully integrated, and +is banned for all use cases. + +Minimum version required: python frontend: `1.14.0`, backend: `91300`. + +Torch compile is NOT yet supported for this backend. + +## Future backends + +We plan to add Flash Attention 4 (CuTeDSL kernels) and any other relevant backends. diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/features.md b/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/features.md new file mode 100644 index 0000000000000000000000000000000000000000..6ebb44447c8d0b70c8c505a822d4d68db8a9c9d4 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/features.md @@ -0,0 +1,151 @@ +# Imaginaire Attention Subpackage Docs > Features + +## Causal mask + +Causal masking requires explicit indication of causal mask type. +For example, simply passing `is_causal=True` will fail: + +```python +output = attention( + query=query, + key=key, + value=value, + is_causal=True +) +``` + +Result: +``` +ValueError: Argument causal_type must be specified when is_causal=True. +``` + +There are currently two types of causal masking that are supported, and many popular backends tend +to support only one. It's therefore critical to to choose the correct one for your application. + + +```python +from cosmos_policy._src.imaginaire.attention.masks import CausalType + +# Causal type choices: +# - CausalType.TopLeft +# - CausalType.BottomRight + +output = attention( + query=query, + key=key, + value=value, + is_causal=True, + causal_type=CausalType.TopLeft, +) +``` + +### Top-left causal mask + +Q sequence length = KV sequence length = 5 + +| | K1 | K2 | K3 | K4 | K5 | +|----|-----------|-----------|-----------|-----------|-----------| +| Q1 | ✓ | ✗ | ✗ | ✗ | ✗ | +| Q2 | ✓ | ✓ | ✗ | ✗ | ✗ | +| Q3 | ✓ | ✓ | ✓ | ✗ | ✗ | +| Q4 | ✓ | ✓ | ✓ | ✓ | ✗ | +| Q5 | ✓ | ✓ | ✓ | ✓ | ✓ | + +Q sequence length = 2, KV sequence length = 5 + +| | K1 | K2 | K3 | K4 | K5 | +|----|-----------|-----------|-----------|-----------|-----------| +| Q1 | ✓ | ✗ | ✗ | ✗ | ✗ | +| Q2 | ✓ | ✓ | ✗ | ✗ | ✗ | + +Q sequence length = 5, KV sequence length = 2 + +| | K1 | K2 | +|----|-----------|-----------| +| Q1 | ✓ | ✗ | +| Q2 | ✓ | ✓ | +| Q3 | ✓ | ✓ | +| Q4 | ✓ | ✓ | +| Q5 | ✓ | ✓ | + +### Bottom-right causal mask + +Q sequence length = KV sequence length = 5 + +| | K1 | K2 | K3 | K4 | K5 | +|----|-----------|-----------|-----------|-----------|-----------| +| Q1 | ✓ | ✗ | ✗ | ✗ | ✗ | +| Q2 | ✓ | ✓ | ✗ | ✗ | ✗ | +| Q3 | ✓ | ✓ | ✓ | ✗ | ✗ | +| Q4 | ✓ | ✓ | ✓ | ✓ | ✗ | +| Q5 | ✓ | ✓ | ✓ | ✓ | ✓ | + +(identical to top-left in this special case) + +Q sequence length = 2, KV sequence length = 5 + +| | K1 | K2 | K3 | K4 | K5 | +|----|-----------|-----------|-----------|-----------|-----------| +| Q1 | ✓ | ✓ | ✓ | ✓ | ✗ | +| Q2 | ✓ | ✓ | ✓ | ✓ | ✓ | + +Q sequence length = 5, KV sequence length = 2 + +| | K1 | K2 | +|----|-----------|-----------| +| Q1 | ✗ | ✗ | +| Q2 | ✗ | ✗ | +| Q3 | ✗ | ✗ | +| Q4 | ✓ | ✗ | +| Q5 | ✓ | ✓ | + +## GQA/MQA + +Simply pass `key` and `value` without repeating attention heads. + +**NOTE**: `key`/`value` heads must evenly divide `query` heads. + +**NOTE**: the behavior is similar to `repeat_interleave`, not `repeat`. + +## Variable length + +**(Less efficient option)** Pass sequence lengths directly: + +```python +output = attention( + query=query, + key=key, + value=value, + seqlens_Q=torch.tensor(sequence_length_list_Q, device=query.device), + seqlens_KV=torch.tensor(sequence_length_list_KV, device=query.device), +) +``` + +This will manually compute the maximum sequence lengths, and cumulative sums (with the additional +padding). + +**(More efficient option)** Compute cumulative sequence lengths and maximums once, and reuse it: + +```python +from cosmos_policy._src.imaginaire.attention.varlen import generate_varlen_parameters + +# NOTE: query, key, and value are only used for verification, so it doesn't matter what model layer +# they correspond to. +( + cumulative_seqlen_Q, + cumulative_seqlen_KV, + max_seqlen_Q, + max_seqlen_KV, +) = generate_varlen_parameters(query, key, value, seqlens_Q, seqlens_KV) + +# in all attention layers that follow: +output = attention( + query=query, + key=key, + value=value, + cumulative_seqlen_Q=cumulative_seqlen_Q, + cumulative_seqlen_KV=cumulative_seqlen_KV, + max_seqlen_Q=max_seqlen_Q, + max_seqlen_KV=max_seqlen_KV, +) +``` diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/multi-dim.md b/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/multi-dim.md new file mode 100644 index 0000000000000000000000000000000000000000..c8d3177d84a69948d95bbdc6295b2d7562e27377 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/docs/multi-dim.md @@ -0,0 +1,111 @@ +# Imaginaire Attention Subpackage Docs > Features > Multi-Dimensional Attention + +Multi-Dimensional Attention is the primary API for handling various complex masks and sparsity +patterns, such as the spatio-temporal mask, and sliding window attention. + +## Basic API + +```python +from cosmos_policy._src.imaginaire.attention import multi_dimensional_attention + +output = multi_dimensional_attention( + query=query, + key=key, + value=value, +) +``` + +Sparsity parameters: +* **Optional** `window_size`: allows reducing the attention span by limiting each token's context to + a local sliding window. References: + * [Image Transformer](https://arxiv.org/abs/1802.05751) + * [Stand-alone self-attention](https://arxiv.org/abs/1906.05909) + * [Neighborhood attention transformer](https://arxiv.org/abs/2204.07143) +* **Optional** `dilation`: introduces gaps between the tokens within a sliding window, capturing + global context without more computation. + Reference: [Dilated neighborhood attention transformer](https://arxiv.org/abs/2209.15001) + +Other masking parameters: +* **Optional** `stride`: introduces delays into the sliding window, for __potential__ efficiency + gains. Reference: [Generalized Neighborhood Attention](https://arxiv.org/abs/2504.16922). +* **Optional** `is_causal`: allows causally masking individual dimensions. This parameter can + implement the spatio-temporal mask (causal masking across temporal dimension, bi-directional + along space). + +All sparsity / masking parameters can be specified **per dimension**. +The key feature of `multi_dimensional_attention` over the standard `attention` API is supporting +multi-dimensional layouts of tokens (i.e. multi-dimensional feature maps). + +This means `query`, `key` and `value` are not necessarily 4-D tensors; they can be 4-D, 5-D, or 6-D, +representing 1-D, 2-D, and 3-D token layouts (see [Tensor layouts](#tensor-layouts)). + +* **Optional** `scale`: attention (softmax/dot product) scale. Defaults to `head_dim ** -0.5`. +* **Optional** `return_lse`: returns logsumexp if `True` +* **Optional** `backend`: explicitly set backend instead of automatically selecting the best compatible + +## Tensor layouts + +In addition to requiring the [contiguous heads-last tensor layout](../README.md#tensor-layouts), +Multi-Dimensional Attention also requires the "sequence length" dimension to be unrolled / unfolded +back into its original representation: + +```python +# 1-D case: language, audio +batch, X, heads, head_dim = query_1d.shape +# _ +# ^ +# | +# |-----> token layout shape + +# 2-D case: images +batch, X, Y, heads, head_dim = query_2d.shape +# ____ +# ^ +# | +# |-----> token layout shape + +# 3-D case: videos / 3-D images +batch, X, Y, Z, heads, head_dim = query_3d.shape +# _______ +# ^ +# | +# |------> token layout shape +``` + +Multi-Dimensional Attention also requires the shapes of `query`, `key` and `value` to match along +those dimensions, henceforth called the **token layout shape**: + +```python +assert query_1d.shape[1:2] == key_1d.shape[1:2] == value_1d.shape[1:2] + +assert query_2d.shape[1:3] == key_2d.shape[1:3] == value_2d.shape[1:3] + +assert query_3d.shape[1:4] == key_3d.shape[1:4] == value_3d.shape[1:4] +``` + +This is because of the large number of sparsity / masking features (and their combinations) +supported, which is mainly possible by making the assumption that query and context coordinate +spaces are the same, eliminating the requirement for a mapping between the two. + +Problems with a different query and key/value token layout shape may be supported in the future. + + +## Backends +The only backend supporting multi-dimensional attention for now is `natten`. + +## Spatio-Temporal Attention + +Spatio-Temporal attention (causal masking across the time dimension, and no masking / bi-directional +across spatial dimensions) is a special case of Multi-Dimensional Attention. +You can either implement it by marking `is_causal` as expected in `multi_dimensional_attention`, or +directly use `spatio_temporal_attention`: + +```python +from cosmos_policy._src.imaginaire.attention import spatio_temporal_attention + +output = spatio_temporal_attention( + query=query, + key=key, + value=value, +) +``` diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/README.md b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/README.md new file mode 100644 index 0000000000000000000000000000000000000000..6982dd56a97425129402c86a8a751d949ea44583 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/README.md @@ -0,0 +1,71 @@ +## Causal Mask + +NOTE: Flash only implements bottom-right-aligned causal mask, but the default +in SDPA, CUTLASS/NATTEN, cuDNN is top-left. +To get the same behavior, we __might__ be able to implement top-left-aligned with +the sliding window argument, but some of Flash's overrides prevent this... + +```python +seqlen_q = query.shape[1] +seqlen_k = key.shape[1] + +# From Flash Attn readme: +# Query at position i will only attend to keys between +# [i + seqlen_k - seqlen_q - window_size[0], i + seqlen_k - seqlen_q + window_size[1]] inclusive. +# +# so our window_size when doing top-left causal masking should satisfy: +# i + seqlen_k - seqlen_q - window_size[0] = 0 +# i + seqlen_k - seqlen_q + window_size[1] = i +# => +# seqlen_k - seqlen_q + window_size[1] = 0 ==> +# window_size[1] = seqlen_q - seqlen_k +# +# and: +# +# i + seqlen_k - seqlen_q = window_size[0] +# +# which has to be satisfied for all 0 <= i < seqlen_q: +# seqlen_k - seqlen_q = window_size[0] +# +# seqlen_q - 1 + seqlen_k - seqlen_q = window_size[0] ==> +# seqlen_k - 1 = window_size[0] +# +# which means ... +# +# +# Other Flash overrides: +# if (window_size_left >= seqlen_k) { window_size_left = -1; } +# if (window_size_right >= seqlen_k) { window_size_right = -1; } +# +# params.is_causal = window_size_left < 0 && window_size_right == 0; +# +# if (window_size_left < 0 && window_size_right >= 0) { window_size_left = seqlen_k; } +# if (window_size_left >= 0 && window_size_right < 0) { window_size_right = seqlen_k; } +# params.window_size_left = window_size_left; +# params.window_size_right = window_size_right; +# +# scheduler: +# n_block_min = max(0, (m_block * kBlockM + seqlen_k - seqlen_q - window_size_left) / kBlockN); +# n_block_max = min(n_block_max, ceil_div((m_block + 1) * kBlockM + seqlen_k - seqlen_q + window_size_right, kBlockN)); +# + +flash_causal = False +window_size = (-1, -1) if not is_causal else (seqlen_k, seqlen_q - seqlen_k) +if is_causal and seqlen_k < seqlen_q: + window_size = (-1, 0) + + padding_KV = seqlen_q - seqlen_k + old_shape = key.shape + key = torch.nn.functional.pad(key, (0, 0, 0, 0, 0, padding_KV), "constant", 0) + value = torch.nn.functional.pad(value, (0, 0, 0, 0, 0, padding_KV), "constant", 0) + log.debug(f"Flash Attention: padded KV from {old_shape} to {key.shape}.") + +print(f"{window_size=}") + +#window_size = (-1, -1) if not is_causal else (-1, 0) +#window_size = (-1, -1) if not is_causal else (-1, seqlen_q - seqlen_k) + +# seqlen_q=7688, seqlen_kv=2048, is_causal=True +# n_block_min = max(0, (q_start - 7688) / kBlockN); +# n_block_max = min(n_block_max, ceil_div(q_end, kBlockN)); +``` diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..1b9331ad7f93f66ca9de878ff3aaa12630a7694c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/__init__.py @@ -0,0 +1,96 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Flash Attention v2 (flash2) Backend +""" + +import torch + +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +# We lock to safe releases of Flash 2 +# We will have a separate backend identifier for 2025 releases with CuTeDSL +# kernels. +FLASH_ATTENTION_V2_MIN_VERSION = [2, 7, 0] +FLASH_ATTENTION_V2_MAX_VERSION = [2, 7, 4] + + +def flash2_supported() -> bool: + """ + Returns whether Flash Attention is supported in this environment. + Requirements are: + * Presence of CUDA Runtime (via PyTorch) + * Presence of Flash Attention, meeting minimum version requirements + + This check guards imports / dependencies on the Flash Attention package. + """ + if not torch.cuda.is_available(): + log.debug("Flash Attention v2 is not supported because PyTorch did not detect CUDA runtime.") + return False + + try: + import flash_attn + + except ImportError: + log.debug("Flash Attention v2 is not supported because the Python package was not found.") + return False + except Exception as e: + log.debug(f"Flash Attention v2 is not supported because importing the Python package failed: {e}") + return False + + flash2_version_str = None + if not hasattr(flash_attn, "__version__"): + from importlib.metadata import version + + flash2_version_str = version("flash_attn") + else: + flash2_version_str = flash_attn.__version__ + + flash2_version_split = flash2_version_str.split(".") + if len(flash2_version_split) < 3: + log.debug(f"Unable to parse Flash Attention v2 version {flash2_version_str}.") + return False + + try: + flash2_version = [int(x) for x in flash2_version_split[:3]] + + except ValueError: + log.debug(f"Unable to parse Flash Attention v2 version as an int list: {flash2_version_str}.") + return False + + if flash2_version > FLASH_ATTENTION_V2_MAX_VERSION or flash2_version < FLASH_ATTENTION_V2_MIN_VERSION: + log.debug( + "Flash Attention v2 build is not supported; this backend only supports versions " + f"{FLASH_ATTENTION_V2_MIN_VERSION} through {FLASH_ATTENTION_V2_MAX_VERSION}, got " + f"{flash2_version}." + ) + return False + + return True + + +FLASH2_SUPPORTED = flash2_supported() + +if FLASH2_SUPPORTED: + from cosmos_policy._src.imaginaire.attention.flash2.functions import flash2_attention + +else: + from cosmos_policy._src.imaginaire.attention.flash2.stubs import flash2_attention + +__all__ = ["flash2_attention", "FLASH2_SUPPORTED"] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/checks.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/checks.py new file mode 100644 index 0000000000000000000000000000000000000000..6dee75988b460d9ef0276048aec7e8786aabc4b9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/checks.py @@ -0,0 +1,112 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Flash Attention v2 (flash2) backend checks +""" + +from functools import partial + +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.checks import attention_param_checks, attention_tensor_checks +from cosmos_policy._src.imaginaire.attention.flash2 import FLASH2_SUPPORTED +from cosmos_policy._src.imaginaire.attention.flash2.meta import get_bwd_dtypes, get_fwd_dtypes +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.utils import get_arch_tag, log_or_raise_error + + +def flash2_attention_check( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool, + causal_type: CausalType, + is_varlen: bool, + raise_error: bool = False, +) -> bool: + """ + Input validation function for the flash2 backend. + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`). + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`). + + is_causal (bool): whether or not causal masking is enabled. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`. Required when `is_causal = True`. + + is_varlen (bool): whether or not a variable length (varlen) use case. Must be inferred + beforehand based on arguments such as seqlens_{Q,KV} or cumulative_seqlen_{Q,KV} being + passed. + + raise_error (bool): whether to raise an error if any checks fail or no backend is selected, + instead of just returning False. Default is False. + + Returns: + success (bool): whether use case is compatible with flash2 backend. + + """ + target_fn = partial(log_or_raise_error, raise_error=raise_error) + + if not FLASH2_SUPPORTED: + target_fn( + "Flash Attention v2 (flash2) is not supported in this environment. Run with debug logs to find out why, or choose another backend.", + exception=RuntimeError, + ) + return False + + arch_tag = get_arch_tag(query.device) + fwd_dtypes = get_fwd_dtypes(arch_tag) + bwd_dtypes = get_bwd_dtypes(arch_tag) + if not attention_tensor_checks( + query=query, + key=key, + value=value, + supported_dtypes_forward=fwd_dtypes, + supported_dtypes_backward=bwd_dtypes, + supports_mla=False, + supports_gqa_mqa=True, + raise_error=raise_error, + backend_name="Flash Attention v2 (flash2)", + ): + target_fn("Flash Attention v2 (flash2) does not support the given inputs.", exception=RuntimeError) + return False + + # Verifies causal_type is a CausalType instance when is_causal + # Verifies DontCare is not used unless seqlen_q == seqlen_kv + attention_param_checks( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + ) + + if is_causal and causal_type not in [CausalType.BottomRight, CausalType.DontCare]: + target_fn("Flash Attention only supports bottom-right causal masking.", exception=RuntimeError) + return False + + return True diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/functions.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/functions.py new file mode 100644 index 0000000000000000000000000000000000000000..5b3c75b256a249c3362711a10c1e9190404a2de4 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/functions.py @@ -0,0 +1,170 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Flash Attention v2 (flash2) Backend: intermediate APIs +Only safe to import when FLASH2_SUPPORTED is True. +""" + +from flash_attn.flash_attn_interface import flash_attn_func, flash_attn_varlen_func +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.flash2.checks import flash2_attention_check +from cosmos_policy._src.imaginaire.attention.masks import CausalType + + +def flash2_attention( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool = False, + causal_type: CausalType | None = None, + scale: float | None = None, + cumulative_seqlen_Q: Tensor | None = None, + cumulative_seqlen_KV: Tensor | None = None, + max_seqlen_Q: int | None = None, + max_seqlen_KV: int | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + """ + Runs Flash Attention v2 on given operands (Q, K, V) with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`) + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`) + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`) + + is_causal (bool): whether or not causal masking is enabled. Default is False. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`. Required when `is_causal = True`. + + scale (float | None): Dot product scale (attention scale). Defaults to head_dim ** -0.5. + + cumulative_seqlen_Q (Tensor | None): (varlen) Optional 1-D tensor with size `batch + 1` + indicating the cumulative sum of number of query tokens in each batch, with an + additional 0 element in the beginning. Must be passed together with + `cumulative_seqlen_KV` and `max_seqlen_{Q,KV}`. + + cumulative_seqlen_KV (Tensor | None): (varlen) Optional 1-D tensor with size `batch + 1` + indicating the cumulative sum of number of key/value tokens in each batch, with an + additional 0 element in the beginning. Must be passed together with + `cumulative_seqlen_Q` and `max_seqlen_{Q,KV}`. + + max_seqlen_Q (int | None): (varlen) Optional integer indicating the maximum query + sequence length in all batches. Must be passed together with `cumulative_seqlen_{Q,KV}` + and `max_seqlen_KV`. + + max_seqlen_KV (int | None): (varlen) Optional integer indicating the maximum key/value + sequence length in all batches. Must be passed together with `cumulative_seqlen_{Q,KV}` + and `max_seqlen_Q`. + + Other Parameters: + return_lse (bool): Whether to return the logsumexp values. Default is False. + + backend_kwargs (dict | None): Key-value pair for passing arguments specific to Flash's + attention operator, if any. + + Returns: + output (Tensor): 4-D output tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim_v]`). + + logsumexp (Tensor): logsumexp tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, 1]`). Only returned when return_lse is True. + """ + + is_varlen = cumulative_seqlen_Q is not None + assert flash2_attention_check( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + is_varlen=is_varlen, + raise_error=True, + ) + + scale = scale if scale is not None else query.shape[-1] ** -0.5 + + backend_kwargs = backend_kwargs if backend_kwargs is not None else {} + + if is_varlen: + assert query.shape[0] == key.shape[0] == value.shape[0] == 1 + q = query.squeeze(0) + k = key.squeeze(0) + v = value.squeeze(0) + assert q.dim() == k.dim() == v.dim() == 3 + out, lse_, _ = flash_attn_varlen_func( + q=query.squeeze(0), + k=key.squeeze(0), + v=value.squeeze(0), + cu_seqlens_q=cumulative_seqlen_Q, + cu_seqlens_k=cumulative_seqlen_KV, + max_seqlen_q=max_seqlen_Q, + max_seqlen_k=max_seqlen_KV, + softmax_scale=scale, + causal=is_causal, + return_attn_probs=True, + **backend_kwargs, + # window_size=(-1, -1), + # dropout_p=0.0, + # softcap=0.0, # 0.0 means deactivated + # alibi_slopes=None, + # deterministic=False, + # block_table=None, + ) + assert out.dim() == 3 + assert lse_.dim() == 2 + + output = out.unsqueeze(0) + lse = lse_.unsqueeze(0) + + else: + output, lse, _ = flash_attn_func( + q=query, + k=key, + v=value, + softmax_scale=scale, + causal=is_causal, + return_attn_probs=True, + **backend_kwargs, + # window_size=(-1, -1), + # dropout_p=0.0, + # softcap=0.0, # 0.0 means deactivated + # alibi_slopes=None, + # deterministic=False, + ) + + assert isinstance(output, Tensor) + assert isinstance(lse, Tensor) + assert output.dim() == 4 + assert lse.dim() == 3 + + lse = lse.permute(0, 2, 1).contiguous() # [batch, seqlen, head_dim] + + if return_lse: + return output, lse + + return output diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/meta.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/meta.py new file mode 100644 index 0000000000000000000000000000000000000000..9f6605344c3ed6ac1de159643fd534f84b48a5fe --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/meta.py @@ -0,0 +1,64 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Flash Attention v2 (flash2) Backend: metadata +Always safe to import (as long as torch is available.) +""" + +import torch + +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + + +def get_fwd_dtypes(arch_tag: int) -> list[torch.dtype]: + """ + Returns data type choices for forward pass according to arch tag (attention.utils.get_arch_tag). + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + data_type_choices (list): a list of PyTorch data types. Empty if device is not supported. + + """ + + if arch_tag < 80: + log.debug("Flash Attention v2 (flash2) is not supported because compute capability is below the minimum (8.0).") + return [] + + return [torch.float16, torch.bfloat16] + + +def get_bwd_dtypes(arch_tag: int) -> list[torch.dtype]: + """ + Returns data type choices for backward pass according to arch tag (attention.utils.get_arch_tag). + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + data_type_choices (list): a list of PyTorch data types. Empty if device is not supported. + + """ + + if arch_tag < 80: + log.debug("Flash Attention v2 (flash2) is not supported because compute capability is below the minimum (8.0).") + return [] + + return [torch.float16, torch.bfloat16] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/stubs.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/stubs.py new file mode 100644 index 0000000000000000000000000000000000000000..45111b62ac95f5b017f290c2118a5c53ae038a67 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash2/stubs.py @@ -0,0 +1,46 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Flash Attention v2 (flash2) Backend: intermediate API stubs +Always safe to import (as long as torch is available.) +""" + +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.masks import CausalType + + +def flash2_attention( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool = False, + causal_type: CausalType | None = None, + scale: float | None = None, + cumulative_seqlen_Q: Tensor | None = None, + cumulative_seqlen_KV: Tensor | None = None, + max_seqlen_Q: int | None = None, + max_seqlen_KV: int | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + raise RuntimeError( + "Tried to run Flash Attention v2, but it is not supported / available. " + "Try running with debug logs enabled to see why." + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..592f096f0ad56dcda645763046fdbc8801d3ac37 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/__init__.py @@ -0,0 +1,94 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Flash Attention v3 (flash3) Backend +""" + +import torch + +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +FLASH_ATTENTION_V3_MIN_VERSION = [3, 0, 0, 0] +FLASH_ATTENTION_V3_MAX_VERSION = [3, 0, 0, 1] + + +def flash3_supported() -> bool: + """ + Returns whether Flash Attention is supported in this environment. + Requirements are: + * Presence of CUDA Runtime (via PyTorch) + * Presence of Flash Attention, meeting minimum version requirements + + This check guards imports / dependencies on the Flash Attention package. + """ + if not torch.cuda.is_available(): + log.debug("Flash Attention v3 is not supported because PyTorch did not detect CUDA runtime.") + return False + + try: + import flash_attn_3 + + except ImportError: + log.debug("Flash Attention v3 is not supported because the Python package was not found.") + return False + except Exception as e: + log.debug(f"Flash Attention v3 is not supported because importing the Python package failed: {e}") + return False + + flash3_version_str = None + if not hasattr(flash_attn_3, "__version__"): + from importlib.metadata import version + + flash3_version_str = version("flash_attn_3") + else: + flash3_version_str = flash_attn_3.__version__ + + flash3_version_split = flash3_version_str.replace("b", ".").split(".") + if len(flash3_version_split) != 4: + log.debug(f"Unable to parse Flash Attention v3 version {flash3_version_str}.") + return False + + try: + flash3_version = [int(x) for x in flash3_version_split] + + except ValueError: + log.debug(f"Unable to parse Flash Attention v3 version as an int list: {flash3_version_str}.") + return False + + if flash3_version > FLASH_ATTENTION_V3_MAX_VERSION or flash3_version < FLASH_ATTENTION_V3_MIN_VERSION: + log.debug( + "Flash Attention v3 build is not supported; this backend only supports versions " + f"{FLASH_ATTENTION_V3_MIN_VERSION} through {FLASH_ATTENTION_V3_MAX_VERSION}, got " + f"{flash3_version}." + ) + return False + + return True + + +FLASH3_SUPPORTED = flash3_supported() + + +if FLASH3_SUPPORTED: + from cosmos_policy._src.imaginaire.attention.flash3.functions import flash3_attention + +else: + from cosmos_policy._src.imaginaire.attention.flash3.stubs import flash3_attention + +__all__ = ["flash3_attention", "FLASH3_SUPPORTED"] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/checks.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/checks.py new file mode 100644 index 0000000000000000000000000000000000000000..5dceaa4cd8ef18815d229b2ce7b2b66f2655ffe6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/checks.py @@ -0,0 +1,134 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Flash Attention v3 (flash3) backend checks +""" + +from functools import partial + +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.checks import attention_param_checks, attention_tensor_checks +from cosmos_policy._src.imaginaire.attention.flash3 import FLASH3_SUPPORTED +from cosmos_policy._src.imaginaire.attention.flash3.meta import get_bwd_dtypes, get_fwd_dtypes +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.utils import get_arch_tag, is_torch_compiling, log_or_raise_error + + +def flash3_attention_check( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool, + causal_type: CausalType, + is_varlen: bool, + raise_error: bool = False, +) -> bool: + """ + Input validation function for the flash3 backend. + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`). + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`). + + is_causal (bool): whether or not causal masking is enabled. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`. Required when `is_causal = True`. + + is_varlen (bool): whether or not a variable length (varlen) use case. Must be inferred + beforehand based on arguments such as seqlens_{Q,KV} or cumulative_seqlen_{Q,KV} being + passed. + + raise_error (bool): whether to raise an error if any checks fail or no backend is selected, + instead of just returning False. Default is False. + + Returns: + success (bool): whether use case is compatible with flash3 backend. + + """ + target_fn = partial(log_or_raise_error, raise_error=raise_error) + + if not FLASH3_SUPPORTED: + target_fn( + "Flash Attention v3 (flash3) is not supported in this environment. Run with debug logs to find out why, or choose another backend.", + exception=RuntimeError, + ) + return False + + if is_torch_compiling(): + target_fn( + "Flash Attention v3 (flash3) backend does not support torch.compile yet.", + exception=RuntimeError, + ) + return False + + arch_tag = get_arch_tag(query.device) + fwd_dtypes = get_fwd_dtypes(arch_tag) + bwd_dtypes = get_bwd_dtypes(arch_tag) + if not attention_tensor_checks( + query=query, + key=key, + value=value, + supported_dtypes_forward=fwd_dtypes, + supported_dtypes_backward=bwd_dtypes, + # flash3 supports MLA, unlike flash2, but with some constraints + # disabled for now due to API bug + supports_mla=False, + supports_gqa_mqa=True, + raise_error=raise_error, + backend_name="Flash Attention v3 (flash3)", + ): + target_fn("Flash Attention v3 (flash3) does not support the given inputs.", exception=RuntimeError) + return False + + # MLA constraints + if query.shape[-1] != value.shape[-1]: + head_dim_q = query.shape[-1] + head_dim_v = value.shape[-1] + if not ((head_dim_q <= 64 and head_dim_v <= 512) or (128 <= head_dim_q <= 192 and 96 <= head_dim_v <= 128)): + target_fn( + "Flash Attention v3 (flash3) does not support this head dim combination. " + "Expected either head_dim_qk <= 64 and head_dim_v <= 512, or 128 <= head_dim_qk <= 192 " + f"and 96 <= head_dim_v <= 128, got {head_dim_q=}, {head_dim_v=}.", + exception=ValueError, + ) + return False + + # Verifies causal_type is a CausalType instance when is_causal + # Verifies DontCare is not used unless seqlen_q == seqlen_kv + attention_param_checks( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + ) + + if is_causal and causal_type not in [CausalType.BottomRight, CausalType.DontCare]: + target_fn("Flash Attention only supports bottom-right causal masking.", exception=ValueError) + return False + + return True diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/functions.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/functions.py new file mode 100644 index 0000000000000000000000000000000000000000..0e03eeae49431c84e6f0430c2a3a5a1db7d714e4 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/functions.py @@ -0,0 +1,184 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Flash Attention v3 (flash3) Backend: intermediate APIs +Only safe to import when FLASH3_SUPPORTED is True. +""" + +import inspect + +from flash_attn_3.flash_attn_interface import flash_attn_func, flash_attn_varlen_func +from torch import Tensor + +# NOTE: older commits didn't have `return_attn_probs` as an argument, and there is no +# reflection of the commit hash in the version, so we have to manually inspect the signatures +HAS_RETURN_ATTN_PROBS = "return_attn_probs" in inspect.signature(flash_attn_func).parameters + +from cosmos_policy._src.imaginaire.attention.flash3.checks import flash3_attention_check +from cosmos_policy._src.imaginaire.attention.masks import CausalType + + +def flash3_attention( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool = False, + causal_type: CausalType | None = None, + scale: float | None = None, + cumulative_seqlen_Q: Tensor | None = None, + cumulative_seqlen_KV: Tensor | None = None, + max_seqlen_Q: int | None = None, + max_seqlen_KV: int | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + """ + Runs Flash Attention v3 on given operands (Q, K, V) with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`) + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`) + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`) + + is_causal (bool): whether or not causal masking is enabled. Default is False. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`. Required when `is_causal = True`. + + scale (float | None): Dot product scale (attention scale). Defaults to head_dim ** -0.5. + + cumulative_seqlen_Q (Tensor | None): (varlen) Optional 1-D tensor with size `batch + 1` + indicating the cumulative sum of number of query tokens in each batch, with an + additional 0 element in the beginning. Must be passed together with + `cumulative_seqlen_KV` and `max_seqlen_{Q,KV}`. + + cumulative_seqlen_KV (Tensor | None): (varlen) Optional 1-D tensor with size `batch + 1` + indicating the cumulative sum of number of key/value tokens in each batch, with an + additional 0 element in the beginning. Must be passed together with + `cumulative_seqlen_Q` and `max_seqlen_{Q,KV}`. + + max_seqlen_Q (int | None): (varlen) Optional integer indicating the maximum query + sequence length in all batches. Must be passed together with `cumulative_seqlen_{Q,KV}` + and `max_seqlen_KV`. + + max_seqlen_KV (int | None): (varlen) Optional integer indicating the maximum key/value + sequence length in all batches. Must be passed together with `cumulative_seqlen_{Q,KV}` + and `max_seqlen_Q`. + + Other Parameters: + return_lse (bool): Whether to return the logsumexp values. Default is False. + + backend_kwargs (dict | None): Key-value pair for passing arguments specific to Flash's + attention operator, if any. + + Returns: + output (Tensor): 4-D output tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim_v]`). + + logsumexp (Tensor): logsumexp tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, 1]`). Only returned when return_lse is True. + """ + + is_varlen = cumulative_seqlen_Q is not None + assert flash3_attention_check( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + is_varlen=is_varlen, + raise_error=True, + ) + + scale = scale if scale is not None else query.shape[-1] ** -0.5 + + backend_kwargs = backend_kwargs if backend_kwargs is not None else {} + + if HAS_RETURN_ATTN_PROBS: + backend_kwargs["return_attn_probs"] = True + + if is_varlen: + assert query.shape[0] == key.shape[0] == value.shape[0] == 1 + q = query.squeeze(0) + k = key.squeeze(0) + v = value.squeeze(0) + assert q.dim() == k.dim() == v.dim() == 3 + out, lse_ = flash_attn_varlen_func( + q=query.squeeze(0), + k=key.squeeze(0), + v=value.squeeze(0), + cu_seqlens_q=cumulative_seqlen_Q, + cu_seqlens_k=cumulative_seqlen_KV, + max_seqlen_q=max_seqlen_Q, + max_seqlen_k=max_seqlen_KV, + softmax_scale=scale, + causal=is_causal, + **backend_kwargs, + # qv=None, + # q_descale=None, k_descale=None, v_descale=None, + # attention_chunk=0, + # num_splits=1, + # pack_gqa=None, + # sm_margin=0, + # window_size=(-1, -1), + # softcap=0.0, # 0.0 means deactivated + # deterministic=False, + ) + assert out.dim() == 3 + assert lse_.dim() == 2 + + output = out.unsqueeze(0) + lse = lse_.unsqueeze(0) + + else: + output, lse = flash_attn_func( + q=query, + k=key, + v=value, + softmax_scale=scale, + causal=is_causal, + **backend_kwargs, + # qv=None, + # q_descale=None, k_descale=None, v_descale=None, + # attention_chunk=0, + # num_splits=1, + # pack_gqa=None, + # sm_margin=0, + # window_size=(-1, -1), + # softcap=0.0, # 0.0 means deactivated + # deterministic=False, + ) + + assert isinstance(output, Tensor) + assert isinstance(lse, Tensor) + assert output.dim() == 4 + assert lse.dim() == 3 + + lse = lse.permute(0, 2, 1).contiguous() # [batch, seqlen, head_dim] + + if return_lse: + return output, lse + + return output diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/meta.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/meta.py new file mode 100644 index 0000000000000000000000000000000000000000..45713024ac0f0bf206a86ddb2696164fbc9d98d9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/meta.py @@ -0,0 +1,64 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Flash Attention v3 (flash3) Backend: metadata +Always safe to import (as long as torch is available.) +""" + +import torch + +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + + +def get_fwd_dtypes(arch_tag: int) -> list[torch.dtype]: + """ + Returns data type choices for forward pass according to arch tag (attention.utils.get_arch_tag). + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + data_type_choices (list): a list of PyTorch data types. Empty if device is not supported. + + """ + + if arch_tag != 90: + log.debug("Flash Attention v3 (flash3) only supports compute capability 9.0 (Hopper).") + return [] + + return [torch.float16, torch.bfloat16] + + +def get_bwd_dtypes(arch_tag: int) -> list[torch.dtype]: + """ + Returns data type choices for backward pass according to arch tag (attention.utils.get_arch_tag). + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + data_type_choices (list): a list of PyTorch data types. Empty if device is not supported. + + """ + + if arch_tag != 90: + log.debug("Flash Attention v3 (flash3) only supports compute capability 9.0 (Hopper).") + return [] + + return [torch.float16, torch.bfloat16] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/stubs.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/stubs.py new file mode 100644 index 0000000000000000000000000000000000000000..fae9be0a81d4eecf5ff6fdc1ade05c5d49ca7dbe --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/flash3/stubs.py @@ -0,0 +1,46 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Flash Attention v3 (flash3) Backend: intermediate API stubs +Always safe to import (as long as torch is available.) +""" + +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.masks import CausalType + + +def flash3_attention( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool = False, + causal_type: CausalType | None = None, + scale: float | None = None, + cumulative_seqlen_Q: Tensor | None = None, + cumulative_seqlen_KV: Tensor | None = None, + max_seqlen_Q: int | None = None, + max_seqlen_KV: int | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + raise RuntimeError( + "Tried to run Flash Attention v3, but it is not supported / available. " + "Try running with debug logs enabled to see why." + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/frontend.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/frontend.py new file mode 100644 index 0000000000000000000000000000000000000000..a3a05d8d77c8188758b9480cee5fc028be21693a --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/frontend.py @@ -0,0 +1,587 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Frontend APIs +""" + +import torch +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.backends import choose_backend, choose_multi_dim_backend +from cosmos_policy._src.imaginaire.attention.checks import ( + attention_param_checks, + attention_tensor_checks, + multi_dim_attention_param_checks, + multi_dim_attention_param_filter, + multi_dim_attention_tensor_checks, + varlen_tensor_checks, +) +from cosmos_policy._src.imaginaire.attention.cudnn import cudnn_attention +from cosmos_policy._src.imaginaire.attention.flash2 import flash2_attention +from cosmos_policy._src.imaginaire.attention.flash3 import flash3_attention +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.natten import natten_attention, natten_multi_dim_attention +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +# Map backend names to their frontend attention API +BACKEND_MAP = { + "cudnn": cudnn_attention, + "natten": natten_attention, + "flash2": flash2_attention, + "flash3": flash3_attention, +} + +MULTI_DIM_BACKEND_MAP = { + "natten": natten_multi_dim_attention, +} + + +def attention( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool = False, + causal_type: CausalType | None = None, + scale: float | None = None, + # varlen parameters + seqlens_Q: Tensor | None = None, + seqlens_KV: Tensor | None = None, + cumulative_seqlen_Q: Tensor | None = None, + cumulative_seqlen_KV: Tensor | None = None, + max_seqlen_Q: int | None = None, + max_seqlen_KV: int | None = None, + # backend & misc parameters + backend: str | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + """ + Runs Attention on given operands (Q, K, V) with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + Varlen Attention is only supported for the sequence-packed layout: QKV tensors have batch size + 1, and tokens from different batches are concatenated without any padding along the sequence + dimension. Sequence lengths for different batches can be provided in two ways: + 1. `seqlens_Q` and `seqlens_KV` (less efficient): only provide the sequence lengths as + integer tensors (must be on the same device as QKV), and cumulative and maximum sequence + lengths are recomputed on each call. + 2. `cumulative_seqlen_{Q,KV}` and `max_seqlen_{Q,KV}` (more efficient): + compute cumulative and maximum sequence lengths. `cumulative_seqlen_{Q,KV}` are integer + tensors on the same device as QKV containing the cumulative sum of `seqlens_{Q,KV}`, + with an additional `0` element in the beginning, therefore sized `batch+1`. + `max_seqlen_{Q,KV}` are integers (not Tensors) that represent the maximum sequence + lengths for Q and KV among all sequence batches. + You can use `generate_varlen_parameters` to generate these + parameters: + ```python3 + from cosmos_policy._src.imaginaire.attention.varlen import generate_varlen_parameters + ( + cumulative_seqlen_Q, + cumulative_seqlen_KV, + max_seqlen_Q, + max_seqlen_KV, + ) = generate_varlen_parameters(q, k, v, seqlens_Q, seqlens_KV) + ``` + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen_q, heads, head_dim]`) + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`) + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`) + + is_causal (bool): whether or not causal masking is enabled. Default is False. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`, `CausalType.DontCare` (only valid when seqlen_q == seqlen_kv). + Required when `is_causal = True`. + + scale (float | None): Dot product scale (attention scale). Defaults to head_dim ** -0.5. + + seqlens_Q (Tensor | None): (varlen) Optional 1-D tensor with size `batch` + indicating the number of query tokens in each batch. Must be passed together with + `seqlens_KV`. + + seqlens_KV (Tensor | None): (varlen) Optional 1-D tensor with size `batch` + indicating the number of key/value tokens in each batch. Must be passed together with + `seqlens_Q`. + + cumulative_seqlen_Q (Tensor | None): (varlen) Optional 1-D tensor with size `batch + 1` + indicating the cumulative sum of number of query tokens in each batch, with an + additional 0 element in the beginning. Must be passed together with + `cumulative_seqlen_KV` and `max_seqlen_{Q,KV}`. + + cumulative_seqlen_KV (Tensor | None): (varlen) Optional 1-D tensor with size `batch + 1` + indicating the cumulative sum of number of key/value tokens in each batch, with an + additional 0 element in the beginning. Must be passed together with + `cumulative_seqlen_Q` and `max_seqlen_{Q,KV}`. + + max_seqlen_Q (int | None): (varlen) Optional integer indicating the maximum query + sequence length in all batches. Must be passed together with `cumulative_seqlen_{Q,KV}` + and `max_seqlen_KV`. + + max_seqlen_KV (int | None): (varlen) Optional integer indicating the maximum key/value + sequence length in all batches. Must be passed together with `cumulative_seqlen_{Q,KV}` + and `max_seqlen_Q`. + + Other Parameters: + backend (str | None): Backend to run with. If unspecified (default), it will try to + select the best available. + + return_lse (bool): Whether to return the logsumexp values. Default is False. + + backend_kwargs (dict | None): Key-value pair for passing arguments specific to the backend's + attention operator, if any. Only valid when a specific backend is selected (backend is + not None). + + Returns: + output (Tensor): 4-D output tensor, with the heads-last contiguous layout + (`[batch, seqlen_q, heads, head_dim_v]`). + + logsumexp (Tensor): logsumexp tensor, with the heads-last contiguous layout + (`[batch, seqlen_q, heads, 1]`). Only returned when return_lse is True. + """ + + assert attention_tensor_checks(query=query, key=key, value=value, raise_error=True) + + attention_param_checks( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + ) + + ( + cumulative_seqlen_Q, + cumulative_seqlen_KV, + max_seqlen_Q, + max_seqlen_KV, + ) = varlen_tensor_checks( + query=query, + key=key, + value=value, + seqlens_Q=seqlens_Q, + seqlens_KV=seqlens_KV, + cumulative_seqlen_Q=cumulative_seqlen_Q, + cumulative_seqlen_KV=cumulative_seqlen_KV, + max_seqlen_Q=max_seqlen_Q, + max_seqlen_KV=max_seqlen_KV, + ) + is_varlen = cumulative_seqlen_Q is not None + + scale = scale if scale is not None else query.shape[-1] ** -0.5 + + if backend is None and backend_kwargs is not None: + backend_kwargs = None + log.debug("A backend was not specified, but got backend_kwargs. Ignoring... ") + + if backend is not None and backend not in BACKEND_MAP: + raise ValueError(f"Selected {backend=}, but available choices are {BACKEND_MAP.keys()}. ") + + compatible_backend = choose_backend( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + is_varlen=is_varlen, + backend=backend, + raise_error=False, + ) + + # Either incompatible backend specified by user, or no compatible backends found + # Try to see if we can handle it with graph transformations + # For now only handling GQA/MQA, but MLA, varlen, and some other features are also + # implementable with graph transformations, but we may need them even if not as efficient. + if compatible_backend is None: + is_gqa_mqa = query.shape[-2] != key.shape[-2] and query.shape[-2] > key.shape[-2] + + # In practice this is the only reason why no backend would be selected, + # but moving forward we should represent support matrices for backends explicitly + # and rely on reasons to make the best decision when it comes to graph transformations. + if is_gqa_mqa: + heads = query.shape[-2] + heads_kv = key.shape[-2] + assert heads % heads_kv == 0 + h_k = heads // heads_kv + + query_t = query + key_t = torch.repeat_interleave(key, repeats=h_k, dim=-2, output_size=heads) + value_t = torch.repeat_interleave(value, repeats=h_k, dim=-2, output_size=heads) + + log.debug("Backend incompatible with GQA/MQA use case. Trying again with graph transformation... ") + return attention( + query=query_t, + key=key_t, + value=value_t, + is_causal=is_causal, + causal_type=causal_type, + scale=scale, + cumulative_seqlen_Q=cumulative_seqlen_Q, + cumulative_seqlen_KV=cumulative_seqlen_KV, + max_seqlen_Q=max_seqlen_Q, + max_seqlen_KV=max_seqlen_KV, + return_lse=return_lse, + backend=backend, + backend_kwargs=backend_kwargs, + ) + + if backend is None: + raise ValueError( + "Could not find a compatible Attention backend for this use case / device. " + "Try running with debug logs to find out why." + ) + else: + raise ValueError( + f"Selected Attention backend {backend} is incompatible with this use case / device. " + "Try running with debug logs to find out why." + ) + + assert compatible_backend in BACKEND_MAP + return BACKEND_MAP[compatible_backend]( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + scale=scale, + cumulative_seqlen_Q=cumulative_seqlen_Q, + cumulative_seqlen_KV=cumulative_seqlen_KV, + max_seqlen_Q=max_seqlen_Q, + max_seqlen_KV=max_seqlen_KV, + return_lse=return_lse, + backend_kwargs=backend_kwargs, + ) + + +def multi_dimensional_attention( + query: Tensor, + key: Tensor, + value: Tensor, + window_size: tuple | int = -1, + stride: tuple | int = 1, + dilation: tuple | int = 1, + is_causal: tuple | bool = False, + scale: float | None = None, + # backend & misc parameters + backend: str | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + """ + Runs Multi-Dimensional Attention on given operands (Q, K, V) with the heads-last contiguous + layout (`[batch, *, heads, head_dim]`). Supports up to and including 3 dimensions: + * 1-D: `[batch, X, heads, head_dim]`, with masking arguments expecting tuples of size 1. + * 2-D: `[batch, X, Y, heads, head_dim]`, with masking arguments expecting tuples of size 2. + * 3-D: `[batch, X, Y, Z, heads, head_dim]`, with masking arguments expecting tuples of size 3. + + The dimensions here refer to the layout of tokens; that is the arrangement of tokens for each + batch/head, or the `[X]`, `[X, Y]`, `[X, Y, Z]` part of the input shape. + We refer to these as the "token layout shape". + + For now, it is always expected that Q, K, and V match in the sizes of those dimensions. + + Masking arguments, all of which can be set uniformly across all dimensions or per dimension, are: + * `window_size`: determines the sliding window size. -1 is interpreted as the maximum window + size. Must be either -1 or at least 2 and at most the token layout shape. + For example, if inputs are `[batch, X, Y, Z, heads_{q,kv}, head_dim_{qk,v}]`, + `window_size` must be either an integer == -1 or an integer <= `min(X, Y, Z)`, + or a tuple of size 3 corresponding to the three dimensions / axes, where: + * `window_size[0] == -1 or 2 <= window_size[0] <= X` + * `window_size[1] == -1 or 2 <= window_size[1] <= Y` + * `window_size[2] == -1 or 2 <= window_size[2] <= Z` + When `window_size` is set to the maximum for any dimension, we're effectively performing + self attention (no sparsity) along that dimension. + Default is -1 (self attention). + + * `stride`: determines the step size of the sliding window. Only matters when the + corresponding `window_size` is not -1 / maximum (self attention). + Default is 1, indicating the smallest sliding window delay. + Larger values trade off translational equivariance for potentially improved efficiency. + Maximum value for `stride` along each dimension is the corresponding `window_size`. + If `stride == window_size` along any dimension, it is equivalent to blocked / windowed + attention (from works such as Swin Transformer, SAM, ViTDet, etc) along that dimension, + meaning no overlap between windows. + For more details, please refer to the GNA paper: + https://arxiv.org/abs/2504.16922 + + * `dilation`: introduces gaps between tokens in a sliding window, similarly to dilated + convolution. + Default is 1, indicating no gaps. + Maximum value is the largest positive integer that satisfies + `window_size * dilation <= token_layout_shape` along that dimension. + Higher dilation means more sparse and global context. Lower dilation means more + locality. + For more details, please refer to the DiNAT paper: + https://arxiv.org/abs/2209.15001 + + * `is_causal`: per-dimension causal mask. + + Parameters: + query (Tensor): 4-D, 5-D, or 6-D query tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads, head_dim]`) + + key (Tensor): 4-D, 5-D, or 6-D key tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim]`) + + value (Tensor): 4-D, 5-D, or 6-D value tensor, with heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim_v]`) + + window_size (tuple | int): Attention window (kernel) size / shape. If an + integer, it will be repeated for all dimensions. For example `window_size=3`, when + `len(token_layout_shape) == 3`, is interpreted as `window_size=(3, 3, 3)`. + `-1`s are replaced with the corresponding `token_layout_shape`. + Final window size must satisfy `2 <= window_size <= token_layout_shape`. + Default is -1 (no sparsity). + + stride (tuple | int): Sliding window step size/shape. If an integer, it will be repeated + for all dimensions. For example `stride=2`, when `len(token_layout_shape) == 3`, is + interpreted as `stride=(2, 2, 2)`. + Final stride must satisfy `1 <= stride <= window_size`. + Default is 1. + + dilation (tuple | int): Dilation step size/shape. If an integer, it will be repeated for + all dimensions. For example `dilation=4`, when `len(token_layout_shape) == 3`, is + interpreted as `dilation=(4, 4, 4)`. + Final dilation must satisfy `2 <= dilation * window_size <= token_layout_shape`. + Default is 1. + + is_causal (tuple | bool): Toggle causal masking. If a boolean, it will be repeated for all + dimensions. For example `is_causal=True`, when `len(token_layout_shape) == 3`, is + interpreted as `is_causal=(True, True, True)`. + Default is False. + + scale (float | None): Dot product scale (attention scale). Defaults to head_dim ** -0.5. + + Other Parameters: + backend (str | None): Backend to run with. If unspecified (default), it will try to + select the best available. + + return_lse (bool): Whether to return the logsumexp values. Default is False. + + backend_kwargs (dict | None): Key-value pair for passing arguments specific to the backend's + multi-dim / sparse attention operator, if any. Only valid when a specific backend is + selected (backend is not None). + + Returns: + output (Tensor): 4-D, 5-D, or 6-D output tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads, head_dim_v]`). + + logsumexp (Tensor): logsumexp tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads, 1]`). Only returned when return_lse is True. + """ + + assert multi_dim_attention_tensor_checks(query=query, key=key, value=value, raise_error=True) + + token_layout_shape, window_size, stride, dilation, is_causal = multi_dim_attention_param_filter( + query, + window_size=window_size, + stride=stride, + dilation=dilation, + is_causal=is_causal, + ) + num_dims = len(token_layout_shape) + + # Automatic transformation for 1s in token layout + # I.e. Attention over a (1, 16, 32) token layout is identical to over a (16, 32) + # NOTE: assumes QKV token layouts match + token_layout_ones = [i for i in range(num_dims) if token_layout_shape[i] == 1] + if len(token_layout_ones) > 0: + token_layout_t = tuple(s for i, s in enumerate(token_layout_shape) if i not in token_layout_ones) + window_size_t = tuple(w for i, w in enumerate(window_size) if i not in token_layout_ones) + stride_t = tuple(s for i, s in enumerate(stride) if i not in token_layout_ones) + dilation_t = tuple(d for i, d in enumerate(dilation) if i not in token_layout_ones) + is_causal_t = tuple(c for i, c in enumerate(is_causal) if i not in token_layout_ones) + + assert all(x >= 2 for x in token_layout_t) + assert all(w >= 2 for w in window_size_t) + + query_t = query.reshape(query.shape[0], *token_layout_t, query.shape[-2], query.shape[-1]) + key_t = key.reshape(key.shape[0], *token_layout_t, key.shape[-2], key.shape[-1]) + value_t = key.reshape(value.shape[0], *token_layout_t, value.shape[-2], value.shape[-1]) + + log.debug( + "This Multi-Dimensional Attention problem has 1s in the token layout, which can be simplified from " + f"<{token_layout_shape=}, {window_size=}, {stride=}, {dilation=}, {is_causal=}> into " + f"<{token_layout_t=}, {window_size_t=}, {stride_t=}, {dilation_t=}, {is_causal_t=}>." + ) + + return multi_dimensional_attention( + query=query_t, + key=key_t, + value=value_t, + window_size=window_size_t, + stride=stride_t, + dilation=dilation_t, + is_causal=is_causal_t, + scale=scale, + backend=backend, + return_lse=return_lse, + backend_kwargs=backend_kwargs, + ) + + multi_dim_attention_param_checks( + query, + window_size=window_size, + stride=stride, + dilation=dilation, + is_causal=is_causal, + ) + + # Fast path for self attention problems + if all(x == w for x, w in zip(token_layout_shape, window_size)) and ( + not any(c for c in is_causal) or num_dims == 1 + ): + log.debug( + "This Multi-Dimensional Attention problem is implementable with standard Attention: " + f"{token_layout_shape=}, {window_size=}, {is_causal=}." + ) + if backend is not None: + log.debug(f"Ignoring {backend=} and backend args...") + + query_1d = query.flatten(1, num_dims) + key_1d = key.flatten(1, num_dims) + value_1d = value.flatten(1, num_dims) + is_causal_1d = is_causal[0] + + return attention( + query_1d, + key_1d, + value_1d, + scale=scale, + is_causal=is_causal_1d, + causal_type=CausalType.DontCare, + return_lse=return_lse, + ) + + scale = scale if scale is not None else query.shape[-1] ** -0.5 + + if backend is None and backend_kwargs is not None: + backend_kwargs = None + log.debug("A backend was not specified, but got backend_kwargs. Ignoring... ") + + backend = choose_multi_dim_backend( + query=query, + key=key, + value=value, + backend=backend, + ) + + if backend not in MULTI_DIM_BACKEND_MAP: + raise ValueError(f"Selected {backend=}, but available choices are {MULTI_DIM_BACKEND_MAP.keys()}. ") + + return MULTI_DIM_BACKEND_MAP[backend]( + query=query, + key=key, + value=value, + window_size=window_size, + stride=stride, + dilation=dilation, + is_causal=is_causal, + scale=scale, + return_lse=return_lse, + backend_kwargs=backend_kwargs, + ) + + +def spatio_temporal_attention( + query: Tensor, + key: Tensor, + value: Tensor, + window_size: tuple | int = -1, + stride: tuple | int = 1, + dilation: tuple | int = 1, + scale: float | None = None, + # backend & misc parameters + backend: str | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + """ + Runs Spatio-Temporal Attention on unflattened QKV with the heads-last contiguous layout + (`[batch, T, H, W, heads, head_dim]`). + For now, it is always expected that Q, K, and V match in their shapes. + + Parameters: + query (Tensor): 6-D query tensor, with the heads-last contiguous layout + (`[batch, T, H, W, heads, head_dim]`) + + key (Tensor): 6-D key tensor, with the heads-last contiguous layout + (`[batch, T, H, W, heads_kv, head_dim]`) + + value (Tensor): 6-D value tensor, with heads-last contiguous layout + (`[batch, T, H, W, heads_kv, head_dim_v]`) + + window_size (tuple | int): Attention window (kernel) size / shape. If an + integer, it will be repeated for all dimensions. For example `window_size=3` is + interpreted as `window_size=(3, 3, 3)`. + `-1`s are replaced with the corresponding value in `(T, H, W)`. + Default is -1 (no sparsity). + + stride (tuple | int): Sliding window step size/shape. If an integer, it will be repeated + for all dimensions. For example `stride=2` is interpreted as `stride=(2, 2, 2)`. + Final stride must satisfy `1 <= stride <= window_size`. + Default is 1. + + dilation (tuple | int): Dilation step size/shape. If an integer, it will be repeated for + all dimensions. For example `dilation=4` is interpreted as `dilation=(4, 4, 4)`. + Final dilation must satisfy `2 <= dilation * window_size <= (T, H, W)`. + Default is 1. + + scale (float | None): Dot product scale (attention scale). Defaults to head_dim ** -0.5. + + Other Parameters: + backend (str | None): Backend to run with. If unspecified (default), it will try to + select the best available. + + return_lse (bool): Whether to return the logsumexp values. Default is False. + + backend_kwargs (dict | None): Key-value pair for passing arguments specific to the backend's + multi-dim / sparse attention operator, if any. Only valid when a specific backend is + selected (backend is not None). + + Returns: + output (Tensor): 6-D output tensor, with the heads-last contiguous layout + (`[batch, T, H, W, heads, head_dim_v]`). + + logsumexp (Tensor): logsumexp tensor, with the heads-last contiguous layout + (`[batch, T, H, W, heads, 1]`). Only returned when return_lse is True. + """ + if query.dim() != 6: + raise ValueError( + "Spatio-Temporal Attention requires 6-D input tensors ([batch, T, H, W, heads, head_dim]), " + f"got {query.shape=})." + ) + + return multi_dimensional_attention( + query=query, + key=key, + value=value, + window_size=window_size, + stride=stride, + dilation=dilation, + is_causal=(True, False, False), + scale=scale, + return_lse=return_lse, + backend_kwargs=backend_kwargs, + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/masks.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/masks.py new file mode 100644 index 0000000000000000000000000000000000000000..3cad6c02dfea9ba2ad7cdb25e26c393678bbedd2 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/masks.py @@ -0,0 +1,61 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Mask utilities +""" + +from enum import Enum + + +class CausalType(Enum): + """ + Different types of causal masking supported by backends of interest. + """ + + # Top-Left: Simplified: mask if q_idx < kv_idx + # CUTLASS / NATTEN default + # Q = 2, KV = 5: + # O____ + # OO___ + # + # Q = 5, KV = 2: + # O_ + # OO + # OO + # OO + # OO + TopLeft = 0 + + # Bottom-right: mask if q_idx + KV - Q < kv_idx + # Flash Attention default + # Q = 2, KV = 5: + # OOOO_ + # OOOOO + # + # Q = 5, KV = 2: + # __ + # __ + # __ + # O_ + # OO + BottomRight = 1 + + # When seqlen_q == seqlen_kv, we don't care about the causal type + # because top-left and bottom-right are equivalent + DontCare = 2 diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..8b02f644691fe44815fc955ee6f0dbb0a6cb1e25 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/__init__.py @@ -0,0 +1,95 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +NATTEN Backend +""" + +import torch + +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +NATTEN_MIN_RELEASE_VERSION = [0, 21, 5] +# 0.21.5.dev1 patches some varlen issues +# 0.21.5.dev2 adds torch compile support +# 0.21.5.dev3 fixes a few compat issues for older torch versions +NATTEN_MIN_DEV_VERSION = ([0, 21, 5], 3) + + +def natten_supported() -> bool: + """ + Returns whether NATTEN is supported in this environment. + Requirements are: + * Presence of CUDA Runtime (via PyTorch) + * Presence of NATTEN, meeting minimum version requirements + + This check guards imports / dependencies on the NATTEN package. + """ + if not torch.cuda.is_available(): + log.debug("NATTEN Attention is not supported because PyTorch did not detect CUDA runtime.") + return False + + try: + import natten + + except ImportError: + log.debug("NATTEN Attention is not supported because the Python package was not found.") + return False + except Exception as e: + log.debug(f"NATTEN Attention is not supported because importing the Python package failed: {e}") + return False + + natten_version_split = natten.__version__.split(".") + if len(natten_version_split) < 3 or len(natten_version_split) > 4: + log.debug(f"Unable to parse NATTEN version {natten.__version__}.") + return False + + try: + natten_version = [int(x) for x in natten_version_split[:3]] + natten_version_dev = None + if len(natten_version_split) >= 4 and natten_version_split[3].startswith("dev"): + natten_version_dev = int(natten_version_split[3].replace("dev", "")) + + except ValueError: + log.debug(f"Unable to parse NATTEN version as an int list: {natten.__version__}.") + return False + + if (natten_version_dev is None and natten_version >= NATTEN_MIN_RELEASE_VERSION) or ( + natten_version_dev is not None + and natten_version >= NATTEN_MIN_DEV_VERSION[0] + and natten_version_dev >= NATTEN_MIN_DEV_VERSION[1] + ): + return True + + log.debug( + "NATTEN Attention is not supported due to insufficient NATTEN version " + f"{natten.__version__=}, expected at least {NATTEN_MIN_RELEASE_VERSION=}, " + f"or {NATTEN_MIN_DEV_VERSION=}." + ) + return False + + +NATTEN_SUPPORTED = natten_supported() + +if NATTEN_SUPPORTED: + from cosmos_policy._src.imaginaire.attention.natten.functions import natten_attention, natten_multi_dim_attention + +else: + from cosmos_policy._src.imaginaire.attention.natten.stubs import natten_attention, natten_multi_dim_attention + +__all__ = ["natten_attention", "natten_multi_dim_attention", "NATTEN_SUPPORTED"] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/checks.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/checks.py new file mode 100644 index 0000000000000000000000000000000000000000..15ceb827ad772d25875c98b0cc9330534178163b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/checks.py @@ -0,0 +1,391 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +NATTEN backend checks +""" + +from functools import partial + +import torch +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.checks import ( + attention_param_checks, + attention_tensor_checks, + multi_dim_attention_tensor_checks, +) +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.natten import NATTEN_SUPPORTED +from cosmos_policy._src.imaginaire.attention.natten.meta import get_bwd_dtypes, get_fwd_dtypes +from cosmos_policy._src.imaginaire.attention.utils import get_arch_tag, log_or_raise_error +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + + +def dtype_supported( + dtype: torch.dtype, is_training: bool, dtypes_fwd: list[torch.dtype], dtypes_bwd: list[torch.dtype] | None = None +) -> bool: + """ + Helper determining whether dtype is supported with different sets of supported dtypes for + training and inference (forward+backward and forward). + + Parameters: + dtype (torch.dtype): tensor element type. + + is_training (bool): whether use case can be used to backpropagate (tensor.requires_grad). + + dtypes_fwd (list[torch.dtype]): list of dtypes allowed for inference only (when not + tensor.requires_grad). + + dtypes_bwd (list[torch.dtype] | None): Optional list of dtypes allowed for training only + (when tensor.requires_grad), if different from dtypes_fwd. + + """ + if is_training and dtypes_bwd is not None: + return dtype in dtypes_bwd + return dtype in dtypes_fwd + + +def choose_natten_backend( + query: Tensor, key: Tensor, value: Tensor, is_causal: bool, is_varlen: bool, raise_error: bool = False +) -> str | None: + """ + Chooses an FMHA backend in NATTEN (cutlass-fmha, hopper-fmha, blackwell-fmha) for the current + use case based on features needed and current GPU architecture. + + Using tensor shapes, it infers whether MLA (head_dim_value != head_dim_qk) or + GQA/MQA (heads_kv != heads_q) are required. + Using tensor device, it infers GPU architecture and compatible backends. + Using arguments is_causal and is_varlen, and other inferred features, it picks the best + available backend. + + It is possible for no backend to be selected, if the combination of features is not available in + any one of the NATTEN backends, in which case it will return None. + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`). + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`). + + is_causal (bool): whether or not causal masking is enabled. + + is_varlen (bool): whether or not a variable length (varlen) use case. Must be inferred + beforehand based on arguments such as seqlens_{Q,KV} or cumulative_seqlen_{Q,KV} being + passed. + + raise_error (bool): whether to raise an error if no backend is selected, instead of just + returning None. Default is False. + + Returns: + backend (str | None): selected NATTEN backend, if any compatible. + + """ + target_fn = partial(log_or_raise_error, raise_error=raise_error) + + # NOTE: assumes attention_tensor_checks have already been run once! + arch_tag = get_arch_tag(query.device) + dtype = query.dtype + is_training = query.requires_grad + + is_mla = query.shape[-1] != value.shape[-1] + is_gqa_mqa = query.shape[-2] != key.shape[-2] + + # banning devices not supported since CUDA 13.0 for simplicity + if arch_tag < 75: + log.debug("NATTEN is not supported because compute capability is below the minimum (7.5).") + return None + + # blackwell-fmha: sm100 and sm103 only. + # limitations: no mla (TBD). + blackwell_fmha_fwd_dtypes = [torch.float16, torch.bfloat16, torch.float8_e5m2, torch.float8_e4m3fn] + blackwell_fmha_bwd_dtypes = [torch.float16, torch.bfloat16] + dtype_supported_blackwell = dtype_supported( + dtype=dtype, is_training=is_training, dtypes_fwd=blackwell_fmha_fwd_dtypes, dtypes_bwd=blackwell_fmha_bwd_dtypes + ) + if arch_tag in [100, 103] and not is_mla and dtype_supported_blackwell: + return "blackwell-fmha" + else: + reason = "" + if arch_tag not in [100, 103]: + reason += f"Incompatible architecture ({arch_tag}, expected 100 or 103). " + if is_mla: + reason += "Use case is MLA (head_dim_qk != head_dim_value). " + if not dtype_supported_blackwell: + if is_training: + reason += ( + f"Data type {dtype} is not in list of supported dtypes for training: {blackwell_fmha_bwd_dtypes}. " + ) + else: + reason += ( + f"Data type {dtype} is not in list of supported dtypes for inference: {blackwell_fmha_fwd_dtypes}. " + ) + log.debug(f"NATTEN backend blackwell-fmha is not compatible. Reason: {reason}") + + # hopper-fmha: sm90 only. + # limitations: no causal masking (TBD), no varlen, no gqa/mqa, no mla. + hopper_fmha_dtypes = [torch.float16, torch.bfloat16] + dtype_supported_hopper = dtype_supported(dtype=dtype, is_training=is_training, dtypes_fwd=hopper_fmha_dtypes) + if arch_tag == 90 and not is_causal and not is_varlen and not is_gqa_mqa and not is_mla and dtype_supported_hopper: + return "hopper-fmha" + else: + reason = "" + if arch_tag != 90: + reason += f"Incompatible architecture ({arch_tag}, expected 90). " + if is_causal: + reason += "Use case is causal. " + if is_varlen: + reason += "Use case is varlen. " + if is_gqa_mqa: + reason += "Use case is GQA/MQA. " + if is_mla: + reason += "Use case is MLA (head_dim_qk != head_dim_value). " + if not dtype_supported_hopper: + reason += f"Data type {dtype} is not in list of supported dtypes: {hopper_fmha_dtypes}. " + log.debug(f"NATTEN backend hopper-fmha is not compatible. Reason: {reason}") + + # cutlass-fmha: targets sm50, sm70, sm75, sm80 (supports sm80+) + # limitations: no gqa/mqa. + cutlass_fmha_dtypes = [torch.float32, torch.float16, torch.bfloat16] + dtype_supported_cutlass = dtype_supported(dtype=dtype, is_training=is_training, dtypes_fwd=cutlass_fmha_dtypes) + if not is_gqa_mqa and dtype_supported_cutlass: + return "cutlass-fmha" + else: + reason = "" + if is_gqa_mqa: + reason += "Use case is GQA/MQA. " + if not dtype_supported_cutlass: + reason += f"Data type {dtype} is not in list of supported dtypes: {cutlass_fmha_dtypes}. " + log.debug(f"NATTEN backend cutlass-fmha is not compatible. Reason: {reason}") + + target_fn( + f"Could not find a compatible NATTEN FMHA backend for {arch_tag=}, {is_causal=}, " + f"{is_varlen=}, {is_mla=}, {is_gqa_mqa=}.", + exception=RuntimeError, + ) + return None + + +def natten_attention_check( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool, + causal_type: CausalType, + is_varlen: bool, + raise_error: bool = False, +) -> bool: + """ + Input validation function for the NATTEN backend. + Runs the common checks in addition to trying to find a compatible NATTEN backend. If any checks + fail, or no compatible backend is found in NATTEN, returns False. + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`). + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`). + + is_causal (bool): whether or not causal masking is enabled. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`. Required when `is_causal = True`. + + is_varlen (bool): whether or not a variable length (varlen) use case. Must be inferred + beforehand based on arguments such as seqlens_{Q,KV} or cumulative_seqlen_{Q,KV} being + passed. + + raise_error (bool): whether to raise an error if any checks fail or no backend is selected, + instead of just returning False. Default is False. + + Returns: + success (bool): whether use case is compatible with NATTEN backend. + + """ + target_fn = partial(log_or_raise_error, raise_error=raise_error) + + if not NATTEN_SUPPORTED: + target_fn( + "NATTEN is not supported in this environment. Run with debug logs to find out why, or choose another backend.", + exception=RuntimeError, + ) + return False + + arch_tag = get_arch_tag(query.device) + fwd_dtypes = get_fwd_dtypes(arch_tag) + bwd_dtypes = get_bwd_dtypes(arch_tag) + if not attention_tensor_checks( + query=query, + key=key, + value=value, + supported_dtypes_forward=fwd_dtypes, + supported_dtypes_backward=bwd_dtypes, + supports_mla=True, + supports_gqa_mqa=True, + raise_error=raise_error, + backend_name="NATTEN Attention", + ): + target_fn("NATTEN does not support the given inputs.", exception=RuntimeError) + return False + + # Verifies causal_type is a CausalType instance when is_causal + # Verifies DontCare is not used unless seqlen_q == seqlen_kv + attention_param_checks( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + ) + + if is_causal and causal_type not in [CausalType.TopLeft, CausalType.DontCare]: + target_fn("NATTEN Attention only supports top-left causal masking for now.", exception=RuntimeError) + return False + + natten_backend = choose_natten_backend( + query, key, value, is_causal=is_causal, is_varlen=is_varlen, raise_error=raise_error + ) + + if natten_backend is None: + return False + + return True + + +def choose_natten_multi_dim_backend(query: Tensor, key: Tensor, value: Tensor, raise_error: bool = False) -> str | None: + """ + Chooses an FNA backend in NATTEN (cutlass-fna, hopper-fna, blackwell-fna) for the current + use case based on features needed and current GPU architecture. + + Using tensor shapes, it infers whether MLA (head_dim_value != head_dim_qk) or + GQA/MQA (heads_kv != heads_q) are required. + Using tensor device, it infers GPU architecture and compatible backends. + Using arguments is_causal and is_varlen, and other inferred features, it picks the best + available backend. + + It is possible for no backend to be selected, if the combination of features is not available in + any one of the NATTEN backends, in which case it will return None. + + Parameters: + query (Tensor): 4-D, 5-D, or 6-D query tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads, head_dim]`). + + key (Tensor): 4-D, 5-D, or 6-D key tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim]`). + + value (Tensor): 4-D, 5-D, or 6-D value tensor, with heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim_v]`). + + raise_error (bool): whether to raise an error if no backend is selected, instead of just + returning None. Default is False. + + Returns: + backend (str | None): selected NATTEN backend, if any compatible. + + """ + + # Reuse choose_natten_backend instead of duplicating code + # NATTEN specifically makes sure the FNA counterparts cover all the features the FMHA kernels + # do. + fmha_backend = choose_natten_backend( + query=query, + key=key, + value=value, + is_causal=False, # causal masking in supported across all multi-dim (FNA) backends + is_varlen=False, # varlen is undefined (so far) for multi-dim + raise_error=raise_error, + ) + + natten_fmha_backend_to_fna_backend = { + "cutlass-fmha": "cutlass-fna", + "hopper-fmha": "hopper-fna", + "blackwell-fmha": "blackwell-fna", + } + + assert fmha_backend in natten_fmha_backend_to_fna_backend + return natten_fmha_backend_to_fna_backend[fmha_backend] + + +def natten_multi_dim_attention_check( + query: Tensor, + key: Tensor, + value: Tensor, + raise_error: bool = False, +) -> bool: + """ + Input validation function for the NATTEN multi-dimensional backend. + Runs the common checks in addition to trying to find a compatible NATTEN backend. If any checks + fail, or no compatible backend is found in NATTEN, returns False. + + Parameters: + query (Tensor): 4-D, 5-D, or 6-D query tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads, head_dim]`). + + key (Tensor): 4-D, 5-D, or 6-D key tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim]`). + + value (Tensor): 4-D, 5-D, or 6-D value tensor, with heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim_v]`). + + raise_error (bool): whether to raise an error if any checks fail or no backend is selected, + instead of just returning False. Default is False. + + Returns: + success (bool): whether use case is compatible with NATTEN backend. + + """ + target_fn = partial(log_or_raise_error, raise_error=raise_error) + + if not NATTEN_SUPPORTED: + target_fn( + "NATTEN is not supported in this environment. Run with debug logs to find out why, or choose another backend.", + exception=RuntimeError, + ) + return False + + arch_tag = get_arch_tag(query.device) + fwd_dtypes = get_fwd_dtypes(arch_tag) + bwd_dtypes = get_bwd_dtypes(arch_tag) + if not multi_dim_attention_tensor_checks( + query=query, + key=key, + value=value, + supported_dtypes_forward=fwd_dtypes, + supported_dtypes_backward=bwd_dtypes, + supports_mla=True, + supports_gqa_mqa=False, # NATTEN's FNA ops don't support GQA/MQA yet + raise_error=raise_error, + backend_name="NATTEN Multi-Dimensional Attention", + ): + target_fn("NATTEN does not support the given inputs.", exception=RuntimeError) + return False + + natten_backend = choose_natten_multi_dim_backend(query, key, value, raise_error=raise_error) + + if natten_backend is None: + return False + + return True diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/functions.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/functions.py new file mode 100644 index 0000000000000000000000000000000000000000..09bdb2ba00d1dc9a1c048c102a05d92c98acf19b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/functions.py @@ -0,0 +1,293 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +NATTEN Backend: intermediate APIs +Only safe to import when NATTEN_SUPPORTED is True. +""" + +from natten.context import set_memory_usage_preference, use_kv_parallelism_in_fused_na +from natten.functional import attention as _natten_attention +from natten.functional import neighborhood_attention_generic as _natten_multi_dim_attention +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.checks import ( + multi_dim_attention_param_checks, + multi_dim_attention_param_filter, +) +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.natten.checks import ( + choose_natten_backend, + choose_natten_multi_dim_backend, + natten_attention_check, + natten_multi_dim_attention_check, +) + +set_memory_usage_preference("unrestricted") +use_kv_parallelism_in_fused_na(True) + + +def natten_attention( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool = False, + causal_type: CausalType | None = None, + scale: float | None = None, + cumulative_seqlen_Q: Tensor | None = None, + cumulative_seqlen_KV: Tensor | None = None, + max_seqlen_Q: int | None = None, + max_seqlen_KV: int | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + """ + Runs NATTEN Attention on given operands (Q, K, V) with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`). + + Parameters: + query (Tensor): 4-D query tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim]`) + + key (Tensor): 4-D key tensor, with the heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim]`) + + value (Tensor): 4-D value tensor, with heads-last contiguous layout + (`[batch, seqlen_kv, heads_kv, head_dim_v]`) + + is_causal (bool): whether or not causal masking is enabled. Default is False. + + causal_type (CausalType): causal masking mode. Choices: `CausalType.TopLeft`, + `CausalType.BottomRight`. Required when `is_causal = True`. + + scale (float | None): Dot product scale (attention scale). Defaults to head_dim ** -0.5. + + cumulative_seqlen_Q (Tensor | None): (varlen) Optional 1-D tensor with size `batch + 1` + indicating the cumulative sum of number of query tokens in each batch, with an + additional 0 element in the beginning. Must be passed together with + `cumulative_seqlen_KV` and `max_seqlen_{Q,KV}`. + + cumulative_seqlen_KV (Tensor | None): (varlen) Optional 1-D tensor with size `batch + 1` + indicating the cumulative sum of number of key/value tokens in each batch, with an + additional 0 element in the beginning. Must be passed together with + `cumulative_seqlen_Q` and `max_seqlen_{Q,KV}`. + + max_seqlen_Q (int | None): (varlen) Optional integer indicating the maximum query + sequence length in all batches. Must be passed together with `cumulative_seqlen_{Q,KV}` + and `max_seqlen_KV`. + + max_seqlen_KV (int | None): (varlen) Optional integer indicating the maximum key/value + sequence length in all batches. Must be passed together with `cumulative_seqlen_{Q,KV}` + and `max_seqlen_Q`. + + Other Parameters: + return_lse (bool): Whether to return the logsumexp values. Default is False. + + backend_kwargs (dict | None): Key-value pair for passing arguments specific to NATTEN's + attention operator, if any. + + Returns: + output (Tensor): 4-D output tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, head_dim_v]`). + + logsumexp (Tensor): logsumexp tensor, with the heads-last contiguous layout + (`[batch, seqlen, heads, 1]`). Only returned when return_lse is True. + """ + + is_varlen = cumulative_seqlen_Q is not None + assert natten_attention_check( + query=query, + key=key, + value=value, + is_causal=is_causal, + causal_type=causal_type, + is_varlen=is_varlen, + raise_error=True, + ) + + scale = scale if scale is not None else query.shape[-1] ** -0.5 + + backend_kwargs = backend_kwargs.copy() if backend_kwargs is not None else {} + + natten_backend = None + if "backend" in backend_kwargs: + natten_backend = backend_kwargs["backend"] + del backend_kwargs["backend"] + else: + natten_backend = choose_natten_backend( + query, key, value, is_causal=is_causal, is_varlen=is_varlen, raise_error=True + ) + + assert natten_backend is not None + + # Override NATTEN's default delta reduction method: using PyTorch + # is more accurate, but slightly slower. + # Only affects NATTEN's "cutlass-fmha" backend (Ampere kernels) + backward_use_pt_reduction = True + if "backward_use_pt_reduction" in backend_kwargs: + backward_use_pt_reduction = backend_kwargs["backward_use_pt_reduction"] + del backend_kwargs["backward_use_pt_reduction"] + + return _natten_attention( + query=query, + key=key, + value=value, + is_causal=is_causal, + scale=scale, + cumulative_seqlen_Q=cumulative_seqlen_Q, + cumulative_seqlen_KV=cumulative_seqlen_KV, + max_seqlen_Q=max_seqlen_Q, + max_seqlen_KV=max_seqlen_KV, + return_lse=return_lse, + backend=natten_backend, + backward_use_pt_reduction=backward_use_pt_reduction, + **backend_kwargs, + ) + + +def natten_multi_dim_attention( + query: Tensor, + key: Tensor, + value: Tensor, + window_size: tuple | int = -1, + stride: tuple | int = 1, + dilation: tuple | int = 1, + is_causal: tuple | bool = False, + scale: float | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + """ + Runs NATTEN's Multi-Dimensional Attention on given operands (Q, K, V) with the heads-last + contiguous layout (`[batch, *, heads, head_dim]`). Supports up to and including 3 dimensions: + * 1-D: `[batch, X, heads, head_dim]`, with masking arguments expecting tuples of size 1. + * 2-D: `[batch, X, Y, heads, head_dim]`, with masking arguments expecting tuples of size 2. + * 3-D: `[batch, X, Y, Z, heads, head_dim]`, with masking arguments expecting tuples of size 3. + + Parameters: + query (Tensor): 4-D, 5-D, or 6-D query tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads, head_dim]`) + + key (Tensor): 4-D, 5-D, or 6-D key tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim]`) + + value (Tensor): 4-D, 5-D, or 6-D value tensor, with heads-last contiguous layout + (`[batch, *token_layout_shape, heads_kv, head_dim_v]`) + + window_size (tuple | int): Attention window (kernel) size / shape. If an + integer, it will be repeated for all dimensions. For example `window_size=3`, when + `len(token_layout_shape) == 3`, is interpreted as `window_size=(3, 3, 3)`. + `-1`s are replaced with the corresponding `token_layout_shape`. + Final window size must satisfy `2 <= window_size <= token_layout_shape`. + Default is -1 (no sparsity). + + stride (tuple | int): Sliding window step size/shape. If an integer, it will be repeated + for all dimensions. For example `stride=2`, when `len(token_layout_shape) == 3`, is + interpreted as `stride=(2, 2, 2)`. + Final stride must satisfy `1 <= stride <= window_size`. + Default is 1. + + dilation (tuple | int): Dilation step size/shape. If an integer, it will be repeated for + all dimensions. For example `dilation=4`, when `len(token_layout_shape) == 3`, is + interpreted as `dilation=(4, 4, 4)`. + Final dilation must satisfy `2 <= dilation * window_size <= token_layout_shape`. + Default is 1. + + is_causal (tuple | bool): Toggle causal masking. If a boolean, it will be repeated for all + dimensions. For example `is_causal=True`, when `len(token_layout_shape) == 3`, is + interpreted as `is_causal=(True, True, True)`. + Default is False. + + scale (float | None): Dot product scale (attention scale). Defaults to head_dim ** -0.5. + + Other Parameters: + return_lse (bool): Whether to return the logsumexp values. Default is False. + + backend_kwargs (dict | None): Key-value pair for passing arguments specific to NATTEN's + multi-dim / sparse attention operator, if any. + + Returns: + output (Tensor): 4-D, 5-D, or 6-D output tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads, head_dim_v]`). + + logsumexp (Tensor): logsumexp tensor, with the heads-last contiguous layout + (`[batch, *token_layout_shape, heads, 1]`). Only returned when return_lse is True. + """ + + assert natten_multi_dim_attention_check( + query=query, + key=key, + value=value, + raise_error=True, + ) + + token_layout, window_size, stride, dilation, is_causal = multi_dim_attention_param_filter( + query, + window_size=window_size, + stride=stride, + dilation=dilation, + is_causal=is_causal, + ) + + multi_dim_attention_param_checks( + query, + window_size=window_size, + stride=stride, + dilation=dilation, + is_causal=is_causal, + ) + + scale = scale if scale is not None else query.shape[-1] ** -0.5 + + backend_kwargs = backend_kwargs.copy() if backend_kwargs is not None else {} + + natten_backend = None + if "backend" in backend_kwargs: + natten_backend = backend_kwargs["backend"] + del backend_kwargs["backend"] + else: + natten_backend = choose_natten_multi_dim_backend(query, key, value, raise_error=True) + + assert natten_backend is not None + + # Override NATTEN's default delta reduction method: using PyTorch + # is more accurate, but slightly slower. + # Only affects NATTEN's "cutlass-fmha" backend (Ampere kernels) + backward_use_pt_reduction = True + if "backward_use_pt_reduction" in backend_kwargs: + backward_use_pt_reduction = backend_kwargs["backward_use_pt_reduction"] + del backend_kwargs["backward_use_pt_reduction"] + + output = _natten_multi_dim_attention( + query=query, + key=key, + value=value, + kernel_size=window_size, + stride=stride, + dilation=dilation, + is_causal=is_causal, + scale=scale, + backend=natten_backend, + backward_use_pt_reduction=backward_use_pt_reduction, + **backend_kwargs, + ) + + if return_lse: + raise NotImplementedError("NATTEN's Multi-Dimensional Attention does not support returning the logsumexp yet.") + + return output diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/meta.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/meta.py new file mode 100644 index 0000000000000000000000000000000000000000..b0c3118e17624bde734703c4040ddc22fa96da46 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/meta.py @@ -0,0 +1,67 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +NATTEN Backend: metadata +Always safe to import (as long as torch is available.) +""" + +import torch + +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + + +def get_fwd_dtypes(arch_tag: int) -> list[torch.dtype]: + """ + Returns data type choices for forward pass according to arch tag (attention.utils.get_arch_tag). + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + data_type_choices (list): a list of PyTorch data types. Empty if device is not supported. + + """ + + if arch_tag < 75: + log.debug("NATTEN is not supported because compute capability is below the minimum (7.5).") + return [] + + if arch_tag in [100, 103]: + return [torch.float32, torch.float16, torch.bfloat16, torch.float8_e5m2, torch.float8_e4m3fn] + + return [torch.float32, torch.float16, torch.bfloat16] + + +def get_bwd_dtypes(arch_tag: int) -> list[torch.dtype]: + """ + Returns data type choices for backward pass according to arch tag (attention.utils.get_arch_tag). + + Parameters: + arch_tag (int): Arch tag for the current CUDA device. Example: 80 for A100, 90 for H100. + + Returns: + data_type_choices (list): a list of PyTorch data types. Empty if device is not supported. + + """ + + if arch_tag < 75: + log.debug("NATTEN is not supported because compute capability is below the minimum (7.5).") + return [] + + return [torch.float32, torch.float16, torch.bfloat16] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/stubs.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/stubs.py new file mode 100644 index 0000000000000000000000000000000000000000..4edb2813a473c9df7388e3760d81bf9ab6bee9e8 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/natten/stubs.py @@ -0,0 +1,64 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +NATTEN Backend: intermediate API stubs +Always safe to import (as long as torch is available.) +""" + +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.masks import CausalType + + +def natten_attention( + query: Tensor, + key: Tensor, + value: Tensor, + is_causal: bool = False, + causal_type: CausalType | None = None, + scale: float | None = None, + cumulative_seqlen_Q: Tensor | None = None, + cumulative_seqlen_KV: Tensor | None = None, + max_seqlen_Q: int | None = None, + max_seqlen_KV: int | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + raise RuntimeError( + "Tried to run NATTEN attention, but it is not supported / available. " + "Try running with debug logs enabled to see why." + ) + + +def natten_multi_dim_attention( + query: Tensor, + key: Tensor, + value: Tensor, + window_size: tuple | int = -1, + stride: tuple | int = 1, + dilation: tuple | int = 1, + is_causal: tuple | bool = False, + scale: float | None = None, + return_lse: bool = False, + backend_kwargs: dict | None = None, +) -> Tensor | tuple[Tensor, Tensor]: + raise RuntimeError( + "Tried to run NATTEN's Multi-Dimensional attention, but it is not supported / available. " + "Try running with debug logs enabled to see why." + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/multi_dim_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/multi_dim_test.py new file mode 100644 index 0000000000000000000000000000000000000000..c68a2b3d2fbb159c2a0f8a708f549f8bd7a20559 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/multi_dim_test.py @@ -0,0 +1,503 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Multi-Dimensional Attention unit tests. +""" + +import math +import random +import unittest +from functools import partial +from itertools import product +from typing import Callable + +import pytest +import torch +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention import multi_dimensional_attention +from cosmos_policy._src.imaginaire.attention.natten import NATTEN_SUPPORTED +from cosmos_policy._src.imaginaire.attention.utils import is_blackwell_dc, is_fp8 +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +RAND_SWEEP_TESTS = 1000 + +skip_if_natten_not_supported = partial( + pytest.mark.skipif, + not NATTEN_SUPPORTED, + reason="NATTEN is disabled, not available, or too old in this environment.", +) + + +def _reset_everything(): + torch.manual_seed(42) + torch.cuda.empty_cache() + + +class MultiDimTester: + def __init__( + self, + reference_fn: Callable, + batch: int, + heads: int, + token_layout_shape: tuple, + head_dim: int, + window_size: tuple, + stride: tuple, + dilation: tuple, + is_causal: tuple, + test_backward: bool = True, + scale: float | None = None, + dtype: torch.dtype = torch.float32, + device: torch.device = "cuda", + heads_kv: int | None = None, + head_dim_v: int | None = None, + ): + self.batch = batch + self.heads = heads + self.heads_kv = heads_kv or heads + self.token_layout_shape = token_layout_shape + self.head_dim = head_dim + self.head_dim_v = head_dim_v or head_dim + self.test_backward = test_backward + self.scale = scale if scale is not None else head_dim**-0.5 + self.dtype = dtype + self.device = device + + self.window_size = window_size + self.stride = stride + self.dilation = dilation + self.is_causal = is_causal + + # Initialize input tensors + self.q = torch.randn( + self.batch, + *self.token_layout_shape, + self.heads, + self.head_dim, + dtype=dtype, + device=device, + requires_grad=test_backward, + ) + self.k = torch.randn( + self.batch, + *self.token_layout_shape, + self.heads_kv, + self.head_dim, + dtype=dtype, + device=device, + requires_grad=test_backward, + ) + self.v = torch.randn( + self.batch, + *self.token_layout_shape, + self.heads_kv, + self.head_dim_v, + dtype=dtype, + device=device, + requires_grad=test_backward, + ) + self.d_output = ( + torch.randn(self.batch, *self.token_layout_shape, self.heads, self.head_dim_v, dtype=dtype, device=device) + if test_backward + else None + ) + + # Run reference implementation + q_ref = self.q.clone().detach().requires_grad_(self.test_backward) + k_ref = self.k.clone().detach().requires_grad_(self.test_backward) + v_ref = self.v.clone().detach().requires_grad_(self.test_backward) + + output_ref = reference_fn( + query=q_ref, + key=k_ref, + value=v_ref, + scale=self.scale, + window_size=self.window_size, + stride=self.stride, + dilation=self.dilation, + is_causal=self.is_causal, + ) + + self.output_ref = output_ref.detach().to(torch.float32) + + # Reference backward pass + if self.test_backward: + d_output = self.d_output.clone().detach() + output_ref.backward(d_output) + self.dq_ref = q_ref.grad.detach().to(torch.float32) + self.dk_ref = k_ref.grad.detach().to(torch.float32) + self.dv_ref = v_ref.grad.detach().to(torch.float32) + + def test( + self, + target_fn: Callable, + dtype: torch.dtype, + atol_fwd: float, + atol_bwd: tuple[float, float, float] | None = None, + rtol_fwd: float = 0.0, + rtol_bwd: float = 0.0, + test_backward: bool | None = None, + ): + test_backward = self.test_backward if test_backward is None else test_backward + + q = self.q.clone().detach().to(dtype).requires_grad_(test_backward) + k = self.k.clone().detach().to(dtype).requires_grad_(test_backward) + v = self.v.clone().detach().to(dtype).requires_grad_(test_backward) + + output = target_fn( + query=q, + key=k, + value=v, + scale=self.scale, + window_size=self.window_size, + stride=self.stride, + dilation=self.dilation, + is_causal=self.is_causal, + ) + + torch.testing.assert_close(output.to(torch.float32), self.output_ref, atol=atol_fwd, rtol=rtol_fwd) + + # Backward pass + if test_backward: + assert atol_bwd is not None + assert rtol_bwd is not None + atol_dq, atol_dk, atol_dv = atol_bwd + + d_output = self.d_output.clone().detach().to(dtype) + output.backward(d_output) + + dq = q.grad.detach().to(torch.float32) + dk = k.grad.detach().to(torch.float32) + dv = v.grad.detach().to(torch.float32) + + torch.testing.assert_close(dq, self.dq_ref, atol=atol_dq, rtol=rtol_bwd) + torch.testing.assert_close(dk, self.dk_ref, atol=atol_dk, rtol=rtol_bwd) + torch.testing.assert_close(dv, self.dv_ref, atol=atol_dv, rtol=rtol_bwd) + + +def idx2crd(index, shape) -> tuple: + rank = len(shape) + coord = [] + residual = index + for i in range(rank - 1, -1, -1): + coord.append(residual % shape[i]) + residual = residual // shape[i] + + # assert residual == 0 + return tuple(coord[::-1]) + + +def multi_dim_mask( + q_idx: int, + kv_idx: int, + token_layout_shape: tuple, + window_size: tuple, + stride: tuple, + dilation: tuple, + is_causal: tuple, +) -> bool: + assert len(token_layout_shape) == len(window_size) == len(stride) == len(dilation) == len(is_causal) + + # Reconstruct global Q and KV coordinates + q_crd = idx2crd(q_idx, token_layout_shape) + kv_crd = idx2crd(kv_idx, token_layout_shape) + + masks = [] + for q, kv, x, w, s, d, c in zip(q_crd, kv_crd, token_layout_shape, window_size, stride, dilation, is_causal): + # Coordinates within dilation group + q_crd_di = q // d + kv_crd_di = kv // d + + # Dilation group coordinates + q_dilation_group_crd = q % d + kv_dilation_group_crd = kv % d + + # Fixup input shape according to dilation group + dilation_group_padding = 1 - ((q_dilation_group_crd + (d - (x % d))) // d) + qkv_shape_corrected = (x // d) + dilation_group_padding + + if c: + # Leader is the last (right-most) query in the stride group. + stride_group_leader = min( + (q_crd_di // s) * s + s - 1, + qkv_shape_corrected - 1, + ) + + if not ( + (q_crd_di - kv_crd_di >= 0) # window still ends at query index + and (stride_group_leader - kv_crd_di < w) + and (q_dilation_group_crd == kv_dilation_group_crd) + ): + return False + + else: + # Window size left and right (non-causal only) + window_size_left = w // 2 + window_size_right = w // 2 + (w % 2 - 1) + + # Leader is the center-most query in the stride group. + # If stride is even, choose the right hand side center query. + stride_group_leader = min( + (q_crd_di // s) * s + (s // 2), + qkv_shape_corrected - 1, + ) + + window_center = min(max(stride_group_leader, window_size_left), qkv_shape_corrected - 1 - window_size_right) + w0 = window_center - kv_crd_di + w1 = kv_crd_di - window_center + if not ( + (((0 <= w0) and (w0 <= window_size_left)) or ((0 <= w1) and (w1 <= window_size_right))) + and (q_dilation_group_crd == kv_dilation_group_crd) + ): + return False + + return True + + +def multi_dim_reference( + query: Tensor, + key: Tensor, + value: Tensor, + window_size: tuple, + stride: tuple, + dilation: tuple, + is_causal: tuple, + scale: float, +): + assert query.dim() in [4, 5, 6] + B, *token_layout_shape, H, D = query.shape + H_kv, _ = key.shape[-2:] + D_v = value.shape[-1] + seqlen = math.prod(token_layout_shape) + + # cast from torch shape to tuple + token_layout_shape = tuple(x for x in token_layout_shape) + + num_dims = len(token_layout_shape) + + assert H % H_kv == 0 + h_k = H // H_kv + + query_t = query.flatten(1, num_dims).transpose(1, 2) + key_t = key.flatten(1, num_dims).transpose(1, 2) + value_t = value.flatten(1, num_dims).transpose(1, 2) + + assert query_t.dim() == key_t.dim() == value_t.dim() == 4 + assert query_t.shape[2] == key_t.shape[2] == value_t.shape[2] == seqlen + + # Decomposed GQA/MQA implementation + if h_k > 1: + key_t = torch.repeat_interleave(key_t, repeats=h_k, dim=1, output_size=H) + value_t = torch.repeat_interleave(value_t, repeats=h_k, dim=1, output_size=H) + + attn_scores = torch.matmul(query_t, key_t.transpose(-2, -1)) * scale + + mask = torch.zeros((seqlen, seqlen), dtype=torch.bool) + is_valid = partial( + multi_dim_mask, + token_layout_shape=token_layout_shape, + window_size=window_size, + stride=stride, + dilation=dilation, + is_causal=is_causal, + ) + for q, kv in product(range(mask.shape[0]), range(mask.shape[1])): + mask[q, kv] = not is_valid(q, kv) + + mask_cu = mask.unsqueeze(0).unsqueeze(0).to(attn_scores.device) + attn_scores = attn_scores.masked_fill(mask_cu, float("-inf")) + + attn_weights = attn_scores.softmax(dim=-1) + + out = torch.matmul(attn_weights, value_t) + + out = out.transpose(1, 2) + + out = out.reshape(B, *token_layout_shape, H, D_v) + + return out + + +class MultiDimTest(unittest.TestCase): + def setUp(self): + _reset_everything() + + def tearDown(self): + _reset_everything() + + def _test_against_bmm_reference( + self, + batch: int, + heads: int, + token_layout_shape: tuple, + head_dim: int, + window_size: tuple, + stride: tuple, + dilation: tuple, + is_causal: tuple, + test_backward: bool, + backend: str, + scale: float | None = None, + heads_kv: int | None = None, + head_dim_v: int | None = None, + ): + reference_dtype = torch.float16 + device = "cuda" + attention_fn = partial(multi_dimensional_attention, backend=backend) + + log.debug( + "Running reference Multi-Dimensional Attention on: " + f"{batch=}, {heads=}, {heads_kv=}, {head_dim=}, {head_dim_v=}, " + f"{token_layout_shape=}, {window_size=}, {stride=}, {dilation=}, {is_causal=}." + ) + tester = MultiDimTester( + reference_fn=multi_dim_reference, + batch=batch, + heads=heads, + heads_kv=heads_kv, + token_layout_shape=token_layout_shape, + head_dim=head_dim, + head_dim_v=head_dim_v, + window_size=window_size, + stride=stride, + dilation=dilation, + is_causal=is_causal, + dtype=reference_dtype, + test_backward=test_backward, + scale=scale, + device=device, + ) + + ALLOWED_DTYPES = [ + # dtype, atol_out, (atol_dq, atol_dk, atol_dv), rtol_fwd, rtol_bwd + (torch.float16, 1e-2, (4e-2, 4e-2, 4e-2), 0, 0), + (torch.bfloat16, 1e-1, (2e-1, 2e-1, 2e-1), 0, 0), + ] + if backend == "natten" and is_blackwell_dc(): + ALLOWED_DTYPES += [ + (torch.float8_e4m3fn, 4e-1, None, 1e-1, 0), + (torch.float8_e5m2, 8e-1, None, 5e-1, 0), + ] + + for dtype, atol_fwd, atol_bwd, rtol_fwd, rtol_bwd in ALLOWED_DTYPES: + test_backward_ = test_backward and not is_fp8(dtype) + log.debug( + f"Testing Multi-Dimensional Attention ({backend}): {batch=}, {heads=}, {heads_kv=}, {head_dim=}, {head_dim_v=}, " + f"{token_layout_shape=}, {window_size=}, {stride=}, {dilation=}, " + f"{is_causal=}, {dtype=}, {test_backward_=}." + ) + tester.test( + target_fn=attention_fn, + dtype=dtype, + atol_fwd=atol_fwd, + atol_bwd=atol_bwd, + rtol_fwd=rtol_fwd, + rtol_bwd=rtol_bwd, + test_backward=test_backward_, + ) + + def _test_randsweep(self, num_dims: int, backend: str, max_tests: int = 1000, max_seqlen: int = 2**17): + random.seed(42) + + for i in range(max_tests): + batch = random.choice(range(1, 2)) + + supports_mla = False + supports_gqa_mqa = False + if backend == "natten": + head_dim_choices = [32, 64, 128] + heads_choices = range(1, 4 + 1) + # GQA/MQA is not supported in FNA ops yet + supports_gqa_mqa = False + + # Enable MLA when supported in hopper or blackwell + head_dim = random.choice(head_dim_choices) + head_dim_v = None + # head_dim_v = random.choice(head_dim_choices) + + else: + raise NotImplementedError() + + heads = random.choice(heads_choices) + heads_kv = ( + heads + if not supports_gqa_mqa + else random.choice([1] + [i for i in range(1, heads + 1) if heads % i == 0]) + ) + assert heads >= heads_kv and heads % heads_kv == 0 + + token_layout_shape = [] + for j in range(num_dims): + max_size = ( + min(max_seqlen, 16384) + if j == 0 + else min(16384, max(10, max_seqlen - math.prod(token_layout_shape))) + ) + token_layout_shape.append(random.choice(range(4, max_size))) + + while math.prod(token_layout_shape) > max_seqlen: + dim_to_cut = random.choice(range(num_dims)) + token_layout_shape[dim_to_cut] = max(4, int(token_layout_shape[dim_to_cut] * 0.1)) + + token_layout_shape = tuple(token_layout_shape) + window_size = tuple(random.choice(range(2, x + 1)) for x in token_layout_shape) + stride = tuple(random.choice(range(1, k + 1)) for k in window_size) + dilation = tuple(random.choice(range(1, x // k + 1)) for x, k in zip(token_layout_shape, window_size)) + is_causal = tuple(random.choice([False, True]) for _ in range(num_dims)) + + self._test_against_bmm_reference( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + token_layout_shape=token_layout_shape, + window_size=window_size, + stride=stride, + dilation=dilation, + is_causal=is_causal, + backend=backend, + test_backward=True, + ) + + @pytest.mark.L1 + @skip_if_natten_not_supported() + def test_natten_fast(self): + random.seed(83) + torch.manual_seed(83) + self._test_randsweep(num_dims=1, backend="natten", max_tests=10, max_seqlen=2**10) + self._test_randsweep(num_dims=2, backend="natten", max_tests=10, max_seqlen=2**10) + self._test_randsweep(num_dims=3, backend="natten", max_tests=10, max_seqlen=2**10) + + @pytest.mark.L1 + @pytest.mark.skip("Extended rand sweep is disabled until we have a faster reference for multi-dim") + @skip_if_natten_not_supported() + def test_natten_randsweep(self): + random.seed(84) + torch.manual_seed(84) + self._test_randsweep(num_dims=1, backend="natten", max_tests=RAND_SWEEP_TESTS // 3, max_seqlen=2**11) + self._test_randsweep(num_dims=2, backend="natten", max_tests=RAND_SWEEP_TESTS // 3, max_seqlen=2**11) + self._test_randsweep(num_dims=3, backend="natten", max_tests=RAND_SWEEP_TESTS // 3, max_seqlen=2**11) + + +if __name__ == "__main__": + random.seed(42) + torch.manual_seed(42) + unittest.main() diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/sdpa_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/sdpa_test.py new file mode 100644 index 0000000000000000000000000000000000000000..6389438b0edec48ef1299807143b9a3eb27daea5 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/sdpa_test.py @@ -0,0 +1,1015 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +SDPA unit tests. +""" + +import random +import unittest +from functools import partial +from typing import Callable + +import pytest +import torch +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention import attention as i4_attention +from cosmos_policy._src.imaginaire.attention.cudnn import CUDNN_DISALLOWED, CUDNN_SUPPORTED +from cosmos_policy._src.imaginaire.attention.flash2 import FLASH2_SUPPORTED +from cosmos_policy._src.imaginaire.attention.flash3 import FLASH3_SUPPORTED +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.natten import NATTEN_SUPPORTED +from cosmos_policy._src.imaginaire.attention.utils import is_blackwell_dc, is_fp8, is_hopper +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +RAND_SWEEP_TESTS = 1000 + +skip_if_cudnn_not_supported = partial( + pytest.mark.skipif, + CUDNN_DISALLOWED or not CUDNN_SUPPORTED, + reason="cuDNN is disabled, not available, or too old in this environment.", +) + +skip_if_natten_not_supported = partial( + pytest.mark.skipif, + not NATTEN_SUPPORTED, + reason="NATTEN is disabled, not available, or too old in this environment.", +) + +skip_if_flash2_not_supported = partial( + pytest.mark.skipif, + not FLASH2_SUPPORTED, + reason="Flash2 is disabled, not available, or too old in this environment.", +) + +skip_if_flash3_not_supported = partial( + pytest.mark.skipif, + not FLASH3_SUPPORTED, + reason="Flash3 is disabled, not available, or too old in this environment.", +) + +# Tests are only enabled on Hopper and Blackwell DC-class for now. +# Will extend to other arches as we integrate more backends. +skip_if_not_supported = partial( + pytest.mark.skipif, + not is_blackwell_dc() and not is_hopper(), + reason="SDPA tests are only allowed for Hopper and Blackwell DC-class GPUs for now.", +) + +skip_if_not_blackwell = partial( + pytest.mark.skipif, not is_blackwell_dc(), reason="This test is only allowed for Blackwell DC-class GPUs." +) + +skip_if_not_hopper = partial(pytest.mark.skipif, not is_hopper(), reason="This test is only allowed for Hopper GPUs.") + + +def _reset_everything(): + torch.manual_seed(42) + torch.cuda.empty_cache() + + +class SdpaTester: + def __init__( + self, + reference_fn: Callable, + batch: int, + heads: int, + seqlen_q: int, + seqlen_kv: int, + head_dim: int, + test_backward: bool = True, + scale: float | None = None, + is_causal: bool = False, + causal_type: CausalType | None = None, + dtype: torch.dtype = torch.float32, + device: torch.device = "cuda", + heads_kv: int | None = None, + head_dim_v: int | None = None, + ): + self.batch = batch + self.heads = heads + self.heads_kv = heads_kv or heads + self.seqlen_q = seqlen_q + self.seqlen_kv = seqlen_kv + self.head_dim = head_dim + self.head_dim_v = head_dim_v or head_dim + self.test_backward = test_backward + self.scale = scale if scale is not None else head_dim**-0.5 + self.is_causal = is_causal + self.causal_type = causal_type + self.dtype = dtype + self.device = device + + # Initialize input tensors + self.q = torch.randn( + self.batch, + self.seqlen_q, + self.heads, + self.head_dim, + dtype=dtype, + device=device, + requires_grad=test_backward, + ) + self.k = torch.randn( + self.batch, + self.seqlen_kv, + self.heads_kv, + self.head_dim, + dtype=dtype, + device=device, + requires_grad=test_backward, + ) + self.v = torch.randn( + self.batch, + self.seqlen_kv, + self.heads_kv, + self.head_dim_v, + dtype=dtype, + device=device, + requires_grad=test_backward, + ) + self.d_output = ( + torch.randn(self.batch, self.seqlen_q, self.heads, self.head_dim_v, dtype=dtype, device=device) + if test_backward + else None + ) + + # Run reference implementation + q_ref = self.q.clone().detach().requires_grad_(self.test_backward) + k_ref = self.k.clone().detach().requires_grad_(self.test_backward) + v_ref = self.v.clone().detach().requires_grad_(self.test_backward) + + output_ref = reference_fn( + query=q_ref, + key=k_ref, + value=v_ref, + scale=self.scale, + is_causal=self.is_causal, + causal_type=self.causal_type, + ) + + self.output_ref = output_ref.detach().to(torch.float32) + + # Reference backward pass + if self.test_backward: + d_output = self.d_output.clone().detach() + output_ref.backward(d_output) + self.dq_ref = q_ref.grad.detach().to(torch.float32) + self.dk_ref = k_ref.grad.detach().to(torch.float32) + self.dv_ref = v_ref.grad.detach().to(torch.float32) + + def test( + self, + target_fn: Callable, + dtype: torch.dtype, + atol_fwd: float, + atol_bwd: tuple[float, float, float] | None = None, + rtol_fwd: float = 0.0, + rtol_bwd: float = 0.0, + test_backward: bool | None = None, + ): + test_backward = self.test_backward if test_backward is None else test_backward + + q = self.q.clone().detach().to(dtype).requires_grad_(test_backward) + k = self.k.clone().detach().to(dtype).requires_grad_(test_backward) + v = self.v.clone().detach().to(dtype).requires_grad_(test_backward) + + output = target_fn( + query=q, key=k, value=v, scale=self.scale, is_causal=self.is_causal, causal_type=self.causal_type + ) + + torch.testing.assert_close(output.to(torch.float32), self.output_ref, atol=atol_fwd, rtol=rtol_fwd) + + # Backward pass + if test_backward: + assert atol_bwd is not None + assert rtol_bwd is not None + atol_dq, atol_dk, atol_dv = atol_bwd + + d_output = self.d_output.clone().detach().to(dtype) + output.backward(d_output) + + dq = q.grad.detach().to(torch.float32) + dk = k.grad.detach().to(torch.float32) + dv = v.grad.detach().to(torch.float32) + + torch.testing.assert_close(dq, self.dq_ref, atol=atol_dq, rtol=rtol_bwd) + torch.testing.assert_close(dk, self.dk_ref, atol=atol_dk, rtol=rtol_bwd) + torch.testing.assert_close(dv, self.dv_ref, atol=atol_dv, rtol=rtol_bwd) + + +def torch_sdpa_reference( + query: Tensor, + key: Tensor, + value: Tensor, + scale: float, + is_causal: bool, + causal_type: CausalType, +): + heads = query.shape[2] + heads_kv = key.shape[2] + assert heads % heads_kv == 0 + h_k = heads // heads_kv + + assert not is_causal or causal_type == CausalType.TopLeft, "Torch SDPA only supports top-left causal mask." + + # Torch requires heads-first layout + query = query.permute(0, 2, 1, 3).contiguous() + key = key.permute(0, 2, 1, 3).contiguous() + value = value.permute(0, 2, 1, 3).contiguous() + + k_final, v_final = key, value + # Decomposed GQA/MQA implementation for torch SDPA via explicit repeats + if h_k > 1: + k_final = torch.repeat_interleave(key, repeats=h_k, dim=1, output_size=heads) + v_final = torch.repeat_interleave(value, repeats=h_k, dim=1, output_size=heads) + + assert k_final.shape[:2] == query.shape[:2] + assert v_final.shape[:2] == query.shape[:2] + assert k_final.shape[-1] == query.shape[-1] + assert v_final.shape[-1] == query.shape[-1] + + with torch.nn.attention.sdpa_kernel(backends=[torch.nn.attention.SDPBackend.EFFICIENT_ATTENTION]): + out = torch.nn.functional.scaled_dot_product_attention( + query, k_final, v_final, is_causal=is_causal, scale=scale + ) + + out = out.permute(0, 2, 1, 3).contiguous() + return out + + +# NOTE: Use ONLY when seqlen_{q,kv} are small! +# Supports MLA and bottom-right causal mask, unlike SDPA +def bmm_sdpa_reference( + query: Tensor, + key: Tensor, + value: Tensor, + scale: float, + is_causal: bool, + causal_type: CausalType, + MAX_QK: int = 16384**2, +): + B, S_q, H, D = query.shape + _, S_kv, H_kv, _ = key.shape + + assert H % H_kv == 0 + h_k = H // H_kv + + if S_q * S_kv > MAX_QK: + raise ValueError(f"Query-key matmul too large: {S_q}*{S_kv} > MAX_QK={MAX_QK}") + + query_t = query.transpose(1, 2) + key_t = key.transpose(1, 2) + value_t = value.transpose(1, 2) + + # Decomposed GQA/MQA implementation + if h_k > 1: + key_t = torch.repeat_interleave(key_t, repeats=h_k, dim=1, output_size=H) + value_t = torch.repeat_interleave(value_t, repeats=h_k, dim=1, output_size=H) + + attn_scores = torch.matmul(query_t, key_t.transpose(-2, -1)) * scale + + if is_causal: + if causal_type == CausalType.TopLeft: + diagonal_offset = 1 + elif causal_type == CausalType.BottomRight: + diagonal_offset = S_kv - S_q + 1 + else: + raise NotImplementedError() + mask = torch.triu(torch.ones(S_q, S_kv, device=query.device, dtype=torch.bool), diagonal=diagonal_offset) + attn_scores = attn_scores.masked_fill(mask, float("-inf")) + + attn_weights = attn_scores.softmax(dim=-1) + + # We can have entirely masked rows (queries) with this mask + if causal_type == CausalType.BottomRight: + attn_weights = torch.nan_to_num(attn_weights, nan=0.0) + + out = torch.matmul(attn_weights, value_t) + + out = out.transpose(1, 2) + + return out + + +class SdpaTest(unittest.TestCase): + def setUp(self): + _reset_everything() + + def tearDown(self): + _reset_everything() + + def _test_against_torch_sdpa( + self, + batch: int, + heads: int, + head_dim: int, + seqlen_q: int, + seqlen_kv: int, + is_causal: bool, + causal_type: CausalType, + test_backward: bool, + backend: str, + scale: float | None = None, + heads_kv: int | None = None, + head_dim_v: int | None = None, + ): + reference_dtype = torch.float16 + device = "cuda" + attention_fn = partial(i4_attention, backend=backend) + + reference_fn = torch_sdpa_reference + if (is_causal and causal_type == CausalType.BottomRight) or (head_dim_v is not None and head_dim_v != head_dim): + reference_fn = bmm_sdpa_reference + + tester = SdpaTester( + reference_fn=reference_fn, + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlen_q=seqlen_q, + seqlen_kv=seqlen_kv, + dtype=reference_dtype, + test_backward=test_backward, + scale=scale, + is_causal=is_causal, + causal_type=causal_type, + device=device, + ) + + ALLOWED_DTYPES = [ + # dtype, atol_out, (atol_dq, atol_dk, atol_dv), rtol_fwd, rtol_bwd + (torch.float16, 1e-2, (4e-2, 4e-2, 4e-2), 0, 0), + (torch.bfloat16, 1e-1, (2e-1, 2e-1, 2e-1), 0, 0), + ] + if backend == "natten" and is_blackwell_dc(): + ALLOWED_DTYPES += [ + (torch.float8_e4m3fn, 4e-1, None, 1e-1, 0), + (torch.float8_e5m2, 8e-1, None, 5e-1, 0), + ] + + for dtype, atol_fwd, atol_bwd, rtol_fwd, rtol_bwd in ALLOWED_DTYPES: + test_backward_ = test_backward and not is_fp8(dtype) + log.debug( + f"Testing SDPA ({backend}) vs torch SDPA: {batch=}, {heads=}, {heads_kv=}, {head_dim=}, {head_dim_v=}, " + f"{seqlen_q=}, {seqlen_kv=}, {is_causal=}, {causal_type=}, {dtype=}, {test_backward_=}" + ) + tester.test( + target_fn=attention_fn, + dtype=dtype, + atol_fwd=atol_fwd, + atol_bwd=atol_bwd, + rtol_fwd=rtol_fwd, + rtol_bwd=rtol_bwd, + test_backward=test_backward_, + ) + + def _test_backend_against_torch_sdpa( + self, + batch: int, + heads: int, + head_dim: int, + seqlen_q: int, + seqlen_kv: int, + is_causal: bool, + backend: str, + scale: float | None = None, + heads_kv: int | None = None, + head_dim_v: int | None = None, + ): + assert backend in ["natten", "flash2", "flash3", "cudnn"] + self._test_against_torch_sdpa( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlen_q=seqlen_q, + seqlen_kv=seqlen_kv, + is_causal=is_causal, + causal_type=CausalType.TopLeft if backend not in ["flash2", "flash3"] else CausalType.BottomRight, + scale=scale, + test_backward=backend != "cudnn", + backend=backend, + ) + + def _test_randsweep_against_torch_sdpa(self, backend: str, max_tests: int = 1000): + random.seed(42) + + max_qk = 2**21 + for i in range(max_tests): + batch = random.choice(range(1, 4)) + + supports_mla = False + supports_gqa_mqa = False + if backend == "natten": + head_dim_choices = [32, 64, 128] + heads_choices = range(1, 8 + 1) + # GQA/MQA is only supported in NATTEN's Blackwell FMHA backend for now + supports_gqa_mqa = is_blackwell_dc() + + # Enable MLA when supported in hopper or blackwell + head_dim = random.choice(head_dim_choices) + head_dim_v = None + # head_dim_v = random.choice(head_dim_choices) + + elif backend in ["flash2", "flash3"]: + head_dim_choices = range(16, 256 + 1, 8) + heads_choices = range(1, 8 + 1) + supports_gqa_mqa = True + + # NOTE: Flash 3 MLA fails a static check in bwd, seems like an FA bug + ## Flash 3 supports MLA, but with some extra constraints + # if backend == "flash3" and random.choice([True, False]): + # # Either head_dim_qk <= 64 and head_dim_v <= 512, or + # # 128 <= head_dim_qk <= 192 and 96 <= head_dim_v <= 128 + # if random.choice([True, False]): + # head_dim = random.choice(range(16, 64 + 1, 8)) + # head_dim_v = random.choice(head_dim_choices) + # else: + # head_dim = random.choice(range(128, 192 + 1, 8)) + # head_dim_v = random.choice(range(96, 128 + 1, 8)) + + # else: + head_dim = random.choice(head_dim_choices) + head_dim_v = None + + elif backend == "cudnn": + head_dim_choices = [32, 64, 128] + heads_choices = range(1, 4) + + # Enable MLA when verified + head_dim = random.choice(head_dim_choices) + head_dim_v = None + # head_dim_v = random.choice(head_dim_choices) + + else: + raise NotImplementedError() + + heads = random.choice(heads_choices) + heads_kv = ( + heads + if not supports_gqa_mqa + else random.choice([1] + [i for i in range(1, heads + 1) if heads % i == 0]) + ) + assert heads >= heads_kv and heads % heads_kv == 0 + + seqlen_q = random.choice(range(16, 2**14, 1)) + seqlen_kv = random.choice(range(16, 2**14, 1)) + + is_causal = random.choice([True, False]) + + while seqlen_q * seqlen_kv > max_qk: + cut_kv = random.choice([True, False]) + if cut_kv: + seqlen_kv = int(seqlen_kv * 0.75) + else: + seqlen_q = int(seqlen_q * 0.75) + + self._test_backend_against_torch_sdpa( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlen_q=seqlen_q, + seqlen_kv=seqlen_kv, + is_causal=is_causal, + backend=backend, + ) + + @pytest.mark.L1 + @skip_if_cudnn_not_supported() + @skip_if_not_blackwell() + def test_cudnn_fast(self): + problem_sizes = [ + #### fp16 NaN??!! + #### batch=1, heads=13, head_dim=128, seqlen_q=7688, seqlen_kv=256, is_causal=False, dtype=torch.float16 + (1, 8, 128, 16384, 16384), + (1, 13, 128, 7688, 256), + #### illegal mem access -- seems intermittent + ## batch=6, heads=2, head_dim=128, seqlen_q=12244, seqlen_kv=123, is_causal=False, dtype=torch.float16 + (6, 2, 128, 12244, 123), + ##### + (2, 1, 128, 2048, 2048), + (2, 1, 64, 2048, 2048), + (4, 1, 64, 2048, 2048), + ### Failing FP16 case: + ### batch=2, heads=1, head_dim=64, seqlen_q=1411, seqlen_kv=1375, is_causal=False, dtype=torch.float16 + (1, 1, 64, 1411, 1375), + (2, 1, 64, 1536, 1280), + (2, 1, 64, 1536, 1536), + (2, 1, 64, 1536, 1376), + (2, 1, 64, 1416, 1376), + (2, 1, 64, 1411, 1375), + ### NaN case + ### batch=3, heads=3, head_dim=64, seqlen_q=9197, seqlen_kv=166, + (1, 1, 64, 10240, 512), + (2, 1, 64, 10240, 512), + (4, 1, 64, 10240, 512), + (8, 1, 64, 10240, 512), + ##### + (3, 1, 64, 10240, 512), + (4, 1, 64, 10240, 512), + (5, 1, 64, 10240, 512), + (6, 1, 64, 10240, 512), + (7, 1, 64, 10240, 512), + (8, 1, 64, 10240, 512), + (3, 3, 64, 10240, 512), + (3, 3, 64, 9216, 512), + (3, 3, 64, 9200, 512), + (3, 3, 64, 9200, 512), + (3, 3, 64, 9200, 512), + (3, 3, 64, 9200, 512), + (3, 3, 64, 9198, 512), + (3, 3, 64, 9197, 512), + # + (3, 1, 64, 10240, 256), + (4, 1, 64, 10240, 256), + (5, 1, 64, 10240, 256), + (6, 1, 64, 10240, 256), + (7, 1, 64, 10240, 256), + (8, 1, 64, 10240, 256), + (3, 3, 64, 10240, 256), + (3, 3, 64, 9216, 256), + (3, 3, 64, 9200, 256), + (3, 3, 64, 9200, 192), + (3, 3, 64, 9200, 168), + (3, 3, 64, 9200, 166), + (3, 3, 64, 9198, 166), + (3, 3, 64, 9197, 166), + # Passing: + (4, 1, 64, 10240, 10240), + (4, 1, 64, 10240, 1024), + (1, 1, 64, 9197, 166), + (3, 3, 64, 2560, 256), + # + (1, 1, 128, 128, 128), + (2, 1, 128, 128, 128), + (1, 2, 128, 128, 128), + (2, 2, 128, 128, 128), + (2, 2, 64, 128, 128), + (1, 1, 32, 32, 32), + (1, 1, 32, 128, 128), + (1, 1, 32, 128, 128), + (1, 1, 128, 128, 64), + (1, 1, 32, 128, 258), + (1, 2, 64, 128, 15), + (1, 1, 32, 8, 17), + (1, 1, 64, 17, 49), + (2, 4, 32, 128, 237), + (4, 3, 64, 256, 33), + (1, 1, 128, 128, 75), + (1, 1, 32, 125, 444), + (1, 2, 64, 125, 231), + (1, 1, 128, 256, 10240), + (1, 1, 32, 128, 4096), + (1, 1, 128, 3584, 381), + (1, 1, 128, 12072, 1680), + ] + for ( + batch, + heads, + head_dim, + seqlen_q, + seqlen_kv, + ) in problem_sizes: + for is_causal in [False, True]: + self._test_backend_against_torch_sdpa( + batch=batch, + heads=heads, + head_dim=head_dim, + seqlen_q=seqlen_q, + seqlen_kv=seqlen_kv, + is_causal=is_causal, + backend="cudnn", + ) + + @pytest.mark.L1 + @skip_if_cudnn_not_supported() + @skip_if_not_blackwell() + def test_cudnn_randsweep(self): + self._test_randsweep_against_torch_sdpa(backend="cudnn", max_tests=RAND_SWEEP_TESTS) + + @pytest.mark.L1 + @skip_if_natten_not_supported() + @skip_if_not_blackwell() + def test_natten_blackwell_fast(self): + problem_sizes = [ + (1, 8, 8, 128, 16384, 16384), + (1, 8, 4, 128, 16384, 16384), + (1, 8, 2, 128, 16384, 16384), + (1, 8, 1, 128, 16384, 16384), + (1, 12, 12, 128, 7688, 256), + (1, 12, 6, 128, 7688, 256), + (1, 12, 4, 128, 7688, 256), + (1, 12, 3, 128, 7688, 256), + (1, 12, 2, 128, 7688, 256), + (1, 12, 1, 128, 7688, 256), + (6, 2, 2, 128, 12244, 123), + (6, 2, 1, 128, 12244, 123), + (2, 1, 1, 128, 2048, 2048), + (2, 1, 1, 64, 2048, 2048), + (4, 1, 1, 64, 2048, 2048), + (1, 1, 1, 64, 1411, 1375), + (2, 1, 1, 64, 1536, 1280), + (2, 1, 1, 64, 1536, 1536), + (2, 1, 1, 64, 1536, 1376), + (2, 1, 1, 64, 1416, 1376), + (2, 1, 1, 64, 1411, 1375), + (1, 1, 1, 64, 10240, 512), + (2, 1, 1, 64, 10240, 512), + (4, 1, 1, 64, 10240, 512), + (8, 1, 1, 64, 10240, 512), + (3, 1, 1, 64, 10240, 512), + (4, 1, 1, 64, 10240, 512), + (5, 1, 1, 64, 10240, 512), + (6, 1, 1, 64, 10240, 512), + (7, 1, 1, 64, 10240, 512), + (8, 1, 1, 64, 10240, 512), + (3, 3, 3, 64, 9197, 512), + (3, 3, 1, 64, 9197, 512), + (7, 1, 1, 64, 10240, 256), + (3, 3, 3, 64, 10240, 256), + (3, 3, 3, 64, 9216, 256), + (3, 3, 3, 64, 9200, 256), + (3, 3, 3, 64, 9200, 192), + (3, 3, 3, 64, 9200, 168), + (3, 3, 3, 64, 9200, 166), + (3, 3, 3, 64, 9198, 166), + (3, 3, 3, 64, 9197, 166), + (4, 1, 1, 64, 10240, 10240), + (4, 1, 1, 64, 10240, 1024), + (1, 1, 1, 64, 9197, 166), + (3, 3, 3, 64, 2560, 256), + (1, 1, 1, 128, 128, 128), + (2, 1, 1, 128, 128, 128), + (1, 2, 2, 128, 128, 128), + (2, 2, 2, 128, 128, 128), + (2, 2, 2, 64, 128, 128), + (1, 1, 1, 32, 32, 32), + (1, 1, 1, 32, 128, 128), + (1, 1, 1, 32, 128, 128), + (1, 1, 1, 128, 128, 64), + (1, 1, 1, 32, 128, 258), + (1, 2, 2, 64, 128, 15), + (1, 1, 1, 32, 8, 17), + (1, 1, 1, 64, 17, 49), + (2, 4, 4, 32, 128, 237), + (2, 4, 2, 32, 128, 237), + (2, 4, 1, 32, 128, 237), + (4, 3, 3, 64, 256, 33), + (4, 3, 1, 64, 256, 33), + (1, 1, 1, 128, 128, 75), + (1, 1, 1, 32, 125, 444), + (1, 2, 2, 64, 125, 231), + (1, 2, 1, 64, 125, 231), + (1, 1, 1, 128, 256, 10240), + (1, 1, 1, 32, 128, 4096), + (1, 1, 1, 128, 3584, 381), + (1, 1, 1, 128, 12072, 1680), + ] + for ( + batch, + heads, + heads_kv, + head_dim, + seqlen_q, + seqlen_kv, + ) in problem_sizes: + for is_causal in [False, True]: + self._test_backend_against_torch_sdpa( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + seqlen_q=seqlen_q, + seqlen_kv=seqlen_kv, + is_causal=is_causal, + backend="natten", + ) + + @pytest.mark.L1 + @skip_if_natten_not_supported() + @skip_if_not_hopper() + def test_natten_hopper_fast(self): + # No GQA/MQA + # No MLA (except when using Ampere kernels) + # No causal masking (except when using Ampere kernels) + problem_sizes = [ + (1, 8, 8, 128, 16384, 16384), + (1, 12, 12, 128, 7688, 256), + (6, 2, 2, 128, 12244, 123), + (2, 1, 1, 128, 2048, 2048), + (2, 1, 1, 64, 2048, 2048), + (4, 1, 1, 64, 2048, 2048), + (1, 1, 1, 64, 1411, 1375), + (2, 1, 1, 64, 1536, 1280), + (2, 1, 1, 64, 1536, 1536), + (2, 1, 1, 64, 1536, 1376), + (2, 1, 1, 64, 1416, 1376), + (2, 1, 1, 64, 1411, 1375), + (1, 1, 1, 64, 10240, 512), + (2, 1, 1, 64, 10240, 512), + (4, 1, 1, 64, 10240, 512), + (8, 1, 1, 64, 10240, 512), + (3, 1, 1, 64, 10240, 512), + (4, 1, 1, 64, 10240, 512), + (5, 1, 1, 64, 10240, 512), + (6, 1, 1, 64, 10240, 512), + (7, 1, 1, 64, 10240, 512), + (8, 1, 1, 64, 10240, 512), + (3, 3, 3, 64, 9197, 512), + (7, 1, 1, 64, 10240, 256), + (3, 3, 3, 64, 10240, 256), + (3, 3, 3, 64, 9216, 256), + (3, 3, 3, 64, 9200, 256), + (3, 3, 3, 64, 9200, 192), + (3, 3, 3, 64, 9200, 168), + (3, 3, 3, 64, 9200, 166), + (3, 3, 3, 64, 9198, 166), + (3, 3, 3, 64, 9197, 166), + (4, 1, 1, 64, 10240, 10240), + (4, 1, 1, 64, 10240, 1024), + (1, 1, 1, 64, 9197, 166), + (3, 3, 3, 64, 2560, 256), + (1, 1, 1, 128, 128, 128), + (2, 1, 1, 128, 128, 128), + (1, 2, 2, 128, 128, 128), + (2, 2, 2, 128, 128, 128), + (2, 2, 2, 64, 128, 128), + (1, 1, 1, 32, 32, 32), + (1, 1, 1, 32, 128, 128), + (1, 1, 1, 32, 128, 128), + (1, 1, 1, 128, 128, 64), + (1, 1, 1, 32, 128, 258), + (1, 2, 2, 64, 128, 15), + (1, 1, 1, 32, 8, 17), + (1, 1, 1, 64, 17, 49), + (2, 4, 4, 32, 128, 237), + (4, 3, 3, 64, 256, 33), + (1, 1, 1, 128, 128, 75), + (1, 1, 1, 32, 125, 444), + (1, 2, 2, 64, 125, 231), + (1, 1, 1, 128, 256, 10240), + (1, 1, 1, 32, 128, 4096), + (1, 1, 1, 128, 3584, 381), + (1, 1, 1, 128, 12072, 1680), + ] + for ( + batch, + heads, + heads_kv, + head_dim, + seqlen_q, + seqlen_kv, + ) in problem_sizes: + for is_causal in [False, True]: + self._test_backend_against_torch_sdpa( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + seqlen_q=seqlen_q, + seqlen_kv=seqlen_kv, + is_causal=is_causal, + backend="natten", + ) + + @pytest.mark.L1 + @skip_if_natten_not_supported() + @skip_if_not_supported() + def test_natten_randsweep(self): + self._test_randsweep_against_torch_sdpa(backend="natten", max_tests=RAND_SWEEP_TESTS) + + @pytest.mark.L1 + @skip_if_flash2_not_supported() + @skip_if_not_supported() + def test_flash2_fast(self): + problem_sizes = [ + (1, 8, 8, 128, 16384, 16384), + (1, 8, 4, 128, 16384, 16384), + (1, 8, 2, 128, 16384, 16384), + (1, 8, 1, 128, 16384, 16384), + (1, 12, 12, 128, 7688, 256), + (1, 12, 6, 128, 7688, 256), + (1, 12, 4, 128, 7688, 256), + (1, 12, 3, 128, 7688, 256), + (1, 12, 2, 128, 7688, 256), + (1, 12, 1, 128, 7688, 256), + (6, 2, 2, 128, 12244, 123), + (6, 2, 1, 128, 12244, 123), + (2, 1, 1, 128, 2048, 2048), + (2, 1, 1, 64, 2048, 2048), + (4, 1, 1, 64, 2048, 2048), + (1, 1, 1, 64, 1411, 1375), + (2, 1, 1, 64, 1536, 1280), + (2, 1, 1, 64, 1536, 1536), + (2, 1, 1, 64, 1536, 1376), + (2, 1, 1, 64, 1416, 1376), + (2, 1, 1, 64, 1411, 1375), + (1, 1, 1, 64, 10240, 512), + (2, 1, 1, 64, 10240, 512), + (4, 1, 1, 64, 10240, 512), + (8, 1, 1, 64, 10240, 512), + (3, 1, 1, 64, 10240, 512), + (4, 1, 1, 64, 10240, 512), + (5, 1, 1, 64, 10240, 512), + (6, 1, 1, 64, 10240, 512), + (7, 1, 1, 64, 10240, 512), + (8, 1, 1, 64, 10240, 512), + (3, 3, 3, 64, 9197, 512), + (3, 3, 1, 64, 9197, 512), + (7, 1, 1, 64, 10240, 256), + (3, 3, 3, 64, 10240, 256), + (3, 3, 3, 64, 9216, 256), + (3, 3, 3, 64, 9200, 256), + (3, 3, 3, 64, 9200, 192), + (3, 3, 3, 64, 9200, 168), + (3, 3, 3, 64, 9200, 166), + (3, 3, 3, 64, 9198, 166), + (3, 3, 3, 64, 9197, 166), + (4, 1, 1, 64, 10240, 10240), + (4, 1, 1, 64, 10240, 1024), + (1, 1, 1, 64, 9197, 166), + (3, 3, 3, 64, 2560, 256), + (1, 1, 1, 128, 128, 128), + (2, 1, 1, 128, 128, 128), + (1, 2, 2, 128, 128, 128), + (2, 2, 2, 128, 128, 128), + (2, 2, 2, 64, 128, 128), + (1, 1, 1, 32, 32, 32), + (1, 1, 1, 32, 128, 128), + (1, 1, 1, 32, 128, 128), + (1, 1, 1, 128, 128, 64), + (1, 1, 1, 32, 128, 258), + (1, 2, 2, 64, 128, 15), + (1, 1, 1, 32, 8, 17), + (1, 1, 1, 64, 17, 49), + (2, 4, 4, 32, 128, 237), + (2, 4, 2, 32, 128, 237), + (2, 4, 1, 32, 128, 237), + (4, 3, 3, 64, 256, 33), + (4, 3, 1, 64, 256, 33), + (1, 1, 1, 128, 128, 75), + (1, 1, 1, 32, 125, 444), + (1, 2, 2, 64, 125, 231), + (1, 2, 1, 64, 125, 231), + (1, 1, 1, 128, 256, 10240), + (1, 1, 1, 32, 128, 4096), + (1, 1, 1, 128, 3584, 381), + (1, 1, 1, 128, 12072, 1680), + ] + for ( + batch, + heads, + heads_kv, + head_dim, + seqlen_q, + seqlen_kv, + ) in problem_sizes: + for is_causal in [False, True]: + self._test_backend_against_torch_sdpa( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + seqlen_q=seqlen_q, + seqlen_kv=seqlen_kv, + is_causal=is_causal, + backend="flash2", + ) + + @pytest.mark.L1 + @skip_if_flash2_not_supported() + @skip_if_not_supported() + def test_flash2_randsweep(self): + self._test_randsweep_against_torch_sdpa(backend="flash2", max_tests=RAND_SWEEP_TESTS) + + @pytest.mark.L1 + @skip_if_flash3_not_supported() + @skip_if_not_supported() + def test_flash3_fast(self): + problem_sizes = [ + (1, 8, 8, 128, 16384, 16384), + (1, 8, 4, 128, 16384, 16384), + (1, 8, 2, 128, 16384, 16384), + (1, 8, 1, 128, 16384, 16384), + (1, 12, 12, 128, 7688, 256), + (1, 12, 6, 128, 7688, 256), + (1, 12, 4, 128, 7688, 256), + (1, 12, 3, 128, 7688, 256), + (1, 12, 2, 128, 7688, 256), + (1, 12, 1, 128, 7688, 256), + (6, 2, 2, 128, 12244, 123), + (6, 2, 1, 128, 12244, 123), + (2, 1, 1, 128, 2048, 2048), + (2, 1, 1, 64, 2048, 2048), + (4, 1, 1, 64, 2048, 2048), + (1, 1, 1, 64, 1411, 1375), + (2, 1, 1, 64, 1536, 1280), + (2, 1, 1, 64, 1536, 1536), + (2, 1, 1, 64, 1536, 1376), + (2, 1, 1, 64, 1416, 1376), + (2, 1, 1, 64, 1411, 1375), + (1, 1, 1, 64, 10240, 512), + (2, 1, 1, 64, 10240, 512), + (4, 1, 1, 64, 10240, 512), + (8, 1, 1, 64, 10240, 512), + (3, 1, 1, 64, 10240, 512), + (4, 1, 1, 64, 10240, 512), + (5, 1, 1, 64, 10240, 512), + (6, 1, 1, 64, 10240, 512), + (7, 1, 1, 64, 10240, 512), + (8, 1, 1, 64, 10240, 512), + (3, 3, 3, 64, 9197, 512), + (3, 3, 1, 64, 9197, 512), + (7, 1, 1, 64, 10240, 256), + (3, 3, 3, 64, 10240, 256), + (3, 3, 3, 64, 9216, 256), + (3, 3, 3, 64, 9200, 256), + (3, 3, 3, 64, 9200, 192), + (3, 3, 3, 64, 9200, 168), + (3, 3, 3, 64, 9200, 166), + (3, 3, 3, 64, 9198, 166), + (3, 3, 3, 64, 9197, 166), + (4, 1, 1, 64, 10240, 10240), + (4, 1, 1, 64, 10240, 1024), + (1, 1, 1, 64, 9197, 166), + (3, 3, 3, 64, 2560, 256), + (1, 1, 1, 128, 128, 128), + (2, 1, 1, 128, 128, 128), + (1, 2, 2, 128, 128, 128), + (2, 2, 2, 128, 128, 128), + (2, 2, 2, 64, 128, 128), + (1, 1, 1, 32, 32, 32), + (1, 1, 1, 32, 128, 128), + (1, 1, 1, 32, 128, 128), + (1, 1, 1, 128, 128, 64), + (1, 1, 1, 32, 128, 258), + (1, 2, 2, 64, 128, 15), + (1, 1, 1, 32, 8, 17), + (1, 1, 1, 64, 17, 49), + (2, 4, 4, 32, 128, 237), + (2, 4, 2, 32, 128, 237), + (2, 4, 1, 32, 128, 237), + (4, 3, 3, 64, 256, 33), + (4, 3, 1, 64, 256, 33), + (1, 1, 1, 128, 128, 75), + (1, 1, 1, 32, 125, 444), + (1, 2, 2, 64, 125, 231), + (1, 2, 1, 64, 125, 231), + (1, 1, 1, 128, 256, 10240), + (1, 1, 1, 32, 128, 4096), + (1, 1, 1, 128, 3584, 381), + (1, 1, 1, 128, 12072, 1680), + ] + for ( + batch, + heads, + heads_kv, + head_dim, + seqlen_q, + seqlen_kv, + ) in problem_sizes: + for is_causal in [False, True]: + self._test_backend_against_torch_sdpa( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + seqlen_q=seqlen_q, + seqlen_kv=seqlen_kv, + is_causal=is_causal, + backend="flash3", + ) + + @pytest.mark.L1 + @skip_if_flash3_not_supported() + @skip_if_not_hopper() + def test_flash3_randsweep(self): + self._test_randsweep_against_torch_sdpa(backend="flash3", max_tests=RAND_SWEEP_TESTS) + + +if __name__ == "__main__": + random.seed(42) + torch.manual_seed(42) + unittest.main() diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/torch_compile_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/torch_compile_test.py new file mode 100644 index 0000000000000000000000000000000000000000..d00f2153f4eb4ea9a7af833797b8f66659a617e8 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/torch_compile_test.py @@ -0,0 +1,365 @@ +################################################################################################# +# Copyright (c) 2022-2025 Ali Hassani. +# +# Permission is hereby granted, free of charge, to any person obtaining a copy +# of this software and associated documentation files (the "Software"), to deal +# in the Software without restriction, including without limitation the rights +# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +# copies of the Software, and to permit persons to whom the Software is +# furnished to do so, subject to the following conditions: +# +# The above copyright notice and this permission notice shall be included in all +# copies or substantial portions of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +# SOFTWARE. +# +################################################################################################# + +import unittest +from functools import partial + +import pytest +import torch +from torch import nn + +from cosmos_policy._src.imaginaire.attention import attention as i4_attention +from cosmos_policy._src.imaginaire.attention.flash2 import FLASH2_SUPPORTED +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.natten import NATTEN_SUPPORTED +from cosmos_policy._src.imaginaire.attention.utils import is_blackwell_dc, is_hopper +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log +from cosmos_policy._src.imaginaire.attention.varlen import generate_varlen_parameters + +skip_if_natten_not_supported = partial( + pytest.mark.skipif, + not NATTEN_SUPPORTED, + reason="NATTEN is disabled, not available, or too old in this environment.", +) + +skip_if_flash2_not_supported = partial( + pytest.mark.skipif, + not FLASH2_SUPPORTED, + reason="Flash2 is disabled, not available, or too old in this environment.", +) + +# Tests are only enabled on Hopper and Blackwell DC-class for now. +# Will extend to other arches as we integrate more backends. +skip_if_not_supported = partial( + pytest.mark.skipif, + not is_blackwell_dc() and not is_hopper(), + reason="Attention tests are only allowed for Hopper and Blackwell DC-class GPUs for now.", +) + +skip_if_not_blackwell = partial( + pytest.mark.skipif, not is_blackwell_dc(), reason="This test is only allowed for Blackwell DC-class GPUs." +) + +skip_if_not_hopper = partial(pytest.mark.skipif, not is_hopper(), reason="This test is only allowed for Hopper GPUs.") + + +def reset_torch_compile(cache_size_limit): + # Torch compile reset and sensible settings for unit testing + log.debug(f"Resetting torch compile cache. New cache size limit: {cache_size_limit}") + torch.compiler.reset() + torch._dynamo.config.cache_size_limit = cache_size_limit + torch._dynamo.config.accumulated_recompile_limit = cache_size_limit * 4 + torch._dynamo.config.fail_on_recompile_limit_hit = True + + +def _reset_everything(): + torch.manual_seed(42) + reset_torch_compile(1024) + torch.cuda.empty_cache() + + +class Block(nn.Module): + def __init__( + self, + embed_dim: int, + num_heads: int, + mlp_ratio: int, + qkv_bias: bool = True, + ): + super().__init__() + + self.embed_dim = embed_dim + self.mlp_ratio = mlp_ratio + self.mlp_dim = int(self.embed_dim * self.mlp_ratio) + self.num_heads = num_heads + self.head_dim = self.embed_dim // self.num_heads + self.scale = self.head_dim**-0.5 + + self.q = nn.Linear(self.embed_dim, self.embed_dim, bias=qkv_bias) + self.kv = nn.Linear(self.embed_dim, self.embed_dim * 2, bias=qkv_bias) + self.proj = nn.Linear(self.embed_dim, self.embed_dim) + + self.mlp = nn.Sequential( + nn.Linear(self.embed_dim, self.mlp_dim), + nn.GELU(), + nn.Linear(self.mlp_dim, embed_dim), + ) + + def forward(self, x: torch.Tensor, c: torch.Tensor, *args, **kwargs): + B, sQ, D = x.shape + B, sK, D = c.shape + q = self.q(x).reshape(B, sQ, self.num_heads, self.head_dim) + k, v = self.kv(c).reshape(B, sK, 2, self.num_heads, self.head_dim).permute(2, 0, 1, 3, 4) + + x0 = i4_attention(q, k, v, *args, **kwargs) + assert isinstance(x0, torch.Tensor) + x0 = x0.reshape(B, sQ, D) + + return self.mlp(x0) + + +class TorchCompileTests(unittest.TestCase): + def setUp(self): + _reset_everything() + + def tearDown(self): + _reset_everything() + + def _test_module( + self, + batch: int, + seqlens_Q: list[int], + seqlens_KV: list[int], + num_heads: int, + head_dim: int, + is_causal: bool, + causal_type: CausalType, + atol: float, + backend: str, + device: str = "cuda", + dtype: torch.dtype = torch.float16, + ): + embed_dim = num_heads * head_dim + + assert len(seqlens_Q) == len(seqlens_KV) + assert len(seqlens_Q) >= 1 + assert len(seqlens_Q) == 1 or batch == len(seqlens_Q) + + seqlen_q = sum(seqlens_Q) + seqlen_kv = sum(seqlens_KV) + is_varlen = len(seqlens_Q) > 1 + + batch_ = 1 if is_varlen else batch + seqlens_Q_ = torch.tensor(seqlens_Q, device=device, dtype=torch.int32) if is_varlen else None + seqlens_KV_ = torch.tensor(seqlens_KV, device=device, dtype=torch.int32) if is_varlen else None + + dummy_q = torch.randn( + (batch_, seqlen_q, num_heads, head_dim), + dtype=dtype, + device=device, + requires_grad=True, + ) + dummy_kv = torch.randn( + (batch_, seqlen_kv, num_heads, head_dim), + dtype=dtype, + device=device, + requires_grad=True, + ) + + # seq maxes MUST be computed ahead of time when using torch compile + # because need to be copied to host + ( + cumulative_seqlen_Q, + cumulative_seqlen_KV, + max_seqlen_Q, + max_seqlen_KV, + ) = generate_varlen_parameters( + query=dummy_q, + key=dummy_kv, + value=dummy_kv, + seqlens_Q=seqlens_Q_, + seqlens_KV=seqlens_KV_, + ) + + _reset_everything() + + log.debug( + f"Testing torch compile on Attention module with input shapes: " + f"{batch=}, {num_heads=}, {head_dim=}, {seqlens_Q=}, {seqlens_KV=}, " + f"{is_causal=}, {causal_type=}, {dtype=}, {device=}, {backend=}." + ) + + model_eager = ( + Block( + embed_dim=embed_dim, + mlp_ratio=2, + num_heads=num_heads, + ) + .to(dtype) + .to(device) + ) + + model_compiled = torch.compile(model_eager, fullgraph=True, backend="inductor") + + x = torch.randn((batch_, seqlen_q, embed_dim), dtype=dtype, device=device) + c = torch.randn((batch_, seqlen_kv, embed_dim), dtype=dtype, device=device) + dy = torch.randn((batch_, seqlen_q, embed_dim), dtype=dtype, device=device) * 0.1 + + x_ref = x.clone().requires_grad_(True) + c_ref = c.clone().requires_grad_(True) + dy_ref = dy.clone() + + # eager + y_ref = model_eager( + x_ref, + c_ref, + is_causal=is_causal, + causal_type=causal_type, + cumulative_seqlen_Q=cumulative_seqlen_Q, + cumulative_seqlen_KV=cumulative_seqlen_KV, + max_seqlen_Q=max_seqlen_Q, + max_seqlen_KV=max_seqlen_KV, + backend=backend, + ) + y_ref.backward(dy_ref) + dx_ref = x_ref.grad + dc_ref = c_ref.grad + + # compile on first attempt + x = x.requires_grad_(True) + c = c.requires_grad_(True) + y = model_compiled( + x, + c, + is_causal=is_causal, + causal_type=causal_type, + cumulative_seqlen_Q=cumulative_seqlen_Q, + cumulative_seqlen_KV=cumulative_seqlen_KV, + max_seqlen_Q=max_seqlen_Q, + max_seqlen_KV=max_seqlen_KV, + backend=backend, + ) + y.backward(dy) + dx = x.grad + dc = c.grad + + torch.testing.assert_close(y, y_ref, atol=atol, rtol=0) + torch.testing.assert_close(dx, dx_ref, atol=atol, rtol=0) + torch.testing.assert_close(dc, dc_ref, atol=atol, rtol=0) + + # Second run, just to make sure it doesn't crash + y = model_compiled( + x, + c, + is_causal=is_causal, + causal_type=causal_type, + cumulative_seqlen_Q=cumulative_seqlen_Q, + cumulative_seqlen_KV=cumulative_seqlen_KV, + max_seqlen_Q=max_seqlen_Q, + max_seqlen_KV=max_seqlen_KV, + backend=backend, + ) + y.backward(dy) + dx = x.grad + dc = c.grad + + @pytest.mark.L1 + @skip_if_natten_not_supported() + @skip_if_not_supported() + def test_compiled_natten(self): + problem_sizes = [ + (1, 4, 128, [128], [128]), + (1, 1, 128, [128], [1024]), + (1, 1, 128, [128], [13568]), + (1, 1, 128, [128], [13496]), + (1, 1, 32, [128], [13496]), + (1, 1, 32, [32], [13496]), + (3, 1, 32, [77], [8504]), + (1, 1, 32, [77], [8504]), + (1, 1, 64, [40], [12296]), + (1, 2, 64, [40], [12296]), + (1, 2, 64, [40], [12296]), + (1, 1, 128, [128], [128]), + (6, 1, 128, [128, 128, 135, 121, 128, 128], [128, 128, 135, 121, 128, 128]), + (5, 1, 128, [128, 128, 135, 128, 128], [128, 128, 135, 128, 128]), + (2, 1, 128, [135, 200], [128, 768]), + (2, 1, 128, [1024, 200], [128, 768]), + (2, 1, 128, [135, 200], [135, 768]), + (2, 1, 128, [1024, 200], [135, 768]), + (2, 1, 128, [1024, 256], [128, 768]), + (4, 1, 128, [1024, 8, 17, 2048], [10, 20, 512, 16]), + (3, 2, 128, [268, 1584, 1571], [2448, 4088, 1925]), + (2, 1, 128, [1024, 256], [512, 768]), + ] + for ( + batch, + num_heads, + head_dim, + seqlens_Q, + seqlens_KV, + ) in problem_sizes: + for is_causal in [False, True]: + self._test_module( + batch=batch, + seqlens_Q=seqlens_Q, + seqlens_KV=seqlens_KV, + num_heads=num_heads, + head_dim=head_dim, + is_causal=is_causal, + causal_type=CausalType.TopLeft, + atol=1e-3, + backend="natten", + ) + + @pytest.mark.L1 + @skip_if_flash2_not_supported() + @skip_if_not_supported() + def test_compiled_flash2(self): + problem_sizes = [ + (1, 4, 128, [128], [128]), + (1, 1, 128, [128], [1024]), + (1, 1, 128, [128], [13568]), + (1, 1, 128, [128], [13496]), + (1, 1, 32, [128], [13496]), + (1, 1, 32, [32], [13496]), + (3, 1, 32, [77], [8504]), + (1, 1, 32, [77], [8504]), + (1, 1, 64, [40], [12296]), + (1, 2, 64, [40], [12296]), + (1, 2, 64, [40], [12296]), + (1, 1, 128, [128], [128]), + (6, 1, 128, [128, 128, 135, 121, 128, 128], [128, 128, 135, 121, 128, 128]), + (5, 1, 128, [128, 128, 135, 128, 128], [128, 128, 135, 128, 128]), + (2, 1, 128, [135, 200], [128, 768]), + (2, 1, 128, [1024, 200], [128, 768]), + (2, 1, 128, [135, 200], [135, 768]), + (2, 1, 128, [1024, 200], [135, 768]), + (2, 1, 128, [1024, 256], [128, 768]), + (4, 1, 128, [1024, 8, 17, 2048], [10, 20, 512, 16]), + (3, 2, 128, [268, 1584, 1571], [2448, 4088, 1925]), + (2, 1, 128, [1024, 256], [512, 768]), + ] + for ( + batch, + num_heads, + head_dim, + seqlens_Q, + seqlens_KV, + ) in problem_sizes: + for is_causal in [False, True]: + self._test_module( + batch=batch, + seqlens_Q=seqlens_Q, + seqlens_KV=seqlens_KV, + num_heads=num_heads, + head_dim=head_dim, + is_causal=is_causal, + causal_type=CausalType.BottomRight, + atol=1e-3, + backend="flash2", + ) + + +if __name__ == "__main__": + torch.manual_seed(42) + unittest.main() diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/varlen_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/varlen_test.py new file mode 100644 index 0000000000000000000000000000000000000000..d3622a3b5fec11f96f95759202955da70d9f38cf --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/tests/varlen_test.py @@ -0,0 +1,711 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +SDPA unit tests. +""" + +import random +import unittest +from functools import partial + +import pytest +import torch + +from cosmos_policy._src.imaginaire.attention import attention as i4_attention +from cosmos_policy._src.imaginaire.attention.flash2 import FLASH2_SUPPORTED +from cosmos_policy._src.imaginaire.attention.flash3 import FLASH3_SUPPORTED +from cosmos_policy._src.imaginaire.attention.masks import CausalType +from cosmos_policy._src.imaginaire.attention.natten import NATTEN_SUPPORTED +from cosmos_policy._src.imaginaire.attention.utils import is_blackwell_dc, is_fp8, is_hopper +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log + +RAND_SWEEP_TESTS = 1000 + +skip_if_natten_not_supported = partial( + pytest.mark.skipif, + not NATTEN_SUPPORTED, + reason="NATTEN is disabled, not available, or too old in this environment.", +) + +skip_if_flash2_not_supported = partial( + pytest.mark.skipif, + not FLASH2_SUPPORTED, + reason="Flash2 is disabled, not available, or too old in this environment.", +) + +skip_if_flash3_not_supported = partial( + pytest.mark.skipif, + not FLASH3_SUPPORTED, + reason="Flash3 is disabled, not available, or too old in this environment.", +) + +# Tests are only enabled on Hopper and Blackwell DC-class for now. +# Will extend to other arches as we integrate more backends. +skip_if_not_supported = partial( + pytest.mark.skipif, + not is_blackwell_dc() and not is_hopper(), + reason="SDPA tests are only allowed for Hopper and Blackwell DC-class GPUs for now.", +) + +skip_if_not_blackwell = partial( + pytest.mark.skipif, not is_blackwell_dc(), reason="This test is only allowed for Blackwell DC-class GPUs." +) + +skip_if_not_hopper = partial(pytest.mark.skipif, not is_hopper(), reason="This test is only allowed for Hopper GPUs.") + + +def _reset_everything(): + torch.manual_seed(42) + torch.cuda.empty_cache() + + +# Computes varlen by breaking up into individual attention calls +def compute_split_reference( + batch: int, + heads: int, + head_dim: int, + seqlens_Q_list: list[int], + seqlens_KV_list: list[int], + is_causal: bool, + causal_type: CausalType | None, + backend: str, + test_backward: bool, + dtype: torch.dtype = torch.float32, + heads_kv: int | None = None, + head_dim_v: int | None = None, + backend_kwargs: dict | None = None, +): + heads_kv = heads_kv or heads + head_dim_v = head_dim_v or head_dim + + assert len(seqlens_Q_list) == len(seqlens_KV_list) == batch + + seqlen_q_total = sum(seqlens_Q_list) + seqlen_kv_total = sum(seqlens_KV_list) + dtype_safe = torch.float16 + with torch.no_grad(): + q_ref, k_ref, v_ref, d_out_ref = ( + torch.randn((1, seqlen_q_total, heads, head_dim), device="cuda", dtype=dtype_safe).to(dtype), + torch.randn( + (1, seqlen_kv_total, heads_kv, head_dim), + device="cuda", + dtype=dtype_safe, + ).to(dtype), + torch.randn( + (1, seqlen_kv_total, heads_kv, head_dim_v), + device="cuda", + dtype=dtype_safe, + ).to(dtype), + torch.randn((1, seqlen_q_total, heads, head_dim_v), device="cuda", dtype=dtype_safe).to(dtype), + ) + q, k, v, d_out = ( + q_ref.clone(), + k_ref.clone(), + v_ref.clone(), + d_out_ref.clone(), + ) + + out_list = [] + lse_list = [] + d_q_list = [] + d_k_list = [] + d_v_list = [] + + q_start, kv_start = 0, 0 + for b in range(batch): + seqlen_q = seqlens_Q_list[b] + seqlen_kv = seqlens_KV_list[b] + + q_ = q_ref[:, q_start : q_start + seqlen_q, :, :].clone() + k_ = k_ref[:, kv_start : kv_start + seqlen_kv, :, :].clone() + v_ = v_ref[:, kv_start : kv_start + seqlen_kv, :, :].clone() + + if test_backward: + q_ = q_.requires_grad_(True) + k_ = k_.requires_grad_(True) + v_ = v_.requires_grad_(True) + d_out_ = d_out_ref[:, q_start : q_start + seqlen_q, :, :].clone().requires_grad_(True) + + out_, lse_ = i4_attention( + q_, + k_, + v_, + is_causal=is_causal, + causal_type=causal_type, + backend=backend, + backend_kwargs=backend_kwargs, + return_lse=True, + ) + + if test_backward: + out_.backward(d_out_) + + with torch.no_grad(): + out_list.append(out_.data.clone().float()) + lse_list.append(lse_.data.clone().float()) + if test_backward: + assert q_.grad is not None + assert k_.grad is not None + assert v_.grad is not None + d_q_list.append(q_.grad.clone().float()) + d_k_list.append(k_.grad.clone().float()) + d_v_list.append(v_.grad.clone().float()) + + q_start += seqlen_q + kv_start += seqlen_kv + + assert q_start == seqlen_q_total + assert kv_start == seqlen_kv_total + + out_ref = torch.cat(out_list, dim=1) + lse_ref = torch.cat(lse_list, dim=1) + assert out_ref.shape[:3] == q_ref.shape[:3] + dq_ref = None + dk_ref = None + dv_ref = None + if test_backward: + dq_ref = torch.cat(d_q_list, dim=1) + dk_ref = torch.cat(d_k_list, dim=1) + dv_ref = torch.cat(d_v_list, dim=1) + + assert dq_ref.shape == q_ref.shape + assert dk_ref.shape == k_ref.shape + assert dv_ref.shape == v_ref.shape + + return (q, k, v, d_out), (out_ref, lse_ref, dq_ref, dk_ref, dv_ref) + + +class VarlenTest(unittest.TestCase): + def setUp(self): + _reset_everything() + + def tearDown(self): + _reset_everything() + + def _test_against_manual_varlen( + self, + batch: int, + heads: int, + head_dim: int, + seqlens_Q_list: list[int], + seqlens_KV_list: list[int], + is_causal: bool, + causal_type: CausalType | None, + dtype: torch.dtype, + atol_fwd: tuple[float, float], + atol_bwd: tuple[float, float, float] | None, + backend: str, + reference_backend: str, + test_backward: bool, + heads_kv: int | None = None, + head_dim_v: int | None = None, + reference_backend_kwargs: dict | None = None, + backend_kwargs: dict | None = None, + ): + heads_kv = heads_kv or heads + head_dim_v = head_dim_v or head_dim + + log.debug( + f"Testing varlen ({backend}) against manual varlen ({reference_backend}): " + f"{batch=}, {heads=}, {heads_kv=}, {head_dim=}, {head_dim_v=}, " + f"{seqlens_Q_list=}, {seqlens_KV_list=}, {is_causal=}, {causal_type=}, {dtype=}." + ) + + inputs, reference = compute_split_reference( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + causal_type=causal_type, + dtype=dtype, + backend=reference_backend, + backend_kwargs=reference_backend_kwargs, + test_backward=test_backward, + ) + + q, k, v, d_out = inputs + out_ref, lse_ref, dq_ref, dk_ref, dv_ref = reference + q = q.to(dtype) + k = k.to(dtype) + v = v.to(dtype) + d_out = d_out.to(dtype) + + # Run target + if test_backward: + q.requires_grad_(test_backward) + k.requires_grad_(test_backward) + v.requires_grad_(test_backward) + d_out.requires_grad_(test_backward) + + seqlens_Q = torch.tensor(seqlens_Q_list, dtype=torch.int32, device=q.device) + seqlens_KV = torch.tensor(seqlens_KV_list, dtype=torch.int32, device=q.device) + + out_, lse_ = i4_attention( + q, + k, + v, + is_causal=is_causal, + causal_type=causal_type, + backend=backend, + return_lse=True, + seqlens_Q=seqlens_Q, + seqlens_KV=seqlens_KV, + backend_kwargs=backend_kwargs, + ) + out = out_.float() + lse = lse_.float() + + if test_backward: + dq, dk, dv = None, None, None + out_.backward(d_out) + with torch.no_grad(): + dq, dk, dv = ( + q.grad.clone().float(), + k.grad.clone().float(), + v.grad.clone().float(), + ) + + atol_out, atol_lse = atol_fwd + assert out.shape == out_ref.shape + + torch.testing.assert_close(out, out_ref, atol=atol_out, rtol=0) + torch.testing.assert_close(lse, lse_ref, atol=atol_lse, rtol=0) + + if test_backward: + assert atol_bwd is not None + atol_dq, atol_dk, atol_dv = atol_bwd + torch.testing.assert_close(dq, dq_ref, atol=atol_dq, rtol=0) + torch.testing.assert_close(dk, dk_ref, atol=atol_dk, rtol=0) + torch.testing.assert_close(dv, dv_ref, atol=atol_dv, rtol=0) + + def _test_natten_varlen( + self, + batch: int, + heads: int, + head_dim: int, + seqlens_Q_list: list[int], + seqlens_KV_list: list[int], + is_causal: bool, + head_dim_v: int | None = None, + heads_kv: int | None = None, + ): + torch.set_default_device("cuda") + + # We're testing against the same backend and same dtype, + # but with varlen implemented as multiple kernel calls, so + # error thresholds should be much smaller here. + # This is therefore only a test of the varlen functionality. + # Correctness per dtype is expected to be verified in the main + # fmha tests. + # dQ still needs a more relaxed threshold because of the non-determinism + ALLOWED_DTYPES = [ + # dtype, (atol_out, atol_lse), (atol_dq, atol_dk, atol_dv) + (torch.float16, (1e-6, 1e-6), (1e-2, 1e-6, 1e-6)), + (torch.bfloat16, (1e-6, 1e-6), (1e-2, 1e-6, 1e-6)), + ] + + if is_blackwell_dc(): + ALLOWED_DTYPES += [ + (torch.float8_e4m3fn, (1e-6, 1e-6), None), + (torch.float8_e5m2, (1e-6, 1e-6), None), + ] + + # NOTE: Hopper FMHA does not support varlen, so natten falls back + # to cutlass-fmha, which means the reference may target hopper-fmha, + # while the varlen target is cutlass-fmha, and this will throw off the + # error limits. + backend_kwargs = None + if is_hopper(): + backend_kwargs = {"backend": "cutlass-fmha"} + + for dtype, atol_fwd, atol_bwd in ALLOWED_DTYPES: + self._test_against_manual_varlen( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + causal_type=CausalType.TopLeft, # Top-left is the only supported mask in natten (for now) + dtype=dtype, + atol_fwd=atol_fwd, + atol_bwd=atol_bwd, + backend="natten", + reference_backend="natten", + backend_kwargs=backend_kwargs, + reference_backend_kwargs=backend_kwargs, + test_backward=not is_fp8(dtype), + ) + + def _test_flash2_varlen( + self, + batch: int, + heads: int, + head_dim: int, + seqlens_Q_list: list[int], + seqlens_KV_list: list[int], + is_causal: bool, + head_dim_v: int | None = None, + heads_kv: int | None = None, + ): + torch.set_default_device("cuda") + + # we can't quite pull the same trick as in natten -- apparently the kernel + # configs for varlen and non-varlen cases are very different. + # Setting deterministic=True doesn't seem to help either + backend_kwargs = None + # backend_kwargs = {"deterministic": True} + ALLOWED_DTYPES = [ + # dtype, (atol_out, atol_lse), (atol_dq, atol_dk, atol_dv) + (torch.float16, (1e-2, 1e-2), (1e-1, 1e-2, 1e-2)), + (torch.bfloat16, (1e-1, 1e-2), (1e-1, 1e-1, 1e-1)), + ] + + for dtype, atol_fwd, atol_bwd in ALLOWED_DTYPES: + self._test_against_manual_varlen( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + causal_type=CausalType.BottomRight, # Bottom-right is the only supported mask in flash2 + dtype=dtype, + atol_fwd=atol_fwd, + atol_bwd=atol_bwd, + backend="flash2", + reference_backend="flash2", + backend_kwargs=backend_kwargs, + reference_backend_kwargs=backend_kwargs, + test_backward=True, + ) + + def _test_flash3_varlen( + self, + batch: int, + heads: int, + head_dim: int, + seqlens_Q_list: list[int], + seqlens_KV_list: list[int], + is_causal: bool, + head_dim_v: int | None = None, + heads_kv: int | None = None, + ): + torch.set_default_device("cuda") + + # We're testing against the same backend and same dtype, + # but with varlen implemented as multiple kernel calls, so + # error thresholds should be much smaller here. + # This is therefore only a test of the varlen functionality. + # Correctness per dtype is expected to be verified in the main + # fmha tests. + # dQ still needs a more relaxed threshold because of the non-determinism + ALLOWED_DTYPES = [ + # dtype, (atol_out, atol_lse), (atol_dq, atol_dk, atol_dv) + (torch.float16, (1e-6, 1e-6), (1e-2, 1e-6, 1e-6)), + (torch.bfloat16, (1e-6, 1e-6), (1e-2, 1e-6, 1e-6)), + ] + backend_kwargs = None + + # GQA/MQA introduce some extra non determinism (possibly due to extra reduction step?) + if heads_kv is not None and heads != heads_kv: + ALLOWED_DTYPES = [ + # dtype, (atol_out, atol_lse), (atol_dq, atol_dk, atol_dv) + (torch.float16, (1e-6, 1e-6), (1e-2, 1e-1, 1e-1)), + (torch.bfloat16, (1e-6, 1e-6), (1e-2, 1e-1, 1e-1)), + ] + backend_kwargs = {"deterministic": True} + + for dtype, atol_fwd, atol_bwd in ALLOWED_DTYPES: + self._test_against_manual_varlen( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + causal_type=CausalType.BottomRight, # Bottom-right is the only supported mask in flash3 + dtype=dtype, + atol_fwd=atol_fwd, + atol_bwd=atol_bwd, + backend="flash3", + reference_backend="flash3", + backend_kwargs=backend_kwargs, + reference_backend_kwargs=backend_kwargs, + test_backward=True, + ) + + def _test_varlen( + self, + batch: int, + heads: int, + head_dim: int, + seqlens_Q_list: list[int], + seqlens_KV_list: list[int], + is_causal: bool, + backend: str, + head_dim_v: int | None = None, + heads_kv: int | None = None, + ): + if backend == "natten": + self._test_natten_varlen( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + ) + elif backend == "flash2": + self._test_flash2_varlen( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + ) + elif backend == "flash3": + self._test_flash3_varlen( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + ) + else: + raise NotImplementedError() + + def _test_varlen_randsweep(self, backend: str, max_tests: int = 1000): + random.seed(42) + + max_seqlen = 2**17 + for i in range(max_tests): + batch = random.choice(range(1, 12)) + + supports_gqa_mqa = False + if backend == "natten": + head_dim_choices = [32, 64, 128] + heads_choices = range(1, 8 + 1) + # GQA/MQA is only supported in NATTEN's Blackwell FMHA backend for now + supports_gqa_mqa = is_blackwell_dc() + elif backend in ["flash2", "flash3"]: + head_dim_choices = range(16, 256 + 1, 8) + heads_choices = range(1, 8 + 1) + supports_gqa_mqa = True + else: + raise NotImplementedError() + + heads = random.choice(heads_choices) + heads_kv = ( + heads + if not supports_gqa_mqa + else random.choice([1] + [i for i in range(1, heads + 1) if heads % i == 0]) + ) + assert heads >= heads_kv and heads % heads_kv == 0 + + head_dim = random.choice(head_dim_choices) + head_dim_v = None + + seqlens_Q_list = [] + seqlens_KV_list = [] + for i in range(batch): + max_q = min(2**12, max(max_seqlen - sum(seqlens_Q_list), 24)) + max_k = min(2**12, max(max_seqlen - sum(seqlens_KV_list), 24)) + new_q = random.choice(range(8, max_q, 1)) + new_k = random.choice(range(8, max_k, 1)) + seqlens_Q_list.append(new_q) + seqlens_KV_list.append(new_k) + + for is_causal in [False, True]: + self._test_varlen( + batch=batch, + heads=heads, + heads_kv=heads_kv, + head_dim=head_dim, + head_dim_v=head_dim_v, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + backend=backend, + ) + + @pytest.mark.L1 + @skip_if_natten_not_supported() + @skip_if_not_supported() + def test_natten_varlen_fast(self): + problem_sizes = [ + ( + 9, + 4, + 128, + [2669, 2240, 910, 2421, 3323, 34, 3308, 2867, 1401], + [2880, 1726, 1847, 1147, 3568, 3116, 661, 1739, 1146], + ), + (6, 1, 128, [128, 128, 135, 121, 128, 128], [128, 128, 135, 121, 128, 128]), + (5, 1, 128, [128, 128, 135, 128, 128], [128, 128, 135, 128, 128]), + (2, 1, 128, [135, 200], [128, 768]), + (2, 1, 128, [1024, 200], [128, 768]), + (2, 1, 128, [135, 200], [135, 768]), + (2, 1, 128, [1024, 200], [135, 768]), + (2, 1, 128, [1024, 256], [128, 768]), + (4, 1, 128, [1024, 8, 17, 2048], [10, 20, 512, 16]), + (3, 2, 128, [268, 1584, 1571], [2448, 4088, 1925]), + (2, 1, 128, [1024, 256], [512, 768]), + ] + for ( + batch, + heads, + head_dim, + seqlens_Q_list, + seqlens_KV_list, + ) in problem_sizes: + for is_causal in [False, True]: + self._test_varlen( + batch=batch, + heads=heads, + head_dim=head_dim, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + backend="natten", + ) + + @pytest.mark.L1 + @skip_if_natten_not_supported() + @skip_if_not_supported() + def test_natten_varlen_randsweep(self): + self._test_varlen_randsweep(backend="natten", max_tests=RAND_SWEEP_TESTS) + + @pytest.mark.L1 + @skip_if_flash2_not_supported() + @skip_if_not_supported() + def test_flash2_varlen_fast(self): + problem_sizes = [ + ( + 9, + 4, + 128, + [2669, 2240, 910, 2421, 3323, 34, 3308, 2867, 1401], + [2880, 1726, 1847, 1147, 3568, 3116, 661, 1739, 1146], + ), + (6, 1, 128, [128, 128, 135, 121, 128, 128], [128, 128, 135, 121, 128, 128]), + (5, 1, 128, [128, 128, 135, 128, 128], [128, 128, 135, 128, 128]), + (2, 1, 128, [135, 200], [128, 768]), + (2, 1, 128, [1024, 200], [128, 768]), + (2, 1, 128, [135, 200], [135, 768]), + (2, 1, 128, [1024, 200], [135, 768]), + (2, 1, 128, [1024, 256], [128, 768]), + (4, 1, 128, [1024, 8, 17, 2048], [10, 20, 512, 16]), + (3, 2, 128, [268, 1584, 1571], [2448, 4088, 1925]), + (2, 1, 128, [1024, 256], [512, 768]), + ] + for ( + batch, + heads, + head_dim, + seqlens_Q_list, + seqlens_KV_list, + ) in problem_sizes: + for is_causal in [False, True]: + self._test_varlen( + batch=batch, + heads=heads, + head_dim=head_dim, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + backend="flash2", + ) + + @pytest.mark.L1 + @skip_if_flash2_not_supported() + @skip_if_not_supported() + def test_flash2_varlen_randsweep(self): + self._test_varlen_randsweep(backend="flash2", max_tests=RAND_SWEEP_TESTS) + + @pytest.mark.L1 + @skip_if_flash3_not_supported() + @skip_if_not_hopper() + def test_flash3_varlen_fast(self): + problem_sizes = [ + ( + 9, + 4, + 128, + [2669, 2240, 910, 2421, 3323, 34, 3308, 2867, 1401], + [2880, 1726, 1847, 1147, 3568, 3116, 661, 1739, 1146], + ), + (6, 1, 128, [128, 128, 135, 121, 128, 128], [128, 128, 135, 121, 128, 128]), + (5, 1, 128, [128, 128, 135, 128, 128], [128, 128, 135, 128, 128]), + (2, 1, 128, [135, 200], [128, 768]), + (2, 1, 128, [1024, 200], [128, 768]), + (2, 1, 128, [135, 200], [135, 768]), + (2, 1, 128, [1024, 200], [135, 768]), + (2, 1, 128, [1024, 256], [128, 768]), + (4, 1, 128, [1024, 8, 17, 2048], [10, 20, 512, 16]), + (3, 2, 128, [268, 1584, 1571], [2448, 4088, 1925]), + (2, 1, 128, [1024, 256], [512, 768]), + ] + for ( + batch, + heads, + head_dim, + seqlens_Q_list, + seqlens_KV_list, + ) in problem_sizes: + for is_causal in [False, True]: + self._test_varlen( + batch=batch, + heads=heads, + head_dim=head_dim, + seqlens_Q_list=seqlens_Q_list, + seqlens_KV_list=seqlens_KV_list, + is_causal=is_causal, + backend="flash3", + ) + + @pytest.mark.L1 + @skip_if_flash3_not_supported() + @skip_if_not_hopper() + def test_flash3_varlen_randsweep(self): + self._test_varlen_randsweep(backend="flash3", max_tests=RAND_SWEEP_TESTS) + + +if __name__ == "__main__": + random.seed(42) + torch.manual_seed(42) + unittest.main() diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/utils/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/utils/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..a1cbab79143d0833bf3a6de03c2484a2c428385a --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/utils/__init__.py @@ -0,0 +1,83 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Utilities: compute capability detection, helpers, and more. +""" + +from typing import Any + +import torch + +from cosmos_policy._src.imaginaire.attention.utils import safe_log as log +from cosmos_policy._src.imaginaire.attention.utils.environment import is_torch_compiling + + +def get_arch_tag(device: torch.device | None = None) -> int: + """ + Returns the compute capability of a given torch device if it's a CUDA device, otherwise returns 0. + + Args: + device (torch.device | None): torch device. Uses default device if None. + + Returns: + device_cc (int): compute capability in the SmXXX format (i.e. 90 for Hopper). + """ + if torch.cuda.is_available() and torch.version.cuda and (device is None or device.type == "cuda"): + major, minor = torch.cuda.get_device_capability(device) + return major * 10 + minor + return 0 + + +def log_or_raise_error(msg: str, raise_error: bool = False, exception: Any = RuntimeError): + if raise_error: + raise exception(msg) + else: + log.debug(msg) + + +def is_full(dtype: torch.dtype) -> bool: + return dtype == torch.float32 + + +def is_half(dtype: torch.dtype) -> bool: + return dtype in [torch.float16, torch.bfloat16] + + +def is_fp8(dtype: torch.dtype) -> bool: + return dtype in [torch.float8_e5m2, torch.float8_e4m3fn] + + +def is_hopper(device: torch.device | None = None) -> bool: + return get_arch_tag(device) == 90 + + +def is_blackwell_dc(device: torch.device | None = None) -> bool: + return get_arch_tag(device) in [100, 103] + + +__all__ = [ + "get_arch_tag", + "log_or_raise_error", + "is_full", + "is_half", + "is_fp8", + "is_hopper", + "is_blackwell_dc", + "is_torch_compiling", +] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/utils/environment.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/utils/environment.py new file mode 100644 index 0000000000000000000000000000000000000000..f2990ca71d681dc9a14b45e8dcf328b3953e6669 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/utils/environment.py @@ -0,0 +1,36 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Environment-related utilities. +""" + +import torch + +from cosmos_policy._src.imaginaire.utils import log + + +# Controls all regions guarded against torch compile +# Logs, and certain assertions cause graph breaks. +def is_torch_compiling() -> bool: + try: + return torch.compiler.is_compiling() + except Exception as e: + log.exception(f"Exception occurred checking whether in torch compiled region: {e}") + # Assume too old to support torch compile + return False diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/utils/safe_log.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/utils/safe_log.py new file mode 100644 index 0000000000000000000000000000000000000000..74d80d0ad5ee711d44b36ad1d991a0a47374b9ef --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/utils/safe_log.py @@ -0,0 +1,65 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Safe logging utilities: logging should be disabled when in a torch.compiled +region. +""" + +from cosmos_policy._src.imaginaire.attention.utils.environment import is_torch_compiling +from cosmos_policy._src.imaginaire.utils import log + + +def trace(message: str, rank0_only: bool = True) -> None: + if not is_torch_compiling(): + log.trace(message=message, rank0_only=rank0_only) + + +def debug(message: str, rank0_only: bool = True) -> None: + if not is_torch_compiling(): + log.debug(message=message, rank0_only=rank0_only) + + +def info(message: str, rank0_only: bool = True) -> None: + if not is_torch_compiling(): + log.info(message=message, rank0_only=rank0_only) + + +def success(message: str, rank0_only: bool = True) -> None: + if not is_torch_compiling(): + log.success(message=message, rank0_only=rank0_only) + + +def warning(message: str, rank0_only: bool = True) -> None: + if not is_torch_compiling(): + log.warning(message=message, rank0_only=rank0_only) + + +def error(message: str, rank0_only: bool = True) -> None: + if not is_torch_compiling(): + log.critical(message=message, rank0_only=rank0_only) + + +def critical(message: str, rank0_only: bool = True) -> None: + if not is_torch_compiling(): + log.critical(message=message, rank0_only=rank0_only) + + +def exception(message: str, rank0_only: bool = True) -> None: + if not is_torch_compiling(): + log.exception(message=message, rank0_only=rank0_only) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/attention/varlen.py b/REGEN-main/cosmos_policy/_src/imaginaire/attention/varlen.py new file mode 100644 index 0000000000000000000000000000000000000000..a0e9160e8eb376cec9f50d831641f11a89efef9d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/attention/varlen.py @@ -0,0 +1,120 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Imaginaire4 Attention Subpackage: +Unified implementation for all Attention implementations. + +Varlen utilities +""" + +import torch +from torch import Tensor + +from cosmos_policy._src.imaginaire.attention.utils import is_torch_compiling + + +def generate_varlen_parameters( + query: Tensor, + key: Tensor, + value: Tensor, + seqlens_Q: Tensor | None = None, + seqlens_KV: Tensor | None = None, +) -> tuple[None, None, int, int] | tuple[Tensor, Tensor, int, int]: + # NOTE: max_seqlen_{Q,KV} require a device-host sync, since they're expected to be ints (with + # which we launch the varlen kernel) and not device tensors. + # .item() introduces control flow and breaks the graph. + # It is also inefficient to repeat this per-op, and mostly there for convenience. + # generate_varlen_parameters should ideally always be called by the user ahead of model + # forward / backward. + if is_torch_compiling(): + raise RuntimeError( + "Running 'generate_varlen_parameters' in a torch-compiled region is disallowed as it " + "results in graph breaks. Please consider calling ahead of time and pass " + "'cumulative_seqlen_{Q,KV}' and 'max_seqlen_{Q,KV}' instead of 'seqlens_{Q,KV}' to " + "'attention'. " + ) + + if query.shape[0] != key.shape[0] or query.shape[0] != value.shape[0]: + raise ValueError( + f"Q, K, and V must match in batch size, got {query.shape[0]=}, {key.shape[0]=}, {value.shape[0]=}." + ) + + if (seqlens_Q is None) ^ (seqlens_KV is None): + raise ValueError( + "Variable length Attention requires both of seqlens_Q and seqlens_KV to be set, got " + f"{seqlens_Q=}, {seqlens_KV=}." + ) + + if seqlens_Q is None and seqlens_KV is None: + # Not varlen + return None, None, 0, 0 + + assert seqlens_Q is not None + assert seqlens_KV is not None + + if not isinstance(seqlens_Q, Tensor) or not isinstance(seqlens_KV, Tensor): + raise ValueError("seqlens_Q and seqlens_KV must both be tensors.") + + if seqlens_Q.device != query.device or seqlens_KV.device != query.device: + raise ValueError( + "seqlens_Q and seqlens_KV must be on the same device as QKV, but " + f"{seqlens_Q.device=}, {seqlens_KV.device=}, {query.device=}." + ) + + if seqlens_Q.dtype != torch.int32 or seqlens_KV.dtype != torch.int32: + raise ValueError( + f"seqlens_Q and seqlens_KV must both be torch.int32 tensors, got {seqlens_Q.dtype=}, {seqlens_KV.dtype=}." + ) + + if seqlens_Q.dim() != 1 or seqlens_KV.dim() != 1: + raise ValueError( + f"seqlens_Q and seqlens_KV must both be 1-D tensors, got {seqlens_Q.dim()=}, {seqlens_KV.dim()=}." + ) + + if seqlens_Q.shape[0] != seqlens_KV.shape[0]: + raise ValueError(f"seqlens_Q and seqlens_KV must match in size, got {seqlens_Q.shape=}, {seqlens_KV.shape=}.") + + if seqlens_Q.shape[0] < 1: + raise ValueError( + f"seqlens_Q and seqlens_KV must contain at least one element, got {seqlens_Q.shape=}, {seqlens_KV.shape=}." + ) + + if query.shape[0] != 1: + raise ValueError( + f"Variable length attention only supports sequence-packed memory layout (batch = 1), got {query.shape[0]=}." + ) + + assert seqlens_Q.dim() == seqlens_KV.dim() == 1 + assert seqlens_Q.shape[0] == seqlens_KV.shape[0] >= 1 + assert seqlens_Q.dtype == seqlens_KV.dtype == torch.int32 + + max_seqlen_Q = seqlens_Q.max().item() # type: ignore + max_seqlen_KV = seqlens_KV.max().item() # type: ignore + + # NOTE: we have to prepend with 0 manually :( + z = torch.tensor([0], dtype=torch.int32, device=seqlens_Q.device) + cumulative_seqlen_Q = torch.cat([z, seqlens_Q.cumsum(0).to(torch.int32)], dim=0) + cumulative_seqlen_KV = torch.cat([z, seqlens_KV.cumsum(0).to(torch.int32)], dim=0) + + assert isinstance(max_seqlen_Q, int) + assert isinstance(max_seqlen_KV, int) + + return ( + cumulative_seqlen_Q, + cumulative_seqlen_KV, + max_seqlen_Q, + max_seqlen_KV, + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/blocklist.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/blocklist.py new file mode 100644 index 0000000000000000000000000000000000000000..ab39eb5bbd5a2c73394eab049601d455b5cb5e1d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/blocklist.py @@ -0,0 +1,248 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import argparse +import os +import re +import string +from difflib import SequenceMatcher + +import nltk +from better_profanity import profanity + +from cosmos_policy._src.imaginaire.auxiliary.guardrail.blocklist.utils import read_keyword_list_from_dir, to_ascii +from cosmos_policy._src.imaginaire.auxiliary.guardrail.common.core import ( + GUARDRAIL1_CHECKPOINT_DIR, + ContentSafetyGuardrail, + GuardrailRunner, +) +from cosmos_policy._src.imaginaire.utils import log, misc + +CENSOR = misc.Color.red("*") + + +class Blocklist(ContentSafetyGuardrail): + def __init__( + self, + guardrail_partial_match_min_chars: int = 6, + guardrail_partial_match_letter_count: float = 0.4, + ) -> None: + """Blocklist model for text filtering safety check. + + Args: + checkpoint_dir (str): Path to the checkpoint directory. + guardrail_partial_match_min_chars (int, optional): Minimum number of characters in a word to check for partial match. Defaults to 6. + guardrail_partial_match_letter_count (float, optional): Maximum allowed difference in characters for partial match. Defaults to 0.4. + """ + self.checkpoint_dir = os.path.join(GUARDRAIL1_CHECKPOINT_DIR, "blocklist") + nltk.data.path.append(os.path.join(self.checkpoint_dir, "nltk_data")) + self.lemmatizer = nltk.WordNetLemmatizer() + self.profanity = profanity + self.guardrail_partial_match_min_chars = guardrail_partial_match_min_chars + self.guardrail_partial_match_letter_count = guardrail_partial_match_letter_count + + # Load blocklist and whitelist keywords + self.blocklist_words = read_keyword_list_from_dir(os.path.join(self.checkpoint_dir, "custom")) + self.whitelist_words = read_keyword_list_from_dir(os.path.join(self.checkpoint_dir, "whitelist")) + self.exact_match_words = read_keyword_list_from_dir(os.path.join(self.checkpoint_dir, "exact_match")) + + self.profanity.load_censor_words(custom_words=self.blocklist_words, whitelist_words=self.whitelist_words) + log.debug(f"Loaded {len(self.blocklist_words)} words/phrases from blocklist") + log.debug(f"Whitelisted {len(self.whitelist_words)} words/phrases from whitelist") + log.debug(f"Loaded {len(self.exact_match_words)} exact match words/phrases from blocklist") + + def uncensor_whitelist(self, input_prompt: str, censored_prompt: str) -> str: + """Explicitly uncensor words that are in the whitelist.""" + input_words = input_prompt.split() + censored_words = censored_prompt.split() + whitelist_words = set(self.whitelist_words) + for i, token in enumerate(input_words): + if token.strip(string.punctuation).lower() in whitelist_words: + censored_words[i] = token + censored_prompt = " ".join(censored_words) + return censored_prompt + + def censor_prompt(self, input_prompt: str) -> tuple[bool, str]: + """Censor the prompt using the blocklist with better-profanity fuzzy matching. + + Args: + input_prompt: input prompt to censor + + Returns: + bool: True if the prompt is blocked, False otherwise + str: A message indicating why the prompt was blocked + """ + censored_prompt = self.profanity.censor(input_prompt, censor_char=CENSOR) + # Uncensor whitelisted words that were censored from blocklist fuzzy matching + censored_prompt = self.uncensor_whitelist(input_prompt, censored_prompt) + if CENSOR in censored_prompt: + return True, f"Prompt blocked by censorship: Censored Prompt: {censored_prompt}" + return False, "" + + @staticmethod + def check_partial_match( + normalized_prompt: str, normalized_word: str, guardrail_partial_match_letter_count: float + ) -> tuple[bool, str]: + """ + Check robustly if normalized word and the matching target have a difference of up to guardrail_partial_match_letter_count characters. + + Args: + normalized_prompt: a string with many words + normalized_word: a string with one or multiple words, its length is smaller than normalized_prompt + guardrail_partial_match_letter_count: maximum allowed difference in characters (float to allow partial characters) + + Returns: + bool: True if a match is found, False otherwise + str: A message indicating why the prompt was blocked + """ + prompt_words = normalized_prompt.split() + word_length = len(normalized_word.split()) + max_similarity_ratio = (len(normalized_word) - float(guardrail_partial_match_letter_count)) / float( + len(normalized_word) + ) + + seq_matcher = SequenceMatcher(None) + seq_matcher.set_seq2(normalized_word) + + for i in range(len(prompt_words) - word_length + 1): + # Extract a substring from the prompt with the same number of words as the normalized_word + substring = " ".join(prompt_words[i : i + word_length]) + seq_matcher.set_seq1(substring) + + # real_quick_ratio and quick_ratio are faster than ratio and both serve as upper bound for similarity ratio. + # If they are less than max_similarity_ratio, it means that also the ratio will be less than max_similarity_ratio and we can skip the expensive ratio computation. + # This saves a lot of time because in practice the tested words are usually dissimilar. + # For details see: https://docs.python.org/3/library/difflib.html#difflib.SequenceMatcher + if ( + seq_matcher.real_quick_ratio() < max_similarity_ratio + or seq_matcher.quick_ratio() < max_similarity_ratio + ): + continue + + similarity_ratio = seq_matcher.ratio() + if similarity_ratio >= max_similarity_ratio: + return ( + True, + f"Prompt blocked by partial match blocklist: Prompt: {normalized_prompt}, Partial Match Word: {normalized_word}", + ) + + return False, "" + + @staticmethod + def check_against_whole_word_blocklist( + prompt: str, + blocklist: list[str], + guardrail_partial_match_min_chars: int = 6, + guardrail_partial_match_letter_count: float = 0.4, + ) -> tuple[bool, str]: + """ + Check if the prompt contains any whole words from the blocklist. + The match is case insensitive and robust to multiple spaces between words. + + Args: + prompt: input prompt to check + blocklist: list of words to check against + guardrail_partial_match_min_chars: minimum number of characters in a word to check for partial match + guardrail_partial_match_letter_count: maximum allowed difference in characters for partial match + + Returns: + tuple[bool, str]: (True if a match is found, False otherwise), message indicating why the prompt was blocked + """ + # Normalize spaces and convert to lowercase + normalized_prompt = re.sub(r"\s+", " ", prompt).strip().lower() + + normalized_words_cache = set() + + for word in blocklist: + # Normalize spaces and convert to lowercase for each blocklist word + normalized_word = re.sub(r"\s+", " ", word).strip().lower() + + if normalized_word in normalized_words_cache: + continue + + normalized_words_cache.add(normalized_word) + + # Use word boundaries to ensure whole word match + if re.search(r"\b" + re.escape(normalized_word) + r"\b", normalized_prompt): + return True, f"Prompt blocked by exact match blocklist: Prompt: {prompt}, Exact Match Word: {word}" + + # Roughly 3/4 of the time this function requires is spent on partial matching. + # We could use just one for loop to check both exact and partial matches but doing it in two loops is faster in practice + # because it delays the partial matching as long as possible with a chance of early exit due to exact match. + # Above we cache the normalized words and here we reuse them in the second loop for partial matching. + + for normalized_word in normalized_words_cache: + # Check for partial match if the word is long enough + if len(normalized_word) >= guardrail_partial_match_min_chars: + match, message = Blocklist.check_partial_match( + normalized_prompt, normalized_word, guardrail_partial_match_letter_count + ) + if match: + return True, message + + return False, "" + + def is_safe(self, input_prompt: str = "") -> tuple[bool, str]: + """Check if the input prompt is safe using the blocklist.""" + # Check if the input is empty + if not input_prompt: + return False, "Input is empty" + input_prompt = to_ascii(input_prompt) + + # Check full sentence for censored words + censored, message = self.censor_prompt(input_prompt) + if censored: + return False, message + + # Check lemmatized words for censored words + tokens = nltk.word_tokenize(input_prompt) + lemmas = [self.lemmatizer.lemmatize(token) for token in tokens] + lemmatized_prompt = " ".join(lemmas) + censored, message = self.censor_prompt(lemmatized_prompt) + if censored: + return False, message + + # Check for exact match blocklist words + censored, message = self.check_against_whole_word_blocklist( + input_prompt, + self.exact_match_words, + self.guardrail_partial_match_min_chars, + self.guardrail_partial_match_letter_count, + ) + if censored: + return False, message + + # If all these checks pass, the input is safe + return True, "Input is safe" + + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--prompt", type=str, required=True, help="Input prompt") + return parser.parse_args() + + +def main(args): + blocklist = Blocklist() + runner = GuardrailRunner(safety_models=[blocklist]) + with misc.timer("blocklist safety check"): + safety, message = runner.run_safety_check(args.prompt) + log.info(f"Input is: {'SAFE' if safety else 'UNSAFE'}") + log.info(f"Message: {message}") if not safety else None + + +if __name__ == "__main__": + args = parse_args() + main(args) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/blocklist_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/blocklist_test.py new file mode 100644 index 0000000000000000000000000000000000000000..0b5fc3396935431c2758763b1cd439d73188e198 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/blocklist_test.py @@ -0,0 +1,57 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import pytest + +from cosmos_policy._src.imaginaire.auxiliary.guardrail.blocklist.blocklist import Blocklist + + +@pytest.mark.L1 +def test_exact_match(): + """Test exact word matching.""" + prompt = "this contains badword in the middle" + word = "badword" + + match, message = Blocklist.check_partial_match(prompt, word, 0.4) + + assert match is True + assert "badword" in message + + +@pytest.mark.L1 +def test_no_match(): + """Test when there's no match.""" + prompt = "this is a clean prompt" + word = "badword" + + match, message = Blocklist.check_partial_match(prompt, word, 0.4) + + assert match is False + assert message == "" + + +@pytest.mark.L1 +def test_partial_match_with_threshold(): + """Test partial matching with character difference threshold.""" + prompt = "this contains badw0rd here" # One character different + word = "badword" + + # With threshold of 1.0 character difference, should match + match, _ = Blocklist.check_partial_match(prompt, word, 1.0) + assert match is True + + # With threshold of 0.5 character difference, should not match + match, _ = Blocklist.check_partial_match(prompt, word, 0.5) + assert match is False diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/profile_blocklist.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/profile_blocklist.py new file mode 100644 index 0000000000000000000000000000000000000000..e2ac5a37a54646f418ecfbbaa26dcb86887ff406 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/profile_blocklist.py @@ -0,0 +1,59 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import time + +from cosmos_policy._src.imaginaire.auxiliary.guardrail.blocklist.blocklist import Blocklist +from cosmos_policy._src.imaginaire.auxiliary.guardrail.common.core import GuardrailRunner + +# Test prompts +prompts = [ + "Create a beautiful landscape", + "Generate a detailed cinematic scene of a bustling medieval marketplace with merchants selling colorful spices, fabrics hanging from wooden stalls, cobblestone streets filled with people in period clothing, warm golden hour lighting filtering through ancient stone buildings, and smoke rising from food vendors cooking over open fires", + "Create an image of Adele", + "lot of words" * 100 + "Create an image of Adele", + "lot of words" * 100 + "Create an image of Adeele", + "Make an image of someone getting hurt badly", + "Generate a disturbing scene with graphic violence showing people fighting brutally with weapons, blood splattered everywhere, gruesome injuries visible in high detail, screaming faces contorted in pain and terror, dark atmosphere with dramatic lighting emphasizing the carnage and destruction, bodies lying motionless on the ground surrounded by pools of crimson liquid", + "Show me a peaceful garden with flowers blooming in spring sunlight", + "Design a futuristic space station orbiting Earth with advanced technology, sleek metallic surfaces reflecting starlight, astronauts in cutting-edge spacesuits conducting research, multiple docking bays with various spacecraft, solar panels gleaming in the cosmic void, and Earth's blue marble visible in the background through massive observation windows", +] + +checkpoint_dir = "/path/to/your/checkpoint/dir" # Change this path + +# Initialize +blocklist = Blocklist(checkpoint_dir=checkpoint_dir) +runner = GuardrailRunner(safety_models=[blocklist]) + +# Warm up +_ = runner.run_safety_check(prompts[0]) + + +times = [] +for prompt in prompts: + start = time.time() + safe, message = runner.run_safety_check(prompt) + end = time.time() + + elapsed = end - start + times.append(elapsed) + + print(f"Prompt: '{prompt[:50]}...'") + print(f"Safe: {safe}, Time: {elapsed:.4f}s") + if message: + print(f"Message: {message}") + print("-" * 40) + +print(f"\nAverage time: {sum(times) / len(times):.4f}s") diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/utils.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/utils.py new file mode 100644 index 0000000000000000000000000000000000000000..a98acf5c65295dc9d9a3a1a274ee9560059fbbea --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/blocklist/utils.py @@ -0,0 +1,45 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import re + +from cosmos_policy._src.imaginaire.utils import log + + +def read_keyword_list_from_dir(folder_path: str) -> list[str]: + """Read keyword list from all files in a folder.""" + output_list = [] + file_list = [] + # Get list of files in the folder + for file in os.listdir(folder_path): + if os.path.isfile(os.path.join(folder_path, file)): + file_list.append(file) + + # Process each file + for file in file_list: + file_path = os.path.join(folder_path, file) + try: + with open(file_path) as f: + output_list.extend([line.strip() for line in f.readlines()]) + except Exception as e: + log.error(f"Error reading file {file}: {e!s}") + + return output_list + + +def to_ascii(prompt: str) -> str: + """Convert prompt to ASCII.""" + return re.sub(r"[^\x00-\x7F]+", " ", prompt) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/common/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/common/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/common/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/common/core.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/common/core.py new file mode 100644 index 0000000000000000000000000000000000000000..d12a61327869c76e5e571b159cac6f5c43765c78 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/common/core.py @@ -0,0 +1,75 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any + +import numpy as np + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.checkpoint_db import get_checkpoint_by_uuid + +GUARDRAIL1_UUID = "9c7b7da4-2d95-45bb-9cb8-2eed954e9736" +GUARDRAIL1_CHECKPOINT_DIR = get_checkpoint_by_uuid(GUARDRAIL1_UUID).path + + +class ContentSafetyGuardrail: + def is_safe(self, **kwargs) -> tuple[bool, str]: + raise NotImplementedError("Child classes must implement the is_safe method") + + +class PostprocessingGuardrail: + def postprocess(self, frames: np.ndarray) -> np.ndarray: + raise NotImplementedError("Child classes must implement the postprocess method") + + +class GuardrailRunner: + def __init__( + self, + safety_models: list[ContentSafetyGuardrail] | None = None, + generic_block_msg: str = "", + generic_safe_msg: str = "", + postprocessors: list[PostprocessingGuardrail] | None = None, + ): + self.safety_models = safety_models + self.generic_block_msg = generic_block_msg + self.generic_safe_msg = generic_safe_msg if generic_safe_msg else "Prompt is safe" + self.postprocessors = postprocessors + + def run_safety_check(self, input: Any) -> tuple[bool, str]: + """Run the safety check on the input.""" + if not self.safety_models: + log.warning("No safety models found, returning safe") + return True, self.generic_safe_msg + + for guardrail in self.safety_models: + guardrail_name = str(guardrail.__class__.__name__).upper() + log.debug(f"Running guardrail: {guardrail_name}") + safe, message = guardrail.is_safe(input) + if not safe: + reasoning = self.generic_block_msg if self.generic_block_msg else f"{guardrail_name}: {message}" + return False, reasoning + return True, self.generic_safe_msg + + def postprocess(self, frames: np.ndarray) -> np.ndarray: + """Run the postprocessing on the video frames.""" + if not self.postprocessors: + log.warning("No postprocessors found, returning original frames") + return frames + + for guardrail in self.postprocessors: + guardrail_name = str(guardrail.__class__.__name__).upper() + log.debug(f"Running guardrail: {guardrail_name}") + frames = guardrail.postprocess(frames) + return frames diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/common/presets.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/common/presets.py new file mode 100644 index 0000000000000000000000000000000000000000..2364c976fda42b5b3828c0aed4ced0859196301c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/common/presets.py @@ -0,0 +1,78 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import numpy as np + +from cosmos_policy._src.imaginaire.auxiliary.guardrail.blocklist.blocklist import Blocklist +from cosmos_policy._src.imaginaire.auxiliary.guardrail.common.core import GuardrailRunner +from cosmos_policy._src.imaginaire.auxiliary.guardrail.face_blur_filter.face_blur_filter import RetinaFaceFilter +from cosmos_policy._src.imaginaire.auxiliary.guardrail.qwen3guard.qwen3guard import Qwen3Guard +from cosmos_policy._src.imaginaire.auxiliary.guardrail.video_content_safety_filter.video_content_safety_filter import ( + VideoContentSafetyFilter, +) +from cosmos_policy._src.imaginaire.utils import log + + +def create_text_guardrail_runner(offload_model_to_cpu: bool = False) -> GuardrailRunner: + """Create the text guardrail runner.""" + return GuardrailRunner( + safety_models=[ + Blocklist(), + Qwen3Guard(offload_model_to_cpu=offload_model_to_cpu), + ] + ) + + +def create_video_guardrail_runner(offload_model_to_cpu: bool = False) -> GuardrailRunner: + """Create the video guardrail runner.""" + return GuardrailRunner( + safety_models=[VideoContentSafetyFilter(offload_model_to_cpu=offload_model_to_cpu)], + postprocessors=[RetinaFaceFilter(offload_model_to_cpu=offload_model_to_cpu)], + ) + + +def run_text_guardrail(prompt: str, guardrail_runner: GuardrailRunner) -> bool: + """Run the text guardrail on the prompt, checking for content safety. + + Args: + prompt: The text prompt. + guardrail_runner: The text guardrail runner. + + Returns: + bool: Whether the prompt is safe. + """ + is_safe, message = guardrail_runner.run_safety_check(prompt) + if not is_safe: + log.critical(f"GUARDRAIL BLOCKED: {message}") + return is_safe + + +def run_video_guardrail(frames: np.ndarray, guardrail_runner: GuardrailRunner) -> np.ndarray | None: + """Run the video guardrail on the frames, checking for content safety and applying face blur. + + Args: + frames: The frames of the generated video. + guardrail_runner: The video guardrail runner. + + Returns: + The processed frames if safe, otherwise None. + """ + is_safe, message = guardrail_runner.run_safety_check(frames) + if not is_safe: + log.critical(f"GUARDRAIL BLOCKED: {message}") + return None + + frames = guardrail_runner.postprocess(frames) + return frames diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/blur_utils.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/blur_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..d52f69d220444a53027b3b4acc3bd192fc6eb76f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/blur_utils.py @@ -0,0 +1,35 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import cv2 +import numpy as np + + +def pixelate_face(face_img: np.ndarray, blocks: int = 5) -> np.ndarray: + """ + Pixelate a face region by reducing resolution and then upscaling. + + Args: + face_img: Face region to pixelate + blocks: Number of blocks to divide the face into (in each dimension) + + Returns: + Pixelated face region + """ + h, w = face_img.shape[:2] + # Shrink the image and scale back up to create pixelation effect + temp = cv2.resize(face_img, (blocks, blocks), interpolation=cv2.INTER_LINEAR) + pixelated = cv2.resize(temp, (w, h), interpolation=cv2.INTER_NEAREST) + return pixelated diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/face_blur_filter.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/face_blur_filter.py new file mode 100644 index 0000000000000000000000000000000000000000..f4e7035c3c36156ad3e6c511055313750bfe3fcf --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/face_blur_filter.py @@ -0,0 +1,242 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import argparse +import os +import warnings + +import numpy as np +import torch +from retinaface.data import cfg_re50 +from retinaface.layers.functions.prior_box import PriorBox +from retinaface.models.retinaface import RetinaFace +from torch.utils.data import DataLoader, TensorDataset +from tqdm import tqdm + +from cosmos_policy._src.imaginaire.auxiliary.guardrail.common.core import ( + GUARDRAIL1_CHECKPOINT_DIR, + GuardrailRunner, + PostprocessingGuardrail, +) +from cosmos_policy._src.imaginaire.auxiliary.guardrail.common.io_utils import ( + get_video_filepaths, + read_video, + save_video, +) +from cosmos_policy._src.imaginaire.auxiliary.guardrail.face_blur_filter.blur_utils import pixelate_face +from cosmos_policy._src.imaginaire.auxiliary.guardrail.face_blur_filter.retinaface_utils import ( + decode_batch, + filter_detected_boxes, + load_model, +) +from cosmos_policy._src.imaginaire.utils import log, misc + +# RetinaFace model constants from https://github.com/biubug6/Pytorch_Retinaface/blob/master/detect.py +TOP_K = 5_000 +KEEP_TOP_K = 750 +NMS_THRESHOLD = 0.4 + + +class RetinaFaceFilter(PostprocessingGuardrail): + def __init__( + self, + batch_size: int = 1, + confidence_threshold: float = 0.7, + offload_model_to_cpu: bool = True, + ) -> None: + """ + Initialize the RetinaFace model for face detection and blurring. + + Args: + checkpoint: Path to the RetinaFace checkpoint file + batch_size: Batch size for RetinaFace inference and processing + confidence_threshold: Minimum confidence score to consider a face detection + offload_model_to_cpu (bool, optional): Whether to offload the model to CPU. Defaults to True. + """ + self.checkpoint = f"{GUARDRAIL1_CHECKPOINT_DIR}/face_blur_filter/Resnet50_Final.pth" + self.cfg = cfg_re50 + self.batch_size = batch_size + self.confidence_threshold = confidence_threshold + self.dtype = torch.float32 + self.offload_model = offload_model_to_cpu + + # Disable loading ResNet pretrained weights + self.cfg["pretrain"] = False + with warnings.catch_warnings(): + warnings.simplefilter("ignore") + self.net = RetinaFace(cfg=self.cfg, phase="test") + + # Load from RetinaFace pretrained checkpoint + if not offload_model_to_cpu: + self.net = load_model(self.net, self.checkpoint, False) + self.net.to("cuda", dtype=self.dtype).eval() + log.debug("Moved face blur filter to GPU") + else: + self.net = load_model(self.net, self.checkpoint, True) + self.net.to("cpu", dtype=self.dtype).eval() + log.debug("Moved face blur filter to CPU") + + def preprocess_frames(self, frames: np.ndarray) -> torch.Tensor: + """Preprocess a sequence of frames for face detection. + + Args: + frames: Input frames + + Returns: + Preprocessed frames tensor + """ + with torch.no_grad(): + frames_tensor = torch.from_numpy(frames).to("cuda", dtype=self.dtype) # Shape: [T, H, W, C] + frames_tensor = frames_tensor.permute(0, 3, 1, 2) # Shape: [T, C, H, W] + frames_tensor = frames_tensor[:, [2, 1, 0], :, :] # RGB to BGR to match RetinaFace model input + means = torch.tensor([104.0, 117.0, 123.0], device="cuda", dtype=self.dtype).view(1, 3, 1, 1) + frames_tensor = frames_tensor - means # Subtract mean BGR values for each channel + return frames_tensor + + def blur_detected_faces( + self, + frames: np.ndarray, + batch_loc: torch.Tensor, + batch_conf: torch.Tensor, + prior_data: torch.Tensor, + scale: torch.Tensor, + min_size: tuple[int] = (20, 20), + ) -> list[np.ndarray]: + """Blur detected faces in a batch of frames using RetinaFace predictions. + + Args: + frames: Input frames + batch_loc: Batched location predictions + batch_conf: Batched confidence scores + prior_data: Prior boxes for the video + scale: Scale factor for resizing detections + min_size: Minimum size of a detected face region in pixels + + Returns: + Processed frames with pixelated faces + """ + with torch.no_grad(): + batch_boxes = decode_batch(batch_loc, prior_data, self.cfg["variance"]) + batch_boxes = batch_boxes * scale + + blurred_frames = [] + for i, boxes in enumerate(batch_boxes): + boxes = boxes.detach().cpu().numpy() + scores = batch_conf[i, :, 1].detach().cpu().numpy() + + filtered_boxes = filter_detected_boxes( + boxes, + scores, + confidence_threshold=self.confidence_threshold, + nms_threshold=NMS_THRESHOLD, + top_k=TOP_K, + keep_top_k=KEEP_TOP_K, + ) + + frame = frames[i] + for box in filtered_boxes: + x1, y1, x2, y2 = map(int, box) + # Ignore bounding boxes smaller than the minimum size + if x2 - x1 < min_size[0] or y2 - y1 < min_size[1]: + continue + max_h, max_w = frame.shape[:2] + face_roi = frame[max(y1, 0) : min(y2, max_h), max(x1, 0) : min(x2, max_w)] + blurred_face = pixelate_face(face_roi) + frame[max(y1, 0) : min(y2, max_h), max(x1, 0) : min(x2, max_w)] = blurred_face + blurred_frames.append(frame) + + return blurred_frames + + def postprocess(self, frames: np.ndarray) -> np.ndarray: + """Blur faces in a sequence of frames. + + Args: + frames: Input frames + + Returns: + Processed frames with pixelated faces + """ + # Create dataset and dataloader + if self.offload_model: + self.net = self.net.to("cuda") + log.debug("Move face blur filter to GPU") + frames_tensor = self.preprocess_frames(frames) + dataset = TensorDataset(frames_tensor) + dataloader = DataLoader(dataset, batch_size=self.batch_size, shuffle=False) + processed_frames, processed_batches = [], [] + + prior_data, scale = None, None + for i, batch in enumerate(dataloader): + batch = batch[0] + h, w = batch.shape[-2:] # Batch shape: [C, H, W] + + with torch.no_grad(): + # Generate priors for the video + if prior_data is None: + priorbox = PriorBox(self.cfg, image_size=(h, w)) + priors = priorbox.forward() + priors = priors.to("cuda", dtype=self.dtype) + prior_data = priors.data + + # Get scale for resizing detections + if scale is None: + scale = torch.Tensor([w, h, w, h]) + scale = scale.to("cuda", dtype=self.dtype) + + batch_loc, batch_conf, _ = self.net(batch) + + # Blur detected faces in each batch of frames + start_idx = i * self.batch_size + end_idx = min(start_idx + self.batch_size, len(frames)) + processed_batches.append( + self.blur_detected_faces(frames[start_idx:end_idx], batch_loc, batch_conf, prior_data, scale) + ) + + processed_frames = [frame for batch in processed_batches for frame in batch] + if self.offload_model: + self.net = self.net.to("cpu") + log.debug("Offload face blur filter to CPU") + return np.array(processed_frames) + + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--input_dir", type=str, required=True, help="Path containing input videos") + parser.add_argument("--output_dir", type=str, required=True, help="Path for saving processed videos") + return parser.parse_args() + + +def main(args): + filepaths = get_video_filepaths(args.input_dir) + if not filepaths: + log.error(f"No video files found in directory: {args.input_dir}") + return + + face_blur = RetinaFaceFilter() + postprocessing_runner = GuardrailRunner(postprocessors=[face_blur]) + os.makedirs(args.output_dir, exist_ok=True) + + for filepath in tqdm(filepaths): + video_data = read_video(filepath) + with misc.timer("face blur filter"): + frames = postprocessing_runner.postprocess(video_data.frames) + + output_path = os.path.join(args.output_dir, os.path.basename(filepath)) + save_video(output_path, frames, video_data.fps) + + +if __name__ == "__main__": + args = parse_args() + main(args) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/retinaface_utils.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/retinaface_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..f1acb5eaf8862ccf3d8ddaae8e1028aff38e477d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/face_blur_filter/retinaface_utils.py @@ -0,0 +1,117 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import numpy as np +import torch +from retinaface.utils.nms.py_cpu_nms import py_cpu_nms + +from cosmos_policy._src.imaginaire.utils import log + + +# Adapted from https://github.com/biubug6/Pytorch_Retinaface/blob/master/detect.py +def filter_detected_boxes(boxes, scores, confidence_threshold, nms_threshold, top_k, keep_top_k): + """Filter boxes based on confidence score and remove overlapping boxes using NMS.""" + # Keep detections with confidence above threshold + inds = np.where(scores > confidence_threshold)[0] + boxes = boxes[inds] + scores = scores[inds] + + # Sort by confidence and keep top K detections + order = scores.argsort()[::-1][:top_k] + boxes = boxes[order] + scores = scores[order] + + # Run non-maximum-suppression (NMS) to remove overlapping boxes + dets = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False) + keep = py_cpu_nms(dets, nms_threshold) + dets = dets[keep, :] + dets = dets[:keep_top_k, :] + boxes = dets[:, :-1] + return boxes + + +# Adapted from https://github.com/biubug6/Pytorch_Retinaface/blob/master/utils/box_utils.py to handle batched inputs +def decode_batch(loc, priors, variances): + """Decode batched locations from predictions using priors and variances. + + Args: + loc (tensor): Batched location predictions for loc layers. + Shape: [batch_size, num_priors, 4] + priors (tensor): Prior boxes in center-offset form. + Shape: [num_priors, 4] + variances: (list[float]): Variances of prior boxes. + + Return: + Decoded batched bounding box predictions + Shape: [batch_size, num_priors, 4] + """ + batch_size = loc.size(0) + priors = priors.unsqueeze(0).expand(batch_size, -1, -1) + + boxes = torch.cat( + ( + priors[:, :, :2] + loc[:, :, :2] * variances[0] * priors[:, :, 2:], + priors[:, :, 2:] * torch.exp(loc[:, :, 2:] * variances[1]), + ), + dim=2, + ) + + boxes[:, :, :2] -= boxes[:, :, 2:] / 2 + boxes[:, :, 2:] += boxes[:, :, :2] + return boxes + + +# Adapted from https://github.com/biubug6/Pytorch_Retinaface/blob/master/detect.py +def _check_keys(model, pretrained_state_dict): + ckpt_keys = set(pretrained_state_dict.keys()) + model_keys = set(model.state_dict().keys()) + used_pretrained_keys = model_keys & ckpt_keys + unused_pretrained_keys = ckpt_keys - model_keys + missing_keys = model_keys - ckpt_keys + log.debug(f"Missing keys:{len(missing_keys)}") + log.debug(f"Unused checkpoint keys:{len(unused_pretrained_keys)}") + log.debug(f"Used keys:{len(used_pretrained_keys)}") + assert len(used_pretrained_keys) > 0, "load NONE from pretrained checkpoint" + return True + + +# Adapted from https://github.com/biubug6/Pytorch_Retinaface/blob/master/detect.py +def _remove_prefix(state_dict, prefix): + """Old version of the model is stored with all names of parameters sharing common prefix 'module.'""" + log.debug(f"Removing prefix '{prefix}'") + + def f(x): + return x.split(prefix, 1)[-1] if x.startswith(prefix) else x + + return {f(key): value for key, value in state_dict.items()} + + +# Adapted from https://github.com/biubug6/Pytorch_Retinaface/blob/master/detect.py +def load_model(model, pretrained_path, load_to_cpu): + log.debug(f"Loading pretrained model from {pretrained_path}") + if load_to_cpu: + pretrained_dict = torch.load(pretrained_path, map_location=lambda storage, loc: storage, weights_only=True) + else: + device = torch.cuda.current_device() + pretrained_dict = torch.load( + pretrained_path, map_location=lambda storage, loc: storage.cuda(device), weights_only=True + ) + if "state_dict" in pretrained_dict.keys(): + pretrained_dict = _remove_prefix(pretrained_dict["state_dict"], "module.") + else: + pretrained_dict = _remove_prefix(pretrained_dict, "module.") + _check_keys(model, pretrained_dict) + model.load_state_dict(pretrained_dict, strict=False) + return model diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/llamaGuard3/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/llamaGuard3/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/llamaGuard3/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/llamaGuard3/categories.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/llamaGuard3/categories.py new file mode 100644 index 0000000000000000000000000000000000000000..f8d5a95d4dce1202e3acec0e10177c97c1e5924e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/llamaGuard3/categories.py @@ -0,0 +1,31 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +UNSAFE_CATEGORIES = { + "S1": "Violent Crimes.", + "S2": "Non-Violent Crimes.", + "S3": "Sex Crimes.", + "S4": "Child Exploitation.", + "S5": "Defamation.", + "S6": "Specialized Advice.", + "S7": "Privacy.", + "S8": "Intellectual Property.", + "S9": "Indiscriminate Weapons.", + "S10": "Hate.", + "S11": "Self-Harm.", + "S12": "Sexual Content.", + "S13": "Elections.", + "s14": "Code Interpreter Abuse.", +} diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/llamaGuard3/llamaGuard3.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/llamaGuard3/llamaGuard3.py new file mode 100644 index 0000000000000000000000000000000000000000..f4f794153fde3a03d9f03bb4ec40ccdf0460ed69 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/llamaGuard3/llamaGuard3.py @@ -0,0 +1,130 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import argparse + +import torch +from transformers import AutoModelForCausalLM, AutoTokenizer + +from cosmos_policy._src.imaginaire.auxiliary.guardrail.common.core import ContentSafetyGuardrail, GuardrailRunner +from cosmos_policy._src.imaginaire.auxiliary.guardrail.llamaGuard3.categories import UNSAFE_CATEGORIES +from cosmos_policy._src.imaginaire.utils import log, misc + +SAFE = misc.Color.green("SAFE") +UNSAFE = misc.Color.red("UNSAFE") + + +class LlamaGuard3(ContentSafetyGuardrail): + def __init__( + self, + offload_model_to_cpu: bool = True, + ) -> None: + """Llama Guard 3 model for text filtering safety check. + + Args: + checkpoint_dir (str): Path to the checkpoint directory. + offload_model_to_cpu (bool, optional): Whether to offload the model to CPU. Defaults to True. + """ + self.offload_model = offload_model_to_cpu + self.dtype = torch.bfloat16 + + model_id = "meta-llama/Llama-Guard-3-8B" + + self.model = AutoModelForCausalLM.from_pretrained(model_id) + self.tokenizer = AutoTokenizer.from_pretrained(model_id) + + # Move model to GPU unless offload_model_to_cpu is True + if not offload_model_to_cpu: + self.model = self.model.to("cuda", dtype=self.dtype).eval() + log.debug("Moved llamaGuard3 model to GPU") + else: + self.model = self.model.to("cpu", dtype=self.dtype).eval() + log.debug("Moved llamaGuard3 model to CPU") + + def get_llamaGuard3_block_message(self, moderation_output: str) -> str: + """Extract the blocked category from the Llama Guard 3 model output.""" + block_msg = "Prompt blocked by Llama Guard 3." + try: + lines = moderation_output.splitlines() + categories_detected = [] + for line in lines[1:]: + line_stripped = line.split("<|eot_id|>")[0].strip() + for catagory in line_stripped.split(","): + catagory = catagory.strip() + if catagory not in UNSAFE_CATEGORIES: + log.warning(f"Unrecognized category from moderation output: {catagory}") + else: + categories_detected.append(catagory) + if len(categories_detected) > 0: + blocked_catagories = ", ".join([UNSAFE_CATEGORIES[catagory][:-1] for catagory in categories_detected]) + block_msg = f"{block_msg} Violations: {blocked_catagories}." + except Exception as e: + log.warning(f"Unable to extract blocked category from Llama Guard 3 output: {e}") + return block_msg + + def filter_llamaGuard3_output(self, prompt: str) -> tuple[bool, str]: + """Filter the Llama Guard 3 model output and return the safety status and message.""" + conversation = [{"role": "user", "content": prompt}] + if self.offload_model: + self.model = self.model.to("cuda") + log.debug("Move llamaGuard3 model to GPU") + input_ids = self.tokenizer.apply_chat_template( + conversation, categories=UNSAFE_CATEGORIES, return_tensors="pt" + ).to("cuda") + prompt_len = input_ids.shape[1] + output = self.model.generate( + input_ids=input_ids, + max_new_tokens=100, + return_dict_in_generate=True, + pad_token_id=0, + ) + generated_tokens = output.sequences[:, prompt_len:] + moderation_output = self.tokenizer.decode(generated_tokens[0], skip_special_tokens=False).strip() + if self.offload_model: + self.model = self.model.to("cpu") + log.debug("Offload llamaGuard3 model to CPU") + if "unsafe" in moderation_output.lower(): + block_msg = self.get_llamaGuard3_block_message(moderation_output) + return False, block_msg + else: + return True, "" + + def is_safe(self, prompt: str) -> tuple[bool, str]: + """Check if the input prompt is safe according to the Llama Guard 3 model.""" + try: + return self.filter_llamaGuard3_output(prompt) + except Exception as e: + log.error(f"Unexpected error occurred when running Llama Guard 3 guardrail: {e}") + return True, "Unexpected error occurred when running Llama Guard 3 guardrail." + + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--prompt", type=str, required=True, help="Input prompt") + return parser.parse_args() + + +def main(args): + llamaGuard3 = LlamaGuard3() + runner = GuardrailRunner(safety_models=[llamaGuard3]) + with misc.timer("Llama Guard 3 safety check"): + safety, message = runner.run_safety_check(args.prompt) + log.info(f"Input is: {'SAFE' if safety else 'UNSAFE'}") + log.info(f"Message: {message}") if not safety else None + + +if __name__ == "__main__": + args = parse_args() + main(args) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/qwen3guard/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/qwen3guard/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/qwen3guard/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/qwen3guard/categories.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/qwen3guard/categories.py new file mode 100644 index 0000000000000000000000000000000000000000..b990489f71ace8ef4c7a7fd896419c343a82c1e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/qwen3guard/categories.py @@ -0,0 +1,23 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +UNSAFE_CATEGORIES = { + "S1": "Violent", + "S2": "Non-violent Illegal Acts", + "S3": "Sexual Content or Sexual Acts", + "S4": "Suicide & Self-Harm", + "S5": "Unethical Acts", + "S6": "Jailbreak", +} diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/qwen3guard/qwen3guard.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/qwen3guard/qwen3guard.py new file mode 100644 index 0000000000000000000000000000000000000000..fe6a4f700e3626f595296fe24f700f4546420668 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/qwen3guard/qwen3guard.py @@ -0,0 +1,102 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import argparse +import re + +import torch +from transformers import AutoModelForCausalLM, AutoTokenizer + +from cosmos_policy._src.imaginaire.auxiliary.guardrail.common.core import ContentSafetyGuardrail, GuardrailRunner +from cosmos_policy._src.imaginaire.auxiliary.guardrail.qwen3guard.categories import UNSAFE_CATEGORIES +from cosmos_policy._src.imaginaire.utils import log, misc + +SAFE = misc.Color.green("SAFE") +UNSAFE = misc.Color.red("UNSAFE") + + +class Qwen3Guard(ContentSafetyGuardrail): + def __init__( + self, + offload_model_to_cpu: bool = True, + ) -> None: + """Llama Guard 3 model for text filtering safety check. + + Args: + checkpoint_dir (str): Path to the checkpoint directory. + offload_model_to_cpu (bool, optional): Whether to offload the model to CPU. Defaults to True. + """ + self.offload_model = offload_model_to_cpu + self.dtype = torch.bfloat16 + + model_id = "Qwen/Qwen3Guard-Gen-0.6B" + + self.model = AutoModelForCausalLM.from_pretrained(model_id) + self.tokenizer = AutoTokenizer.from_pretrained(model_id) + + # Move model to GPU unless offload_model_to_cpu is True + if not offload_model_to_cpu: + self.model = self.model.to("cuda", dtype=self.dtype).eval() + log.debug("Moved llamaGuard3 model to GPU") + else: + self.model = self.model.to("cpu", dtype=self.dtype).eval() + log.debug("Moved Qwen3Guard model to CPU") + + def extract_label_and_categories(self, prompt): + safe_pattern = r"Safety: (Safe|Unsafe|Controversial)" + category_pattern = r"(" + "|".join(UNSAFE_CATEGORIES.values()) + ")" + messages = [{"role": "user", "content": prompt}] + + text = self.tokenizer.apply_chat_template(messages, tokenize=False) + model_inputs = self.tokenizer([text], return_tensors="pt").to(self.model.device) + generated_ids = self.model.generate(**model_inputs, max_new_tokens=128) + output_ids = generated_ids[0][len(model_inputs.input_ids[0]) :].tolist() + content = self.tokenizer.decode(output_ids, skip_special_tokens=True) + + safe_label_match = re.search(safe_pattern, content) + label = safe_label_match.group(1) if safe_label_match else None + categories = re.findall(category_pattern, content) + if label.lower() == "unsafe": + return False, f"Prompt blocked by Qwen3Guard. Safety: {label}, Categories: {categories}" + else: + return True, "" + + def is_safe(self, prompt: str) -> tuple[bool, str]: + """Check if the input prompt is safe according to the Qwen3Guard model.""" + try: + return self.extract_label_and_categories(prompt) + except Exception as e: + log.error(f"Unexpected error occurred when running Qwen3Guard guardrail: {e}") + return True, "Unexpected error occurred when running Qwen3Guard guardrail." + + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--prompt", type=str, required=True, help="Input prompt") + return parser.parse_args() + + +def main(args): + qwen3guard = Qwen3Guard() + runner = GuardrailRunner(safety_models=[qwen3guard]) + with misc.timer("Qwen3Guard safety check"): + safety, message = runner.run_safety_check(args.prompt) + log.info(f"Input is: {'SAFE' if safety else 'UNSAFE'}") + log.info(f"Message: {message}") if not safety else None + + +if __name__ == "__main__": + args = parse_args() + main(args) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/model.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/model.py new file mode 100644 index 0000000000000000000000000000000000000000..858da6603ff64229ac76f1fe98dc4bd9d2fa6ad7 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/model.py @@ -0,0 +1,60 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import attrs +import torch +import torch.nn as nn + +from cosmos_policy._src.imaginaire.config import make_freezable + + +@make_freezable +@attrs.define(slots=False) +class ModelConfig: + input_size: int = 1152 + num_classes: int = 7 + + +class SafetyClassifier(nn.Module): + def __init__(self, input_size: int = 1024, num_classes: int = 2): + super().__init__() + self.input_size = input_size + self.num_classes = num_classes + self.layers = nn.Sequential( + nn.Linear(self.input_size, 512), + nn.BatchNorm1d(512), + nn.ReLU(), + nn.Linear(512, 256), + nn.BatchNorm1d(256), + nn.ReLU(), + nn.Linear(256, self.num_classes), + # Note: No activation function here; CrossEntropyLoss expects raw logits + ) + + def forward(self, x): + return self.layers(x) + + +class VideoSafetyModel(nn.Module): + def __init__(self, config: ModelConfig) -> None: + super().__init__() + self.config = config + self.num_classes = config.num_classes + self.network = SafetyClassifier(input_size=config.input_size, num_classes=self.num_classes) + + @torch.inference_mode() + def forward(self, data_batch: dict[str, torch.Tensor]) -> dict[str, torch.Tensor]: + logits = self.network(data_batch["data"].cuda()) + return {"logits": logits} diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/video_content_safety_filter.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/video_content_safety_filter.py new file mode 100644 index 0000000000000000000000000000000000000000..aac527fcfabaa4562ee95887a602c9ac945faae4 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/video_content_safety_filter.py @@ -0,0 +1,187 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import argparse +import json +import os +from collections.abc import Iterable + +import torch +from PIL import Image + +from cosmos_policy._src.imaginaire.auxiliary.guardrail.common.core import ( + GUARDRAIL1_CHECKPOINT_DIR, + ContentSafetyGuardrail, + GuardrailRunner, +) +from cosmos_policy._src.imaginaire.auxiliary.guardrail.common.io_utils import get_video_filepaths, read_video +from cosmos_policy._src.imaginaire.auxiliary.guardrail.video_content_safety_filter.model import ( + ModelConfig, + VideoSafetyModel, +) +from cosmos_policy._src.imaginaire.auxiliary.guardrail.video_content_safety_filter.vision_encoder import SigLIPEncoder +from cosmos_policy._src.imaginaire.utils import log, misc + +# Define the class index to class name mapping for multi-class classification +CLASS_IDX_TO_NAME = { + 0: "Safe", + 1: "Sexual_Content", + 3: "Drugs", + 4: "Child_Abuse", + 5: "Hate_and_Harassment", + 6: "Self-Harm", +} + +CUTOFF_UNSAFE_FRAMES_PERCENT = 10 # 10% of frames are unsafe, then the video is unsafe + + +class VideoContentSafetyFilter(ContentSafetyGuardrail): + def __init__( + self, + offload_model_to_cpu: bool = True, + ) -> None: + """Video content safety filter model. + + Args: + checkpoint_dir (str): Path to the checkpoint directory. + offload_model_to_cpu (bool, optional): Whether to offload the model to CPU. Defaults to True. + """ + self.offload_model = offload_model_to_cpu + self.dtype = torch.float32 + self.checkpoint_dir = os.path.join(GUARDRAIL1_CHECKPOINT_DIR, "video_content_safety_filter") + + # Use ModelConfig directly for inference configuration + model_config = ModelConfig(input_size=1152, num_classes=7) + + # Load the multi-class classifier and initialize the SigLIP encoder + self.model = VideoSafetyModel(model_config) + safety_filter_local_path = os.path.join(self.checkpoint_dir, "safety_filter.pt") + checkpoint = torch.load(safety_filter_local_path, map_location=torch.device("cpu"), weights_only=True) + self.model.load_state_dict(checkpoint["model"]) + self.encoder = SigLIPEncoder(device="cuda", dtype=self.dtype) + if offload_model_to_cpu: + self.encoder.to("cpu") + self.model = self.model.to("cpu", dtype=self.dtype).eval() + log.debug("Moved video content safety filter to CPU") + else: + self.encoder.to("cuda") + self.model = self.model.to("cuda", dtype=self.dtype).eval() + log.debug("Moved video content safety filter to GPU") + + @torch.inference_mode() + def __infer(self, pil_image: Image.Image) -> int: + """Infer the class of the image.""" + image_embs = self.encoder.encode_image(pil_image) + logits = self.model.network(image_embs) + probabilities = torch.nn.functional.softmax(logits, dim=-1) + predicted_class = int(torch.argmax(probabilities, dim=-1).item()) + return predicted_class + + def _to_cuda_if_offload(self): + if self.offload_model: + self.encoder = self.encoder.to("cuda") + self.model = self.model.to("cuda") + log.debug("Move video content safety filter to GPU") + + def _to_cpu_if_offload(self): + if self.offload_model: + self.encoder = self.encoder.to("cpu") + self.model = self.model.to("cpu") + log.debug("Offload video content safety filter to CPU") + + def is_safe_file(self, filepath: str) -> bool: + """Check if the video file is safe.""" + video_data = read_video(filepath) + + # Sample frames at 2 FPS + sample_rate = 2 # frames per second + frame_interval = int(video_data.fps / sample_rate) + frame_numbers = list(range(0, int(video_data.fps * video_data.duration), frame_interval)) + frames = [video_data.frames[frame_number] for frame_number in frame_numbers] + return self.is_safe_frames(frames) + + def is_safe_frames(self, frames: Iterable) -> bool: + """Check if the video frames are safe.""" + is_safe = True + frame_scores = [] + unsafe_frame_count = 0 + total_frame_count = 0 + + self._to_cuda_if_offload() + for frame_number, frame in enumerate(frames): + total_frame_count += 1 + try: + pil_image = Image.fromarray(frame) + predicted_class = self.__infer(pil_image) + class_name = CLASS_IDX_TO_NAME.get(predicted_class, "Unknown") + frame_scores.append({"frame_number": frame_number, "class": class_name}) + + # If any frame considered in the list of unsafe categories, mark the video as unsafe + if class_name != "Safe" and class_name in CLASS_IDX_TO_NAME.values(): + log.warning(f"Unsafe frame detected in frame_number {frame_number}: {class_name}") + unsafe_frame_count += 1 + + except Exception as e: + log.warning(f"Warning: Failed to run safety classifier on frame_number {frame_number}. Exception: {e}") + continue + + if (unsafe_frame_count / total_frame_count) > (CUTOFF_UNSAFE_FRAMES_PERCENT / 100): + is_safe = False + log.warning( + f"Unsafe frame count {unsafe_frame_count} is greater than {CUTOFF_UNSAFE_FRAMES_PERCENT}% of total frames {total_frame_count}" + ) + + video_data = { + "is_safe": is_safe, + "frame_scores": frame_scores, + } + self._to_cpu_if_offload() + log.debug(f"Frames data: {json.dumps(video_data, indent=4)}") + return is_safe + + def is_safe(self, input: str | Iterable) -> tuple[bool, str]: + if isinstance(input, str): + is_safe = self.is_safe_file(input) + return is_safe, "safe video detected" if is_safe else "unsafe video detected" + elif isinstance(input, Iterable): + is_safe = self.is_safe_frames(input) + return is_safe, "safe frames detected" if is_safe else "unsafe frames detected" + else: + raise ValueError(f"Input type {type(input)} not supported.") + + +def parse_args(): + parser = argparse.ArgumentParser() + parser.add_argument("--input_dir", type=str, required=True, help="Path containing input videos") + return parser.parse_args() + + +def main(args): + filepaths = get_video_filepaths(args.input_dir) + if not filepaths: + log.error(f"No video files found in directory: {args.input_dir}") + return + + video_filter = VideoContentSafetyFilter() + runner = GuardrailRunner(safety_models=[video_filter], generic_safe_msg="Video is safe") + + for filepath in filepaths: + with misc.timer("video content safety filter"): + _ = runner.run_safety_check(filepath) + + +if __name__ == "__main__": + args = parse_args() + main(args) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/vision_encoder.py b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/vision_encoder.py new file mode 100644 index 0000000000000000000000000000000000000000..283446e380a3ef43871f06aeebae5ebd68f65707 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/auxiliary/guardrail/video_content_safety_filter/vision_encoder.py @@ -0,0 +1,42 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import torch +from PIL import Image +from transformers import SiglipModel, SiglipProcessor + + +class SigLIPEncoder(torch.nn.Module): + def __init__( + self, + device="cuda" if torch.cuda.is_available() else "cpu", # noqa: B008 + dtype=torch.float32, + ) -> None: + super().__init__() + self.device = device + self.dtype = dtype + model_id = "google/siglip-so400m-patch14-384" + self.model = SiglipModel.from_pretrained(model_id) + self.processor = SiglipProcessor.from_pretrained(model_id) + self.model.to(self.device, dtype=self.dtype).eval() + + @torch.inference_mode() + def encode_image(self, input_img: Image.Image) -> torch.Tensor: + """Encode an image into a feature vector.""" + with torch.no_grad(): + inputs = self.processor(images=input_img, return_tensors="pt").to(self.device, dtype=self.dtype) + image_features = self.model.get_image_features(**inputs) + image_features /= image_features.norm(dim=-1, keepdim=True) + return image_features diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/every_n.py b/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/every_n.py new file mode 100644 index 0000000000000000000000000000000000000000..0124b2570910295e55761f05a9c2d23a9d59a17d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/every_n.py @@ -0,0 +1,85 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from abc import abstractmethod +from typing import Optional + +import torch + +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.callback import Callback + + +class EveryN(Callback): + def __init__( + self, + every_n: Optional[int] = None, + step_size: int = 1, + barrier_after_run: bool = True, + run_at_start: bool = False, + ) -> None: + """Constructor for `EveryN`. + + Args: + every_n (int): Frequency with which callback is run during training. + step_size (int): Size of iteration step count. Default 1. + barrier_after_run (bool): Whether to have a distributed barrier after each execution. Default True, to avoid timeouts. + run_at_start (bool): Whether to run at the beginning of training. Default False. + """ + self.every_n = every_n + if self.every_n == 0: + log.warning( + f"every_n is set to 0. Callback {self.__class__.__name__} will be invoked only once in the beginning of the training. Calls happens on_training_step_end will be skipped." + ) + + self.step_size = step_size + self.barrier_after_run = barrier_after_run + self.run_at_start = run_at_start + + def on_training_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + # every_n = 0 is a special case which means every_n_impl will be called only once in the beginning of the training + if self.every_n != 0: + trainer = self.trainer + global_step = iteration // self.step_size + should_run = (iteration == 1 and self.run_at_start) or ( + global_step % self.every_n == 0 + ) # (self.every_n - 1) + if should_run: + log.debug(f"Callback {self.__class__.__name__} fired on train_batch_end step {global_step}") + self.every_n_impl(trainer, model, data_batch, output_batch, loss, iteration) + log.debug(f"Callback {self.__class__.__name__} finished on train_batch_end step {global_step}") + # add necessary barrier to avoid timeout + if self.barrier_after_run: + distributed.barrier() + + @abstractmethod + def every_n_impl( + self, + trainer: ImaginaireTrainer, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int, + ) -> None: ... diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/image_grad_clip.py b/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/image_grad_clip.py new file mode 100644 index 0000000000000000000000000000000000000000..dc85b145c246b3814cc1e61fd94d7ad6d218bfc1 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/image_grad_clip.py @@ -0,0 +1,78 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import List, Optional + +import torch +import wandb +from torch.distributed.fsdp import FullyShardedDataParallel as FSDP + +from cosmos_policy._src.imaginaire.utils import distributed +from cosmos_policy._src.imaginaire.utils.callback import Callback + + +@torch.jit.script +def _fused_nan_to_num(params: List[torch.Tensor]): + for param in params: + torch.nan_to_num(param, nan=0.0, posinf=0.0, neginf=0.0, out=param) + + +class GradClip(Callback): + def __init__( + self, clip_norm=1.0, force_finite: bool = True, model_key: Optional[str] = None, fsdp_enabled: bool = False + ): + self.clip_norm = clip_norm + self.force_finite = force_finite + self.model_key = model_key + self.fsdp_enabled = fsdp_enabled + + def on_before_optimizer_step( + self, + model_ddp: distributed.DistributedDataParallel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int = 0, + ) -> None: + del optimizer, scheduler + if isinstance(model_ddp, distributed.DistributedDataParallel): + model = model_ddp.module + else: + model = model_ddp + + # select sub-network if specified + if self.model_key is not None: + items = self.model_key.split(".") + for item in items: + model = getattr(model, item) + + if self.force_finite: + params = [] + for param in model.parameters(): + if param.grad is not None: + params.append(param.grad) + # torch.nan_to_num(param.grad, nan=0, posinf=0, neginf=0, out=param.grad) + _fused_nan_to_num(params) + + # check if FSDP is used + if isinstance(model, FSDP) and self.fsdp_enabled: + total_norm = model.clip_grad_norm_(self.clip_norm) + else: + total_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), self.clip_norm, foreach=True) + + # log + if iteration % self.config.trainer.logging_iter == 0: + if wandb.run: + wandb.log({"clip_grad_norm": total_norm.item()}, step=iteration) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/manual_gc.py b/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/manual_gc.py new file mode 100644 index 0000000000000000000000000000000000000000..ed4d7352324a265e14cf2bda8e8b73616510e5cb --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/callbacks/manual_gc.py @@ -0,0 +1,49 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import gc + +from cosmos_policy._src.imaginaire.callbacks.every_n import EveryN +from cosmos_policy._src.imaginaire.utils import log + + +class ManualGarbageCollection(EveryN): + """ + Disable auto gc and manually trigger garbage collection every N iterations + It is super useful for large scale training to reduce gpu sync time! + Can reach 50% speedup. + + It is important to note that this callback only disables gc in main process and have auto gc enabled in subprocesses. + + We start disable gc after warm_up iterations to avoid disabling gc in subprocesses, such as dataloader, which can cause OOM + """ + + def __init__(self, *args, warm_up: int = 5, **kwargs): + kwargs["barrier_after_run"] = False + super().__init__(*args, **kwargs) + + self.counter = 0 + self.warm = warm_up + + def every_n_impl(self, trainer, model, data_batch, output_batch, loss, iteration): + del trainer, model, data_batch, output_batch, loss + self.counter += 1 + if self.counter < self.warm: + return + if self.counter == self.warm: + gc.disable() + log.critical("Garbage collection disabled") + + gc.collect(1) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/base.py b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/base.py new file mode 100644 index 0000000000000000000000000000000000000000..673d5ac7decff88fb3c3267c0103ff27720edff5 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/base.py @@ -0,0 +1,177 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +from abc import ABC, abstractmethod +from typing import Optional + +import torch + +from cosmos_policy._src.imaginaire.config import CheckpointConfig, JobConfig +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import callback +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + + +class AbstractCheckpointer(ABC): + """The checkpointer class. Supports checkpoint saving/loading to both local disk or object store.""" + + def __init__( + self, + config_checkpoint: CheckpointConfig, + config_job: JobConfig, + callbacks: Optional[callback.CallBackGroup] = None, + ): + """Constructor of the checkpointer. + + Args: + config_checkpoint (CheckpointConfig): The config object for the checkpointer. + """ + self.config_checkpoint = config_checkpoint + # Set the callback functions. + self.callbacks = callbacks + self.save_to_object_store = config_checkpoint.save_to_object_store.enabled + self.load_from_object_store = config_checkpoint.load_from_object_store.enabled + + # Set checkpoint directories for local and object store paths + self._local_dirname = os.path.join(config_job.path_local, "checkpoints") + self._object_store_dirname = os.path.join(config_job.path, "checkpoints") + + self.strict_resume = config_checkpoint.strict_resume + self.load_path = config_checkpoint.load_path or None + self.load_training_state = config_checkpoint.load_training_state + self.only_load_scheduler_state = config_checkpoint.only_load_scheduler_state + self.save_thread = None + self.verbose = config_checkpoint.verbose + self.keys_not_to_resume = config_checkpoint.keys_not_to_resume + self.broadcast_via_filesystem = config_checkpoint.broadcast_via_filesystem + # Create the object store client interface. + if config_checkpoint.load_from_object_store.enabled: + self.load_s3_backend_key = "_ckpt_s3_loader" + easy_io.set_s3_backend( + key="_ckpt_s3_loader", + backend_args={ + "backend": "s3", + "path_mapping": { + "s3://ckpt/": f"s3://{config_checkpoint.load_from_object_store.bucket}/", + }, + "s3_credential_path": config_checkpoint.load_from_object_store.credentials, + }, + ) + else: + self.load_s3_backend_key = None + + if config_checkpoint.save_to_object_store.enabled: + self.save_s3_backend_key = "_ckpt_s3_saver" + easy_io.set_s3_backend( + key="_ckpt_s3_saver", + backend_args={ + "backend": "s3", + "path_mapping": { + "s3://ckpt/": f"s3://{config_checkpoint.save_to_object_store.bucket}/", + }, + "s3_credential_path": config_checkpoint.save_to_object_store.credentials, + }, + ) + else: + self.save_s3_backend_key = None + + @abstractmethod + def save( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int, + ) -> None: + pass + + @abstractmethod + def load( + self, + model: ImaginaireModel, + optimizer: Optional[torch.optim.Optimizer] = None, + scheduler: Optional[torch.optim.lr_scheduler.LRScheduler] = None, + grad_scaler: Optional[torch.amp.GradScaler] = None, + ) -> int: + pass + + @property + def save_bucket(self): + """Get the bucket name for saving checkpoints.""" + return self.config_checkpoint.save_to_object_store.bucket if self.save_to_object_store else None + + @property + def load_bucket(self): + """Get the bucket name for loading checkpoints.""" + return self.config_checkpoint.load_from_object_store.bucket if self.load_from_object_store else None + + @property + def save_dirname(self): + return ( + f"s3://{self.save_bucket}/{self._object_store_dirname}" + if self.save_to_object_store + else self._local_dirname + ) + + @property + def load_dirname(self): + return ( + f"s3://{self.load_bucket}/{self._object_store_dirname}" + if self.load_from_object_store + else self._local_dirname + ) + + def finalize(self) -> None: + """Finalize the checkpointer.""" + if self.save_thread: + self.save_thread.join() + + def _read_latest_checkpoint_file(self) -> str | None: + """Get the file name of the latest saved checkpoint. If it doesn't exist, return None. + + Returns: + checkpoint_file (str | None): file name of the latest saved checkpoint. + """ + checkpoint_file = None + checkpoint_path = os.path.join(self.load_dirname, "latest_checkpoint.txt") + if easy_io.exists(f"{checkpoint_path}", backend_key=self.load_s3_backend_key): + checkpoint_file = easy_io.load(f"{checkpoint_path}", backend_key=self.load_s3_backend_key).strip() + + return checkpoint_file + + def _write_latest_checkpoint_file(self, checkpoint_file: str) -> None: + """Track the file name of the latest saved checkpoint. + + Args: + checkpoint_file (str): file name of the latest saved checkpoint. + """ + content = f"{checkpoint_file}\n" + checkpoint_path = os.path.join(self.save_dirname, "latest_checkpoint.txt") + easy_io.dump( + content, + checkpoint_path, + backend_key=self.save_s3_backend_key, + ) + + def _check_checkpoint_exists(self, checkpoint_path: str) -> None: + """If the file checkpoint_path does not exist, raise an error. + + Args: + checkpoint_path (str): full path to the checkpoint. + """ + if not easy_io.exists(f"{checkpoint_path}", backend_key=self.load_s3_backend_key): + raise FileNotFoundError(f"File not found (object store): {checkpoint_path}") diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/ddp.py b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/ddp.py new file mode 100644 index 0000000000000000000000000000000000000000..3bc63b3556a165041d58429e5156b1b7fb3c83be --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/ddp.py @@ -0,0 +1,445 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import threading +from collections import namedtuple +from typing import Any, Dict, Optional, Set, Tuple, Union + +import torch +import torch.distributed +from megatron.core import parallel_state +from torch.distributed import ProcessGroup, get_process_group_ranks + +from cosmos_policy._src.imaginaire.checkpointer.base import AbstractCheckpointer +from cosmos_policy._src.imaginaire.checkpointer.safe_broadcast import broadcast_object +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import distributed, log, misc +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + +StateDictItemPath = namedtuple("StateDictItemPath", ["state_dict", "save_path"]) + + +class Checkpointer(AbstractCheckpointer): + """ + Checkpointer for DDP. + Note: This implementation only supports local filesystem. + """ + + KEYS_TO_SAVE = ["model", "optim", "scheduler", "trainer"] + KEYS_TO_POSTFIX = { + "model": "model", + "optim": "optim", + "scheduler": "scheduler", + "trainer": "", + } + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + pp_world_size = parallel_state.get_pipeline_model_parallel_world_size() + ep_world_size = parallel_state.get_expert_model_parallel_world_size() + assert pp_world_size < 2, "Pipeline Parallelism (PP) is not tested yet." + assert ep_world_size < 2, "Expert Parallelism (EP) is not tested yet." + self.mp_world_size = parallel_state.get_model_parallel_group().size() + if self.mp_world_size > 1 and self.__class__ == Checkpointer: + raise NotImplementedError( + "Model Parallelism (MP) is enabled - you should use TensorParallel Checkpointer instead of DDP Checkpointer." + ) + # DDP rank (with context parallelism considered) + self.rank_dp_w_cp = parallel_state.get_data_parallel_rank(with_context_parallel=True) + # Context parallelism rank + self.cp_rank = parallel_state.get_context_parallel_rank() + # Model parallelism rank (including Tensor+Pipeline+Expert Parallelisms) + self.mp_rank = parallel_state.get_model_parallel_group().rank() + + # self.mp_rank = parallel_state.get_model_parallel_group(with_expert_parallel=ep_world_size > 1).rank() + if self.broadcast_via_filesystem: + log.info("Broadcasting checkpoint data via the local filesystem.") + if not self.strict_resume: + log.warning("Strict resume mode is off. Some model parameters may not be loaded.") + + # collect ranks of all model parallel groups + all_ranks = [None for _ in range(distributed.get_world_size())] + torch.distributed.all_gather_object( + all_ranks, get_process_group_ranks(parallel_state.get_model_parallel_group()) + ) + all_ranks = list(set(tuple(rank) if isinstance(rank, list) else rank for rank in all_ranks)) + for ranks in all_ranks: + group = torch.distributed.new_group(list(ranks), backend="gloo") + if distributed.get_rank() in ranks: + self.mp_gloo_pg = group + + self.print("Checkpointer Initialized.") + + def print(self, message: str): + """ + Print message to the console. Include the parallelism rank information when verbose is set to True. + """ + if self.verbose: + log.info( + f"[Parallelism Rank: DP-{self.rank_dp_w_cp}, TP-{self.mp_rank}, CP-{self.cp_rank}]: {message}", + rank0_only=False, + ) + else: + log.info(message, rank0_only=True) + + def add_type_postfix_to_checkpoint_path(self, key: str, checkpoint_path: str, model: ImaginaireModel) -> str: + del model + assert key in self.KEYS_TO_SAVE + post_fix = self.KEYS_TO_POSTFIX[key] + + if post_fix: + _ckpt_path = checkpoint_path.replace(".pt", f"_{post_fix}.pt") + else: + _ckpt_path = checkpoint_path + return _ckpt_path + + def save( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int, + ) -> None: + """Save network weights, optimizer parameters, scheduler parameters to a checkpoint. + + Args: + model (ImaginaireModel): The PyTorch model. + optimizer (torch.optim.Optimizer): The model optimizer. + scheduler (torch.optim.lr_scheduler.LRScheduler): The optimization scheduler. + grad_scaler (torch.amp.GradScaler): The gradient scaler (for mixed precision training). + iteration (int): Current iteration number. + """ + self.callbacks.on_save_checkpoint_start(model, iteration) + + checkpoint_file = self.format_checkpoint_filename(model, iteration) + state_dict = self.generate_save_state_dict(model, optimizer, scheduler, grad_scaler, iteration) + state_dict = self._map_state_dict_path_during_save(state_dict, checkpoint_file, model) + if state_dict: + # Wait for previous saver thread to end. + if self.save_thread: + self.save_thread.join() + # Run the checkpoint saver in a separate thread. + self.save_thread = threading.Thread( + target=self._save_worker, + daemon=False, + args=(state_dict, checkpoint_file, distributed.get_rank()), + ) + self.save_thread.start() + + # Note: Checkpoints are saved on a separate thread and this callback is not accurate. + # Please check logs from on_save_checkpoint_success() for better accuracy + self.callbacks.on_save_checkpoint_end(model=None, iteration=iteration) + + def _map_state_dict_path_during_save(self, state_dict, checkpoint_file, model) -> dict[str, StateDictItemPath]: + new_dict = {} + for key, _state_dict in state_dict.items(): + _ckpt_path = self.add_type_postfix_to_checkpoint_path(key, checkpoint_file, model) + checkpoint_path = os.path.join(self.save_dirname, _ckpt_path) + new_dict[key] = StateDictItemPath(_state_dict, checkpoint_path) + return new_dict + + @misc.timer("checkpoint saving") + def _save_worker(self, state_dict: dict[str, StateDictItemPath], checkpoint_file: str, rank: int = 0) -> None: + """Worker to upload checkpoint to object store, spawned with a child thread (in parallel with the training). + + Args: + state_dict (dict[str, StateDictItemPath]): The state dict of the model/optimizer/scheduler. + checkpoint_file (str): The file name of the model checkpoint. + rank (int): GPU device (default: 0). + """ + try: + for key, item in state_dict.items(): + self.print(f"Saving {key} to {item.save_path}") + try: + easy_io.dump( + item.state_dict, + item.save_path, + fast_backend=True, # optional for fast backend, cpu heavy + backend_key=self.save_s3_backend_key, + ) + self.print(f"Saved {key} to {item.save_path}") + except Exception as e: + self.print(f"Failed to save {key} to {item.save_path}: {str(e)}") + raise # Re-raise the exception after logging + + # Synchronize only rank 0 of each model parallel group + if self.mp_world_size > 1: + torch.distributed.barrier(group=self.mp_gloo_pg) + + # Only rank 0 of MP group and rank 0 of DP with CP updates latest_checkpoint.txt + if self.mp_rank == 0 and self.rank_dp_w_cp == 0: + self._write_latest_checkpoint_file(checkpoint_file) + + if distributed.get_rank() == 0: # only rank 0 saves trained_data_record + if "trained_data_record" in state_dict["model"].state_dict: + self._write_trained_data_record( + checkpoint_file, state_dict["model"].state_dict["trained_data_record"] + ) + + iteration = int(checkpoint_file.replace("iter_", "").replace(".pt", "")) + self.callbacks.on_save_checkpoint_success(iteration=iteration) + except Exception as e: # noqa: BLE001 + log.exception(f"Checkpoint failed to upload: {e}", rank0_only=not self.verbose) + + def format_checkpoint_filename(self, model: ImaginaireModel, iteration: int) -> str: + """Generate the checkpoint file name. + + Args: + iteration (int): The current iteration number. + + Returns: + checkpoint_file (str): The checkpoint file name. + """ + del self, model + return f"iter_{iteration:09}.pt" + + @misc.timer("generate saving state dict") + def generate_save_state_dict( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int, + ) -> Optional[Dict[str, Any]]: + state_dict = {} + + if self.rank_dp_w_cp == 0: + trainer_state = dict( + grad_scaler=grad_scaler.state_dict(), + iteration=iteration, + ) + model_state = model.state_dict() + optim_state = optimizer.state_dict() + scheduler_state = scheduler.state_dict() + self.callbacks.on_save_checkpoint(model, state_dict=trainer_state) + + trainer_state, model_state, optim_state, scheduler_state = misc.to( + [trainer_state, model_state, optim_state, scheduler_state], device="cpu" + ) + + state_dict = { + "model": model_state, + "optim": optim_state, + "scheduler": scheduler_state, + } + if distributed.get_rank() == 0: # only rank 0 saves trainer state + state_dict["trainer"] = trainer_state + return state_dict + return state_dict + + def load_broadcast_state_dict( + self, checkpoint_path: str, model: ImaginaireModel, resume_keys: Set + ) -> dict[str, Any]: + """ + Load state_dict and broadcast. + + The main steps are: + 1. Download TP-rank-specific checkpoints for every GPU of DDP-rank 0 and CP-rank 0. + 2. Each rank loads its corresponding checkpoint from the local cache or receives it via broadcast. + + This approach ensures that each MP rank loads its specific part of the model, which is + crucial for Model Parallelism where different parts of the model are distributed across + multiple GPUs. + + When using Model Parallelism (e.g., Tensor Parallelism), the `broadcast_via_filesystem` option can + be set to True. This allows each rank to load its specific checkpoint from the local filesystem + instead of receiving it via network broadcast, which could be more efficient in some cases. + + For standard DDP without TP, `broadcast_via_filesystem` should remain False (default). + + Args: + checkpoint_path (str): The base path of the checkpoint. + model (ImaginaireModel): The model being loaded. + resume_keys (Set): Set of keys to resume from the checkpoint. + + Returns: + dict[str, Any]: A dictionary containing the loaded state for each resumed key. + """ + state_dict = {} + sorted_resume_keys = sorted(resume_keys) + # Step 1: Download TP-rank-specific checkpoints for every GPU of DDP-rank 0 and CP-rank 0. + if self.rank_dp_w_cp == 0: + for key in sorted_resume_keys: + _ckpt_path = self.add_type_postfix_to_checkpoint_path(key, checkpoint_path, model) + local_cache_path = os.path.join(self.load_dirname, os.path.basename(_ckpt_path)) + if os.path.exists(local_cache_path): + # If the local checkpoint exists, we can directly load it + self.print(f"Checkpoint is already in local cache: {local_cache_path}. Loading...") + _state_dict = easy_io.load(local_cache_path, fast_backend=True) + else: + _state_dict = easy_io.load(_ckpt_path, fast_backend=True, backend_key=self.load_s3_backend_key) + self.print(f"Downloading checkpoint from: {_ckpt_path}") + if self.broadcast_via_filesystem: + # Save the checkpoint to the local filesystem + easy_io.dump(_state_dict, local_cache_path, fast_backend=True) + state_dict[key] = _state_dict + # Ensure all ranks wait for the download to complete + distributed.barrier() + + # Step 2: Broadcast checkpoint data + log.info( + "Start broadcasting checkpoint from the source rank to all other ranks in the same DDP group.", + rank0_only=True, + ) + for key in sorted_resume_keys: + if self.broadcast_via_filesystem: + # Load the checkpoint from the local filesystem for other ranks + if self.rank_dp_w_cp != 0: + _ckpt_path = self.add_type_postfix_to_checkpoint_path(key, checkpoint_path, model) + local_cache_path = os.path.join(self.load_dirname, os.path.basename(_ckpt_path)) + self.print(f"Loading checkpoint from: {local_cache_path}") + state_dict[key] = easy_io.load(local_cache_path, fast_backend=True) + else: + # Broadcast the checkpoint to all GPUs of the current DDP rank + group: ProcessGroup = parallel_state.get_data_parallel_group(with_context_parallel=True) + min_rank = min(get_process_group_ranks(group)) + + _state_dict = broadcast_object( + state_dict[key] if self.rank_dp_w_cp == 0 else None, + min_rank, + group=group, + device=torch.device(torch.cuda.current_device()), + ) + if self.rank_dp_w_cp == 0: + self.print(f'Broadcasted checkpoint["{key}"] to all other ranks in the same DDP group.') + else: + state_dict[key] = _state_dict + self.print(f'Received checkpoint["{key}"] from source rank {min_rank}.') + + return state_dict + + def keys_to_resume_during_load(self) -> Tuple[Set, Union[str, None]]: + latest_checkpoint_file = self._read_latest_checkpoint_file() + + resume_keys = [] + + if latest_checkpoint_file is not None: + # 1. Resume training from latest_checkpoint.txt under the same name. + checkpoint_path = os.path.join(self.load_dirname, latest_checkpoint_file) + resume_keys.extend(self.KEYS_TO_SAVE) + else: + if self.load_path: + # 2. Load the module weights specified by config_checkpoint.path. + checkpoint_path = self.load_path + if self.load_s3_backend_key: + checkpoint_path = f"s3://ckpt/{checkpoint_path}" + if self.load_training_state: + resume_keys.extend(self.KEYS_TO_SAVE) + else: + resume_keys.append("model") + if self.only_load_scheduler_state: + resume_keys.append("scheduler") + else: + checkpoint_path = None + if len(self.keys_not_to_resume) > 0: + for key in self.keys_not_to_resume: + assert key in self.KEYS_TO_SAVE, f"Invalid key to resume: {key} not in {self.KEYS_TO_SAVE}" + resume_keys = [key for key in resume_keys if key not in self.keys_not_to_resume] + return set(resume_keys), checkpoint_path + + @misc.timer("checkpoint loading") + def load( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer | None = None, + scheduler: torch.optim.lr_scheduler.LRScheduler | None = None, + grad_scaler: torch.amp.GradScaler | None = None, + ) -> int: + """Load network weights and optimizer states from a checkpoint in a single process. + + The priority of the checkpoint loading logic is: + 1. Attempt to resume training if possible by looking for latest_checkpoint.txt under the same name. + 2. If no latest checkpoint were found, it loads the model weights specified by config_checkpoint.path. + - This is typically used for inference mode. + - If config_checkpoint.load_optimizer_state is True, then also load the optimizer and scheduler states. + 3. If none of the above, randomly initialize the model parameters and train from scratch. + + Args: + model (ImaginaireModel): The PyTorch model. + optimizer (torch.optim.Optimizer | None): The model optimizer (default: None). + scheduler (torch.optim.lr_scheduler.LRScheduler | None): The optimization scheduler (default: None). + grad_scaler (torch.amp.GradScaler | None): The gradient scaler (for mixed precision training). + + Returns: + iteration (int): the iteration number to start/resume from. + """ + self.callbacks.on_load_checkpoint_start(model) + + resume_keys, checkpoint_path = self.keys_to_resume_during_load() + + iteration = 0 + + # Load checkpoint. + if checkpoint_path is not None: + self._check_checkpoint_exists(checkpoint_path) + state_dict = self.load_broadcast_state_dict(checkpoint_path, model, set(resume_keys)) + + if "trainer" in state_dict: + trainer_state = state_dict["trainer"] + log.critical(state_dict.keys(), rank0_only=False) + log.critical(trainer_state, rank0_only=False) + log.info("- Loading the gradient scaler...") + grad_scaler.load_state_dict(trainer_state["grad_scaler"]) + self.callbacks.on_load_checkpoint(model, state_dict=trainer_state) + iteration = trainer_state["iteration"] + if "optim" in state_dict: + assert optimizer + optimizer_state = state_dict["optim"] + log.info("- Loading the optimizer...") + optimizer.load_state_dict(optimizer_state) + if "scheduler" in state_dict: + assert scheduler + scheduler_state = state_dict["scheduler"] + log.info("- Loading the scheduler...") + scheduler.load_state_dict(scheduler_state) + scheduler.last_epoch = iteration + if "model" in state_dict: + model_state = state_dict["model"] + log.info("- Loading the model...") + # model.load_state_dict(model_state) + if self.strict_resume: + log.info("\t Strict resume mode is on.") + else: + log.info("\t Strict resume mode is off.") + model_load_info = model.load_state_dict(model_state, strict=self.strict_resume) + log.info(f"\t {model_load_info}") + self.print(f"Loaded checkpoint from {checkpoint_path} in iteration {iteration}") + else: + log.info("Training from scratch.") + torch.cuda.empty_cache() + + self.callbacks.on_load_checkpoint_end(model, iteration=iteration, checkpoint_path=checkpoint_path) + + return iteration + + def _write_trained_data_record(self, checkpoint_file: str, trained_data_record: dict[str, int]) -> None: + """Write json file to save number of seen samples and number of iterations. + + Args: + checkpoint_file (str): iteration number for the saved checkpoint + trained_data_record (dict[str, int]): example {"image": 0, "video": 0, "iteration": 0}. + """ + # filename: iter_xxxxxxxxx_trained_data_record.json + checkpoint_path = os.path.join( + self.save_dirname, f"{checkpoint_file.replace('.pt', '')}_trained_data_record.json" + ) + easy_io.dump( + trained_data_record, + checkpoint_path, + backend_key=self.save_s3_backend_key, + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/ddp_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/ddp_test.py new file mode 100644 index 0000000000000000000000000000000000000000..700e52c380029007687ef0a7e0acb0eaf856dc95 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/ddp_test.py @@ -0,0 +1,198 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +torchrun --nproc_per_node=2 -m pytest cosmos_policy/_src/imaginaire/checkpointer/ddp_test.py +""" + +import os +import shutil + +import pytest +import torch +import torch.distributed as dist +import torch.nn as nn +import torch.optim as optim +from megatron.core import parallel_state + +from cosmos_policy._src.imaginaire.checkpointer.ddp import Checkpointer +from cosmos_policy._src.imaginaire.config import CheckpointConfig, Config, JobConfig, ObjectStoreConfig +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import instantiate +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer +from cosmos_policy._src.imaginaire.utils import distributed, log + + +class SimpleModel(nn.Module): + def __init__(self): + super().__init__() + self.fc = nn.Linear(10, 5) + + def forward(self, x): + return self.fc(x) + + +class SimpleImaginaireModel(ImaginaireModel): + def __init__(self): + super().__init__() + self.model = SimpleModel() + + def forward(self, x): + return self.model(x) + + +@pytest.fixture(scope="module") +def setup_trainer_config(): + config = Config( + model=L(SimpleImaginaireModel)(), + optimizer=L(torch.optim.Adam)(params=None, lr=0.001), + scheduler=L(torch.optim.lr_scheduler.StepLR)(optimizer=None, step_size=1, gamma=0.1), + dataloader_train=None, + dataloader_val=None, + ) + # construct the trainer, which will also construct distributed init process group + trainer = ImaginaireTrainer(config) + model = instantiate(config.model).cuda() + optimizer, scheduler = model.init_optimizer_scheduler(config.optimizer, config.scheduler) + grad_scaler = torch.amp.GradScaler("cuda", **config.trainer.grad_scaler_args) + yield trainer, model, optimizer, scheduler, grad_scaler, config + # Cleanup + dist.destroy_process_group() + parallel_state.destroy_model_parallel() + + +def setup_checkpointer(trainer, object_store_enabled, job_name): + _object_store = ObjectStoreConfig( + enabled=object_store_enabled, + bucket="checkpoints" if object_store_enabled else "test-bucket", + credentials="credentials/pbss_dir.secret" if object_store_enabled else "test-credentials", + ) + config_checkpoint = CheckpointConfig( + save_to_object_store=_object_store, + load_from_object_store=_object_store, + strict_resume=True, + load_path=None, + load_training_state=True, + only_load_scheduler_state=False, + ) + config_job = JobConfig( + project="imaginaire4", + group="test_checkpointer", + name=job_name, + ) + ckpt = Checkpointer(config_checkpoint, config_job, trainer.callbacks) + return ckpt + + +def train_and_save(model, optimizer, scheduler, grad_scaler, checkpointer): + x = torch.randn(32, 10).cuda() + y = torch.randn(32, 5).cuda() + criterion = nn.MSELoss() + + for _ in range(2): + optimizer.zero_grad() + output = model(x) + loss = criterion(output, y) + loss.backward() + grad_scaler.step(optimizer) + grad_scaler.update() + scheduler.step() + scheduler.step() + checkpointer.save(model, optimizer, scheduler, grad_scaler, iteration=10) + + +def load_and_compare( + model, + optimizer, + scheduler, + grad_scaler, + checkpointer: Checkpointer, + original_model, + original_optimizer, + original_scheduler, +): + iteration = checkpointer.load(model, optimizer, scheduler, grad_scaler) + + dist.barrier() + if iteration != 10: + log.critical(f"Iteration number does not match after loading checkpoint: {iteration}", rank0_only=False) + assert iteration == 10, "Iteration number does not match after loading checkpoint" + + # compare model parameters + for param1, param2 in zip(original_model.parameters(), model.parameters()): + assert torch.allclose(param1, param2), f"Model parameters differ after loading checkpoint {param1} {param2}" + distributed.barrier() + log.success("Model parameters match after loading checkpoint", rank0_only=False) + + # compare optimizer states + for param1, param2 in zip( + original_optimizer.state_dict()["state"].values(), optimizer.state_dict()["state"].values() + ): + for k in param1: + assert torch.allclose(param1[k], param2[k]), f"Optimizer state {k} differs after loading checkpoint" + + log.success("Optimizer states match after loading checkpoint", rank0_only=False) + + # compare scheduler states + orginal_lr = original_scheduler.get_last_lr() + loaded_lr = scheduler.get_last_lr() + if orginal_lr != loaded_lr: + log.critical(f"Learning rate differs after loading checkpoint: {orginal_lr} {loaded_lr}", rank0_only=False) + assert orginal_lr == loaded_lr, "Learning rate differs after loading checkpoint" + log.success("Learning rate matches after loading checkpoint", rank0_only=False) + + +@pytest.mark.skip(reason="Tests are available in the test environment, run manually") +def test_checkpointer_local(setup_trainer_config): + trainer, model, optimizer, scheduler, grad_scaler, config = setup_trainer_config + checkpointer = setup_checkpointer(trainer, object_store_enabled=False, job_name="local_ddp") + if distributed.get_rank() == 0: + if os.path.exists(checkpointer.checkpoint_dir_local): + shutil.rmtree(checkpointer.checkpoint_dir_local) + + ddp_model = distributed.DistributedDataParallel(model) + train_and_save(ddp_model, optimizer, scheduler, grad_scaler, checkpointer) + checkpointer.finalize() + distributed.barrier() + checkpointer = setup_checkpointer(trainer, object_store_enabled=False, job_name="local_ddp") + + model2 = SimpleImaginaireModel().cuda() + optimizer2 = optim.Adam(model2.parameters()) + scheduler2 = optim.lr_scheduler.StepLR(optimizer2, step_size=1, gamma=0.1) + grad_scaler2 = torch.amp.GradScaler("cuda", **config.trainer.grad_scaler_args) + ddp_model2 = distributed.DistributedDataParallel(model2) + + load_and_compare(ddp_model2, optimizer2, scheduler2, grad_scaler2, checkpointer, ddp_model, optimizer, scheduler) + + +@pytest.mark.skip(reason="Tests are available in the test environment, run manually") +def test_checkpointer_remote(setup_trainer_config): + trainer, model, optimizer, scheduler, grad_scaler, config = setup_trainer_config + checkpointer = setup_checkpointer(trainer, object_store_enabled=True, job_name="remote_ddp") + + ddp_model = distributed.DistributedDataParallel(model) + train_and_save(ddp_model, optimizer, scheduler, grad_scaler, checkpointer) + checkpointer.finalize() + distributed.barrier() + checkpointer = setup_checkpointer(trainer, object_store_enabled=True, job_name="remote_ddp") + + model2 = SimpleImaginaireModel().cuda() + optimizer2 = optim.Adam(model2.parameters()) + scheduler2 = optim.lr_scheduler.StepLR(optimizer2, step_size=1, gamma=0.1) + grad_scaler2 = torch.amp.GradScaler(**config.trainer.grad_scaler_args) + ddp_model2 = distributed.DistributedDataParallel(model2) + + load_and_compare(ddp_model2, optimizer2, scheduler2, grad_scaler2, checkpointer, ddp_model, optimizer, scheduler) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/dummy.py b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/dummy.py new file mode 100644 index 0000000000000000000000000000000000000000..43bf89f94e8109857fa7fa43b283baedcc0ee654 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/dummy.py @@ -0,0 +1,47 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Optional + +import torch +import torch.distributed + +from cosmos_policy._src.imaginaire.checkpointer.base import AbstractCheckpointer +from cosmos_policy._src.imaginaire.model import ImaginaireModel + + +class Checkpointer(AbstractCheckpointer): + """ + A dummy checkpointer that does not save or load anything. This is useful for debugging jobs or share workload with collobrators. + """ + + def save( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int, + ) -> None: + pass + + def load( + self, + model: ImaginaireModel, + optimizer: Optional[torch.optim.Optimizer] = None, + scheduler: Optional[torch.optim.lr_scheduler.LRScheduler] = None, + grad_scaler: Optional[torch.amp.GradScaler] = None, + ) -> int: + return 0 diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/s3_filesystem.py b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/s3_filesystem.py new file mode 100644 index 0000000000000000000000000000000000000000..cf79b6b81567d5b432fe91efacd23e0d27eba1fa --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/s3_filesystem.py @@ -0,0 +1,330 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import io +import os +import time +from contextlib import contextmanager +from typing import Generator, Union +from urllib.parse import urlparse + +from botocore.exceptions import ClientError +from torch.distributed.checkpoint import FileSystemReader, FileSystemWriter +from torch.distributed.checkpoint.filesystem import FileSystemBase + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + + +class S3Stream(io.BytesIO): + """ + Workaround for PyTorch manually closing the stream before we can upload it to S3. We override the close() as noop + and instead call our own _true_close() method to close the stream after we are done using it. + The commit at fault is https://github.com/pytorch/pytorch/commit/9c909bf3bb122db2cce95e2eb7459bbe50dfa15a + """ + + def close(self): + self.flush() + # No close + + def _true_close(self): + super().close() + + +class S3FileSystem(FileSystemBase): + """Implementation of FileSystemBase for AWS S3 storage.""" + + def __init__( + self, + credential_path: str, + max_attempts: int = 20, + initial_backoff: float = 1.0, + max_backoff: float = 30.0, + backoff_factor: float = 2.0, + enable_gcs_patch_in_boto3: bool = False, + ) -> None: + """ + Initialize S3FileSystem with retry configuration. + + Args: + credential_path: Path to AWS credentials JSON file + max_attempts: Maximum number of retry attempts + initial_backoff: Initial backoff time in seconds + max_backoff: Maximum backoff time in seconds + backoff_factor: Multiplicative factor for backoff time + enable_gcs_patch_in_boto3: Whether to enable GCS patch in boto3 + """ + self.easy_io_backend = easy_io.get_file_backend( + backend_args={ + "backend": "s3", + "s3_credential_path": credential_path, + "path_mapping": None, + } + ) + self.max_attempts = max_attempts + self.initial_backoff = initial_backoff + self.max_backoff = max_backoff + self.backoff_factor = backoff_factor + self.enable_gcs_patch_in_boto3 = enable_gcs_patch_in_boto3 + if enable_gcs_patch_in_boto3: + log.info("enable_gcs_patch_in_boto3: True") + + def _retry_with_backoff(self, operation_func, *args, **kwargs): + """ + Execute an operation with exponential backoff retry logic. + + Args: + operation_func: Function to execute + *args: Positional arguments for the function + **kwargs: Keyword arguments for the function + + Returns: + Result of the operation function + + Raises: + Exception: If all retry attempts fail + """ + last_exception = None + backoff = self.initial_backoff + + for attempt in range(self.max_attempts): + try: + return operation_func(*args, **kwargs) + except ClientError as e: + error_code = e.response.get("Error", {}).get("Code", "") + log.info(f"S3 Filesystem: Received ClientError: {error_code}", rank0_only=False) + + # Handle specific error cases + if error_code in ["SlowDown", "ThrottlingException", "RequestLimitExceeded", "InternalError"]: + last_exception = e + if attempt < self.max_attempts - 1: # Don't sleep on last attempt + current_backoff = min(backoff, self.max_backoff) + log.info(f"S3 Filesystem: Retrying in {current_backoff} seconds", rank0_only=False) + time.sleep(current_backoff) + backoff *= self.backoff_factor + continue + # For other client errors, raise immediately + raise + except Exception as e: + log.info(f"S3 Filesystem: Received Exception: {str(e)}", rank0_only=False) + last_exception = e + if attempt < self.max_attempts - 1: + current_backoff = min(backoff, self.max_backoff) + log.info(f"S3 Filesystem: Retrying in {current_backoff} seconds", rank0_only=False) + time.sleep(current_backoff) + backoff *= self.backoff_factor + continue + + # pyrefly: ignore [bad-raise] + raise last_exception + + @contextmanager + def create_stream(self, path: Union[str, os.PathLike], mode: str) -> Generator[io.IOBase, None, None]: + """Create a stream for reading from or writing to S3 with retry logic.""" + path_str = str(path) + bucket, key = self._parse_s3_uri(path_str) + log.info(f"S3 Filesystem: Creating stream for {key} in bucket {bucket}", rank0_only=False) + + if mode == "rb": + stream = io.BytesIO() + try: + + def download_operation(): + stream.write(self.easy_io_backend.get(filepath=path_str)) + stream.seek(0) + + log.info(f"S3 Filesystem: Downloading {key} from bucket {bucket}", rank0_only=False) + self._retry_with_backoff(download_operation) + log.info("S3 Filesystem: Download complete", rank0_only=False) + yield stream + finally: + stream.close() + elif mode == "wb": + stream = S3Stream() + try: + yield stream + + def upload_operation(): + stream.seek(0) + self.easy_io_backend.put(obj=stream, filepath=path_str) + + log.info(f"S3 Filesystem: Uploading {key} to bucket {bucket}", rank0_only=False) + self._retry_with_backoff(upload_operation) + log.info("S3 Filesystem: Upload complete", rank0_only=False) + finally: + stream._true_close() + else: + raise ValueError(f"Unsupported mode: {mode}") + + def concat_path(self, path: Union[str, os.PathLike], suffix: str) -> Union[str, os.PathLike]: + """Concatenate S3 path with suffix.""" + path_str = str(path) + if path_str.endswith("/"): + return f"{path_str}{suffix}" + return f"{path_str}/{suffix}" + + def init_path(self, path: Union[str, os.PathLike]) -> Union[str, os.PathLike]: + """Initialize and validate S3 path.""" + path_str = str(path) + if not path_str.startswith("s3://"): + raise ValueError(f"Invalid S3 URI: {path_str}. Must start with 's3://'") + return path_str + + def rename(self, path: Union[str, os.PathLike], new_path: Union[str, os.PathLike]) -> None: + """Rename (move) an object in S3 with retry logic.""" + src_path = str(path) + dst_path = str(new_path) + + def copy_operation(): + self.easy_io_backend.copyfile(src=src_path, dst=dst_path) + + self._retry_with_backoff(copy_operation) + + def delete_operation(): + self.easy_io_backend.remove(filepath=src_path) + + self._retry_with_backoff(delete_operation) + + def mkdir(self, path: Union[str, os.PathLike]) -> None: + """ + Create a "directory" in S3. + + Note: S3 doesn't have real directories, but we can create an empty object + with a trailing slash to simulate a directory. + """ + # Creating same buckets from different ranks can cause rate limit issues in GCP. + # In object store, we don't need to create a directory. + pass + + def ls(self, path: Union[str, os.PathLike]) -> list[str]: + """List objects under the given S3 path (prefix) and return s3:// URIs.""" + path_str = str(path) + return [ + f"{path_str.removesuffix('/')}/{obj_suffix}" + for obj_suffix in self.easy_io_backend.list_dir_or_file(dir_path=path_str, list_dir=False, list_file=True) + ] + + @classmethod + def validate_checkpoint_id(cls, checkpoint_id: Union[str, os.PathLike]) -> bool: + """Validate if the checkpoint_id is a valid S3 URI.""" + checkpoint_id_str = str(checkpoint_id) + try: + if not checkpoint_id_str.startswith("s3://"): + return False + parsed = urlparse(checkpoint_id_str) + return bool(parsed.netloc and parsed.path) # Must have bucket and key + except Exception: + return False + + def exists(self, path: Union[str, os.PathLike]) -> bool: + """Check if an object exists in S3 with retry logic.""" + try: + + def head_operation() -> bool: + return self.easy_io_backend.exists(filepath=str(path)) + + return self._retry_with_backoff(head_operation) + except ClientError as e: + if e.response.get("Error", {}).get("Code", "") == "404": + return False + raise + + def rm_file(self, path: Union[str, os.PathLike]) -> None: + """Remove a file from S3 with retry logic.""" + + def delete_operation(): + self.easy_io_backend.remove(filepath=str(path)) + + self._retry_with_backoff(delete_operation) + + def _parse_s3_uri(self, uri: str) -> tuple[str, str]: + """ + Parse an S3 URI into bucket and key. + + Args: + uri: S3 URI in the format s3://bucket-name/key + + Returns: + Tuple of (bucket_name, key) + + Raises: + ValueError: If the URI is invalid + """ + uri = uri if isinstance(uri, str) else str(uri) + if not uri.startswith("s3://"): + raise ValueError(f"Invalid S3 URI: {uri}. Must start with 's3://'") + + parsed = urlparse(uri) + bucket = parsed.netloc + + # Remove leading slash from key + key = parsed.path.lstrip("/") + + if not bucket: + raise ValueError(f"Invalid S3 URI: {uri}. No bucket specified") + + return bucket, key + + +class S3StorageWriter(FileSystemWriter): + def __init__( + self, + credential_path: str, + path: str, + enable_gcs_patch_in_boto3: bool = False, + **kwargs, + ) -> None: + """ + Initialize an S3 writer for distributed checkpointing. + + Args: + region (str): The AWS region for S3. + path (str): The S3 URI to write checkpoints to. + kwargs (dict): Keyword arguments to pass to the parent :class:`FileSystemWriter`. + enable_gcs_patch_in_boto3 (bool): Whether to enable GCS patch in boto3 + """ + super().__init__( + path=path, + sync_files=False, + **kwargs, + ) + self.fs = S3FileSystem(credential_path, enable_gcs_patch_in_boto3=enable_gcs_patch_in_boto3) # type: ignore + self.path = self.fs.init_path(path) + + @classmethod + def validate_checkpoint_id(cls, checkpoint_id: Union[str, os.PathLike]) -> bool: + return S3FileSystem.validate_checkpoint_id(checkpoint_id) + + +class S3StorageReader(FileSystemReader): + def __init__( + self, credential_path: str, path: Union[str, os.PathLike], enable_gcs_patch_in_boto3: bool = False + ) -> None: + """ + Initialize an S3 reader for distributed checkpointing. + + Args: + region (str): The AWS region for S3. + path (Union[str, os.PathLike]): The S3 path to read checkpoints from. + enable_gcs_patch_in_boto3 (bool): Whether to enable GCS patch in boto3 + """ + super().__init__(path) + self.fs = S3FileSystem(credential_path, enable_gcs_patch_in_boto3=enable_gcs_patch_in_boto3) # type: ignore + self.path = self.fs.init_path(path) + self.sync_files = False + + @classmethod + def validate_checkpoint_id(cls, checkpoint_id: Union[str, os.PathLike]) -> bool: + return S3FileSystem.validate_checkpoint_id(checkpoint_id) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/safe_broadcast.py b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/safe_broadcast.py new file mode 100644 index 0000000000000000000000000000000000000000..0fa6f483be34be52c5fdf4919826491861a25a3b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/safe_broadcast.py @@ -0,0 +1,96 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import collections +import io +import pickle +from typing import Any + +import torch +import torch.distributed as dist + +# https://github.com/pytorch/pytorch/blob/main/torch/distributed/optim/zero_redundancy_optimizer.py#L29 + + +def broadcast_object( + obj: Any, + src_rank: int, + group: object = dist.group.WORLD, + device: torch.device = torch.device("cpu"), +) -> Any: + r""" + Broadcasts an object to the given group. + + It will be sending the object if called from the source rank and receiving + the object otherwise. + + Arguments: + obj: object to broadcast; only used if called on the source rank. + src_rank (int): source rank. + group (``ProcessGroup``, optional): group used for the broadcast + (default: ``dist.group.WORLD``). + device (``torch.device``, optional): device to send from or receive + to (default: ``torch.device("cpu")``). + + Returns: + The broadcasted object. + """ + if dist.get_rank() == src_rank: + # Send the object + buffer = io.BytesIO() + torch.save(obj, buffer, pickle_protocol=pickle.HIGHEST_PROTOCOL) + data = bytearray(buffer.getbuffer()) + length_tensor = torch.LongTensor([len(data)]).to(device) + data_send_tensor = torch.ByteTensor(data).to(device) + dist.broadcast(length_tensor, src=src_rank, group=group, async_op=False) + dist.broadcast(data_send_tensor, src=src_rank, group=group, async_op=False) + else: + # Receive the object + length_tensor = torch.LongTensor([0]).to(device) + dist.broadcast(length_tensor, src=src_rank, group=group, async_op=False) + data_recv_tensor = torch.empty([int(length_tensor.item())], dtype=torch.uint8, device=device) + dist.broadcast(data_recv_tensor, src=src_rank, group=group, async_op=False) + buffer = io.BytesIO(data_recv_tensor.cpu().numpy()) + obj = torch.load(buffer, map_location=device, weights_only=False) + return obj + + +def _recursive_copy_to_device( + value: Any, + non_blocking: bool, + device: torch.device, +) -> Any: + r""" + Recursively searches lists, tuples, dicts and copies tensors to device if possible. + + Non-tensor values are passed as-is in the result. + + .. note: These are all copies, so if there are two objects that reference + the same object, then after this call, there will be two different objects + referenced on the device. + """ + if isinstance(value, torch.Tensor): + return value.to(device, non_blocking=non_blocking) + + if isinstance(value, (list, tuple)): + values = [_recursive_copy_to_device(val, non_blocking=non_blocking, device=device) for val in value] + return values if isinstance(value, list) else tuple(values) + + if isinstance(value, collections.abc.Mapping): + return { + key: _recursive_copy_to_device(val, non_blocking=non_blocking, device=device) for key, val in value.items() + } + + return value diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/tp.py b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/tp.py new file mode 100644 index 0000000000000000000000000000000000000000..26f50c516f2096a71fd19c79470e0e13e9fe8169 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/tp.py @@ -0,0 +1,42 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from cosmos_policy._src.imaginaire.checkpointer.ddp import Checkpointer as DDPCheckpointer +from cosmos_policy._src.imaginaire.model import ImaginaireModel + + +class Checkpointer(DDPCheckpointer): + """ + Checkpointer class for Tensor Parallelism (TP) in distributed training. + + This implementation supports the combination of Tensor Parallelism (TP) and Data Parallel Processing (DDP), with optional Context Parallelism (CP). + + Note: + - Fully Sharded Data Parallelism (FSDP) is not supported by this checkpointer. + - In principle, this implementation is also compatible with Pipeline Parallelism (PP) and Expert Parallelism (EP), which are other forms of model parallelism. However, PP and EP have not been tested yet. + """ + + def add_type_postfix_to_checkpoint_path(self, key: str, checkpoint_path: str, model: ImaginaireModel) -> str: + """ + Overwrite the `add_type_postfix_to_checkpoint_path` function of the base class (DDP checkpointer) + to append the TP-rank postfix to the checkpoint path. + """ + checkpoint_path = super().add_type_postfix_to_checkpoint_path(key, checkpoint_path, model) + if key == "trainer": + return checkpoint_path + else: + checkpoint_path = checkpoint_path.replace(".pt", f"_mp_{self.mp_rank}.pt") + + return checkpoint_path diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/tp_ema.py b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/tp_ema.py new file mode 100644 index 0000000000000000000000000000000000000000..6aea81eec01c9ad6959f27fec3f32f32f42333d8 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/checkpointer/tp_ema.py @@ -0,0 +1,93 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any, Dict, Optional + +import torch +from megatron.core import parallel_state + +from cosmos_policy._src.imaginaire.checkpointer.tp import Checkpointer as BaseCheckpointer +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import misc + + +class Checkpointer(BaseCheckpointer): + KEYS_TO_SAVE = ["model", "optim", "trainer", "scheduler", "ema"] + KEYS_TO_POSTFIX = { + "model": "model", + "optim": "optim", + "ema": "ema", + "scheduler": "scheduler", + "trainer": "", + } + + @misc.timer("generate saving state dict") + def generate_save_state_dict( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int, + ) -> Optional[Dict[str, Any]]: + state_dict = {} + if parallel_state.get_data_parallel_rank() == 0: + trainer_state = dict( + grad_scaler=grad_scaler.state_dict(), + iteration=iteration, + ) + model_state = model.state_dict() + optim_state = optimizer.state_dict() + scheduler_state = scheduler.state_dict() + self.callbacks.on_save_checkpoint(model, state_dict=trainer_state) + + trainer_state, model_state, optim_state, scheduler_state = misc.to( + [trainer_state, model_state, optim_state, scheduler_state], device="cpu" + ) + + state_dict = { + "trainer": trainer_state, + "model": model_state, + "optim": optim_state, + "scheduler": scheduler_state, + } + + if parallel_state.get_data_parallel_rank() < 3: + ema_state = model.ema.state_dict() + state_dict["ema"] = ema_state + + return state_dict + + def add_type_postfix_to_checkpoint_path(self, key: str, checkpoint_path: str, model: ImaginaireModel) -> str: + # we need to get which ema should be saved + assert key in self.KEYS_TO_SAVE + post_fix = self.KEYS_TO_POSTFIX[key] + + if post_fix: + checkpoint_path = checkpoint_path.replace(".pt", f"_{post_fix}.pt") + else: + checkpoint_path = checkpoint_path + + if key == "ema": + dp_rank = parallel_state.get_data_parallel_rank() + checkpoint_path = checkpoint_path.replace(".pt", f"{dp_rank}.pt") + + if key == "trainer": + return checkpoint_path + else: + mp_rank = parallel_state.get_model_parallel_group().rank() + checkpoint_path = checkpoint_path.replace(".pt", f"_mp_{mp_rank}.pt") + + return checkpoint_path diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/config.py b/REGEN-main/cosmos_policy/_src/imaginaire/config.py new file mode 100644 index 0000000000000000000000000000000000000000..54c8239060fbe9910a4d7fecac3f085d5e10769c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/config.py @@ -0,0 +1,517 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""Training config system for Imaginare4""" + +from __future__ import annotations + +import importlib +import os +import time +from typing import Any, Dict, Optional, Type, TypeVar, Union + +import attrs +import torch +import torch.utils.data +import torch.utils.data.distributed +from loguru import logger as logging + +try: + from megatron.core import ModelParallelConfig + + USE_MEGATRON = True +except ImportError: + USE_MEGATRON = False + print("Megatron-core is not installed.") + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.serialization import from_yaml, load_callable +from cosmos_policy._src.imaginaire.utils import callback, distributed +from cosmos_policy._src.imaginaire.utils.misc import Color + +T = TypeVar("T") + + +def _is_attrs_instance(obj: object) -> bool: + """ + Helper function to check if an object is an instance of an attrs-defined class. + + Args: + obj: The object to check. + + Returns: + bool: True if the object is an instance of an attrs-defined class, False otherwise. + """ + return hasattr(obj, "__attrs_attrs__") + + +def make_freezable(cls: T) -> T: + """ + A decorator that adds the capability to freeze instances of an attrs-defined class. + + NOTE: This requires the wrapped attrs to be defined with attrs.define(slots=False) because we need + to hack on a "_is_frozen" attribute. + + This decorator enhances an attrs-defined class with the ability to be "frozen" at runtime. + Once an instance is frozen, its attributes cannot be changed. It also recursively freezes + any attrs-defined objects that are attributes of the class. + + Usage: + @make_freezable + @attrs.define(slots=False) + class MyClass: + attribute1: int + attribute2: str + + obj = MyClass(1, 'a') + obj.freeze() # Freeze the instance + obj.attribute1 = 2 # Raises AttributeError + + Args: + cls: The class to be decorated. + + Returns: + The decorated class with added freezing capability. + """ + + if not hasattr(cls, "__dict__"): + raise TypeError( + "make_freezable cannot be used with classes that do not define __dict__. Make sure that the wrapped " + "class was defined with `@attrs.define(slots=False)`" + ) + + original_setattr = cls.__setattr__ + + def setattr_override(self, key, value) -> None: # noqa: ANN001 + """ + Override __setattr__ to allow modifications during initialization + and prevent modifications once the instance is frozen. + """ + if hasattr(self, "_is_frozen") and self._is_frozen and key != "_is_frozen": + raise AttributeError("Cannot modify frozen instance") + original_setattr(self, key, value) # type: ignore + + cls.__setattr__ = setattr_override # type: ignore + + def freeze(self: object) -> None: + """ + Freeze the instance and all its attrs-defined attributes. + """ + for _, value in attrs.asdict(self, recurse=False).items(): + if _is_attrs_instance(value) and hasattr(value, "freeze"): + value.freeze() + self._is_frozen = True # type: ignore + + cls.freeze = freeze # type: ignore + + return cls + + +def _pretty_print_attrs_instance(obj: object, indent: int = 0, use_color: bool = False) -> str: + """ + Recursively pretty prints attrs objects with color. + """ + + assert attrs.has(obj.__class__) + + lines: list[str] = [] + for attribute in attrs.fields(obj.__class__): + value = getattr(obj, attribute.name) + if attrs.has(value.__class__): + if use_color: + lines.append(" " * indent + Color.cyan("* ") + Color.green(attribute.name) + ":") + else: + lines.append(" " * indent + "* " + attribute.name + ":") + lines.append(_pretty_print_attrs_instance(value, indent + 1, use_color)) + else: + if use_color: + lines.append( + " " * indent + Color.cyan("* ") + Color.green(attribute.name) + ": " + Color.yellow(value) + ) + else: + lines.append(" " * indent + "* " + attribute.name + ": " + str(value)) + return "\n".join(lines) + + +def pretty_print_overrides(overrides: Optional[list[str]] = None, use_color: bool = False) -> str: + """ + Pretty prints overrides. + """ + + lines: list[str] = [] + lines.append(Color.cyan("* ") + Color.green("overrides") + ": ") + for override in overrides: + if override == "--": + continue + if override.startswith("~"): + attribute_name = override[1:] + attribute_value = None + else: + attribute_name, attribute_value = override.split("=") + if use_color: + lines.append(" " + Color.cyan("* ") + Color.green(attribute_name) + ": " + Color.yellow(attribute_value)) + else: + lines.append(" " + "* " + attribute_name + ": " + str(attribute_value)) + + return "\n".join(lines) + + +@make_freezable +@attrs.define(slots=False) # slots=False is required for make_freezable. See the make_freezable notes for more info. +class ObjectStoreConfig: + # Whether the file I/O is from object store instead of local disk. + enabled: bool = False + # Path to the object store credentials file. + credentials: str = "" + # Object store bucket to read from / write to the objects. + bucket: str = "" + + +@make_freezable +@attrs.define(slots=False) +class JobConfig: + # Project name. + project: str = "" + # Experiment name. + group: str = "" + # Run/job name. + name: str = "" + # W&B mode, can be "online", or "disabled". + wandb_mode: str = "online" + # Cluster configuration (optional, for cluster-specific settings). + cluster: Optional[Any] = None + + @property + def path(self) -> str: + return f"{self.project}/{self.group}/{self.name}" + + @property + def path_local(self) -> str: + local_root = os.environ.get("IMAGINAIRE_OUTPUT_ROOT", "./checkpoints/imaginaire4-output") + return f"{local_root}/{self.path}" + + +@make_freezable +@attrs.define(slots=False) +class EMAConfig: + # Enable tracking a set of exponential moving average (EMA) weights. + enabled: bool = False + # EMA decay rate. + beta: float = 0.9999 + # Enable removing "_orig_mod-" from buffer names that is added by torch.compile + torch_compile_buffer_renaming: bool = False + + +@make_freezable +@attrs.define(slots=False) +class PowerEMAConfig: + # Enable tracking a set of exponential moving average (EMA) weights. + enabled: bool = False + # EDM2 paper EMA decay rate. + s: float = 0.1 + # Enable removing "_orig_mod-" from buffer names that is added by torch.compile + torch_compile_buffer_renaming: bool = False + + +@make_freezable +@attrs.define(slots=False) +class DDPConfig: + # Traverse the computation graph to find parameters that don't receive gradients. + find_unused_parameters: bool = False + # Set to True if the computation graph does not change during the whole training loop. + static_graph: bool = True + # Set to True if we want to synchronize buffers. Set to False if the sync is going to be handled elsewhere. + broadcast_buffers: bool = True + + +@make_freezable +@attrs.define(slots=False) +class CuDNNConfig: + # Set to True for better reproducibility of the results (only using deterministic cudnn functions). + deterministic: bool = False + # If set to True, cudnn will benchmark several algorithms and pick the fastest one. + benchmark: bool = True + + +@make_freezable +@attrs.define(slots=False) +class JITConfig: + # Enable exporting a JIT compiled model. + enabled: bool = False + # Input tensor shape, for example input. + input_shape: Union[list[int], None] = None + # Device to compile onto. + device: str = "cuda" + # # Data type to compile onto. + dtype: str = "bfloat16" + # Strict mode for PyTorch JIT. + strict: bool = True + + +@make_freezable +@attrs.define(slots=False) +class CheckpointConfig: + # possible checkpoint class + type: Optional[Dict] = None + # for dcp, whether to use async mode + dcp_async_mode_enabled: bool = False + # Configs for saving the checkpoints to object store. + save_to_object_store: ObjectStoreConfig = attrs.field(factory=ObjectStoreConfig) + # Save the checkpoint every N iterations. + save_iter: int = 999999999 + # Configs for loading the checkpoints from object store. + load_from_object_store: ObjectStoreConfig = attrs.field(factory=ObjectStoreConfig) + # Path of model weights to resume the checkpoint from. + load_path: str = "" + # Whether to load the training states (optimizer/scheduler/grad-scaler) from the checkpoint path. + load_training_state: bool = False + # Whether to load the scheduler state only from the checkpoint path. If load_training_state is True, this will be ignored. + only_load_scheduler_state: bool = False + # Load state_dict to the models in strict mode. + strict_resume: bool = True + # Configs for JIT compiling EMA model. + jit: JITConfig = attrs.field(factory=JITConfig) + # Print detailed information during checkpoint saving/loading. + verbose: bool = True + # keys not to resume from the checkpoint, choices: ["model", "optim", "scheduler", "trainer"] + keys_not_to_resume: list[str] = [] + # Whether to use the local filesystem for broadcasting checkpoint data (used for Tensor Parallel Checkpointer). + broadcast_via_filesystem: bool = False + load_ema_to_reg: bool = False + # In dcp planner, skip the weight shape check, load weights into the model even weight shape is different + dcp_allow_mismatched_size: bool = False + # Enable GCS patch in boto3 for loading/saving checkpoints from/to GCS + enable_gcs_patch_in_boto3: bool = False + + +@make_freezable +@attrs.define(slots=False) +class NVTXConfig: + """Config for NVTX ranges used in the main training loop. + + See tutorials/nanogpt for more details on how to integrate profiling into your model.""" + + # Enable the NVTX ranges. + enabled: bool = False + # Synchronize everything in each NVTX range. + cuda_synchronize: bool = False + + +@make_freezable +@attrs.define(slots=False) +class StragglerDetectionConfig: + """Config for Straggler detection tool: https://gitlab-master.nvidia.com/dl/gwe/fault_tolerance_related/straggler/-/tree/cupti?ref_type=heads""" + + # Enable the Straggler Detection. + enabled: bool = False + # How frequently should the Straggler reports be generated. + report_freq: int = 100 + # How frequently iterations should be profiled + profile_freq: int = 1 + # What is the maximum relative difference between GPUs after they are considered stragglers + max_diff: float = 2.0 + # Should the error be raised when straggler is detected + raise_error: bool = True + # Analyze kernels in the forward pass. + analyze_forward: bool = True + # Analyze kernels in the backward pass. + analyze_backward: bool = True + # Analyze kernels in the optimizer. + analyze_optimizer: bool = True + # Analyze dataloading time. + analyze_dataloading: bool = True + + +@make_freezable +@attrs.define(slots=False) +class Profiling: + enable_profiling: bool = False + enable_memory_snapshot: bool = False + save_s3: bool = False + profile_freq: int = 1 + # Target ranks for profiling, each entry must be >=0 and < world_size. + target_ranks: list[int] = list(range(8)) + # Set `record_shape` and `profile_memory` to False to reduce profile size. + record_shape: bool = False + profile_memory: bool = False + with_stack: bool = True + with_modules: bool = True + + +@make_freezable +@attrs.define(slots=False) +class TrainerConfig: + from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer + + type: Type[ImaginaireTrainer] = ImaginaireTrainer + # Set the callback class. + # Defaults to the callbacks below. + callbacks: LazyDict = LazyDict( + dict( + ema=L(callback.EMAModelCallback)(), + progress_bar=L(callback.ProgressBarCallback)(), + wandb=L(callback.WandBCallback)(), + ) + ) + # distributed parallelism strategy + distributed_parallelism: str = "ddp" + # Distributed data parallel configs. + ddp: DDPConfig = attrs.field(factory=DDPConfig) + # cuDNN configs. + cudnn: CuDNNConfig = attrs.field(factory=CuDNNConfig) + # Set the random seed. + seed: int = 0 + # Gradient scaler arguments (for torch.amp.GradScaler). + grad_scaler_args: dict = attrs.field(factory=lambda: dict(enabled=False)) + # Maximum number of iterations to train the model. + max_iter: int = 999999999 + # Maximum number of iterations to validate the model. If None, validate on the entire dataset. + max_val_iter: int | None = None + # How often we log the training stats. + logging_iter: int = 100 + # Whether we want to run the validation routines. + run_validation: bool = True + # How often we evaluate on the validation set. + validation_iter: int = 999999999 + # Whether to run the validation on the start of the training. + run_validation_on_start: bool = False + # Kill the process after N seconds since the last iteration (usually means dead job). + timeout_period: int = 999999999 + # Tensor memory organization format. + memory_format: torch.memory_format = torch.preserve_format + # Gradient accumulation (update step every N iteration). + grad_accum_iter: int = 1 + # Straggler Detection config + straggler_detection: StragglerDetectionConfig = attrs.field(factory=StragglerDetectionConfig) + # Profiling config + profiling: Profiling = attrs.field(factory=Profiling) + + +@make_freezable +@attrs.define(slots=False) +class Config: + """Config for an imaginaire4 job. + + See /README.md/Configuration System for more info. + """ + + # Model configs. + model: LazyDict + # Optimizer configs. + optimizer: LazyDict + # Scheduler configs. + scheduler: LazyDict + # Training data configs. + dataloader_train: LazyDict + # Validation data configs. + dataloader_val: LazyDict + + # Training job configs. + job: JobConfig = attrs.field(factory=JobConfig) + + # Trainer configs. + trainer: TrainerConfig = attrs.field(factory=TrainerConfig) + + if USE_MEGATRON: + # Megatron-Core configs + model_parallel: ModelParallelConfig = attrs.field(factory=ModelParallelConfig) + else: + model_parallel: None = None + + # Checkpointer configs. + checkpoint: CheckpointConfig = attrs.field(factory=CheckpointConfig) + + # enable upload reproducible setup to s3 + upload_reproducible_setup: bool = False + + def pretty_print(self, use_color: bool = False) -> str: + return _pretty_print_attrs_instance(self, 0, use_color) + + def to_dict(self) -> dict[str, Any]: + return attrs.asdict(self) + + def validate(self) -> None: + """Validate that the config has all required fields.""" + + # broadcast job.name across all ranks to make sure it is consistent + # otherwise, unaligned job names leads unaligned path to save checkpoints + job_name_tensor = torch.ByteTensor(bytearray(self.job.name, "utf-8")).cuda() + distributed.broadcast(job_name_tensor, 0) + self.job.name = job_name_tensor.cpu().numpy().tobytes().decode("utf-8") + + assert self.job.project != "" + assert self.job.group != "" + assert self.job.name != "" + + +def load_config(config_path: str, opts: list[str], enable_one_logger: bool = False) -> Config: + t1 = time.monotonic_ns() + if config_path.endswith(".yaml"): + config = from_yaml(config_path) + # for registration of dataloaders, etc. + _ = load_callable(config.__module__).make_config() + + from cosmos_policy._src.imaginaire.utils.config_helper import override + + config = override(config, opts, remove_defaults=True) + else: + config = _load_py_config(config_path, opts, validate=False) + + if enable_one_logger: + try: + # pyrefly: ignore # missing-import + from cosmos_policy._src.imaginaire.utils.one_logger.one_logger_override_utils import ( + override_one_logger_callback, + ) + + ol_t1 = time.monotonic_ns() + config = override_one_logger_callback(config) + ol_t2 = time.monotonic_ns() + logging.debug(f"override_one_logger_callback: took {(ol_t2 - ol_t1) / 1e6:.2f}ms") + except ImportError: + pass + + t2 = time.monotonic_ns() + logging.debug(f"toal time to load config: {(t2 - t1) / 1e6:.2f}ms") + return config + + +def _load_py_config(config_path: str, opts: list[str], validate: bool = True) -> Config: + # NOTE: circular dependency + from cosmos_policy._src.imaginaire.utils.config_helper import get_config_module, override + + t1 = time.monotonic_ns() + config_module = get_config_module(config_path) + t2 = time.monotonic_ns() + logging.debug(f"get_config_module: took {(t2 - t1) / 1e6:.2f}ms") + + t1 = time.monotonic_ns() + config = importlib.import_module(config_module).make_config() + t2 = time.monotonic_ns() + logging.debug(f"importlib.import_module: took {(t2 - t1) / 1e6:.2f}ms") + + t1 = time.monotonic_ns() + config = override(config, opts) + t2 = time.monotonic_ns() + logging.debug(f"override: took {(t2 - t1) / 1e6:.2f}ms") + + if validate: + t1 = time.monotonic_ns() + config.validate() + t2 = time.monotonic_ns() + logging.debug(f"config.validate: took {(t2 - t1) / 1e6:.2f}ms") + + return config diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/configs/lr_scheduler.py b/REGEN-main/cosmos_policy/_src/imaginaire/configs/lr_scheduler.py new file mode 100644 index 0000000000000000000000000000000000000000..750bee19fc839356d902ed7d4e3f7bf354bd0d9c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/configs/lr_scheduler.py @@ -0,0 +1,26 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from cosmos_policy._src.imaginaire.functional.lr_scheduler import LambdaLinearScheduler +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict + +LambdaLinearSchedulerConfig: LazyDict = L(LambdaLinearScheduler)( + warm_up_steps=[1000], + cycle_lengths=[10000000000000], + f_start=[1.0e-6], + f_max=[1.0], + f_min=[1.0], +) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/datasets/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/datasets/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/datasets/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/datasets/joint_training.py b/REGEN-main/cosmos_policy/_src/imaginaire/datasets/joint_training.py new file mode 100644 index 0000000000000000000000000000000000000000..3546200f8b91d15418a12017d3e15be63022ac6b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/datasets/joint_training.py @@ -0,0 +1,105 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""Utility funcitons to use joint dataloader for training.""" + +from typing import Dict, Iterator # For multiview training + +import torch + +import cosmos_policy._src.imaginaire.config +import cosmos_policy._src.imaginaire.datasets.webdataset.dataloader +from cosmos_policy._src.imaginaire.config import Config +from cosmos_policy._src.imaginaire.lazy_config import instantiate +from cosmos_policy._src.imaginaire.utils import log + + +def create_dataloader_dict( + config: Config, dataloader_train: cosmos_policy._src.imaginaire.datasets.webdataset.dataloader.DataLoader +) -> Dict: + """Create the dataloader dictionary. + + Example config: + + ``` + config: + joint_train: + data_sample_prob: + dataloader_train: 0.5 # sampling probability for default dataloader + dataloader_1: 0.2 # sampling probability for dataloader_1 + dataloader_2: 0.3 # sampling probability for dataloader_2 + dataloader_1: + ... # dataloader config for dataloader_1 + dataloader_2: + ... # dataloader config for dataloader_2 + ``` + + Args: + config (Config): The config object for the Imaginaire codebase. + + Returns: + dict: The dataloader dictionary. + """ + dataloader_list = list(config.joint_train.data_sample_prob.keys()) + + dataloader_dict = {} + for dataloader_name in dataloader_list: + if dataloader_name == "dataloader_train": + continue + log.info( + f"Creating dataloader: {dataloader_name}, sampling probability: {config.joint_train.data_sample_prob[dataloader_name]}" + ) + dataloader_dict[dataloader_name] = iter(instantiate(getattr(config.joint_train, dataloader_name))) + dataloader_dict["dataloader_train"] = iter(dataloader_train) + return dataloader_dict + + +def data_batch_iterator(dataloader_dict: Dict, data_sample_prob: Dict) -> Iterator[Dict]: + """Sample data batches continuously from the dataloader dictionary based on sampling probabilities.""" + dataloader_list = list(data_sample_prob.keys()) + + while True: + selected_dataloader_id = torch.multinomial( + torch.tensor([data_sample_prob[dataloader_name] for dataloader_name in dataloader_list]), 1 + ).item() + selected_dataloader_name = dataloader_list[selected_dataloader_id] + selected_dataloader = dataloader_dict[selected_dataloader_name] + + try: + data_batch = next(selected_dataloader) + except StopIteration: + # Reinitialize the iterator for the selected dataloader once it is exhausted + dataloader_dict[dataloader_list[selected_dataloader_id]] = iter(selected_dataloader) + data_batch = next(dataloader_dict[dataloader_list[selected_dataloader_id]]) + data_batch["dataloader_name"] = selected_dataloader_name + yield data_batch + + +def init_and_wrap_data_loaders(config: Config, dataloader_train: torch.utils.data.DataLoader) -> Dict: + """Wrap the dataloaders for multiview training. + + Args: + config (Config): The config object for the Imaginaire codebase. + dataloader_train (torch.utils.data.DataLoader): The training data loader. + + Returns: + dict: The dataloader dictionary. + """ + # Create the dataloader dictionary with multiple dataloaders + dataloader_dict = create_dataloader_dict(config, dataloader_train) + + # Create the data batch iterator sample from the dataloader dictionary based on sampling probabilities + dataloader_train = data_batch_iterator(dataloader_dict, config.joint_train.data_sample_prob) + return dataloader_train diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/datasets/mock_dataset.py b/REGEN-main/cosmos_policy/_src/imaginaire/datasets/mock_dataset.py new file mode 100644 index 0000000000000000000000000000000000000000..8d0c2d7dccf9c1a25eaaa754617c8e2094a4ecbe --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/datasets/mock_dataset.py @@ -0,0 +1,186 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Copied from jam_data by Qinsheng Zhang, with unknown license. +""" + +import inspect +from typing import Any, Callable, Dict + +import torch +from torch.utils.data import Dataset + +MAX_LENGTH = 1 << 15 + + +class LambdaDataset(torch.utils.data.Dataset): + """ + A dataset that generates items by applying a function. This allows for creating + dynamic datasets where the items are the result of function calls. The function can optionally + accept an index argument. + + Attributes: + length (int): The total number of items in the dataset. + fn (Callable): The function to generate dataset items. + is_index_in_params (bool): Flag to determine whether 'index' should be passed + to the function `fn`. + """ + + def __init__(self, fn: Callable, length: int = MAX_LENGTH) -> None: + """ + Initializes the LambdaDataset with a function and the total length. + + Args: + fn (Callable): A function that returns a dataset item. It can optionally accept an + index argument to generate data items based on their index. + length (int): The total number of items in the dataset, defaults to MAX_LENGTH. + """ + self.length = length + self.fn = fn + + try: + # Attempt to inspect the function signature to determine if it accepts an 'index' parameter. + signature = inspect.signature(fn) + self.is_index_in_params = "index" in signature.parameters + except ValueError: + # If the function signature is not inspectable, assume 'index' is not a parameter. + self.is_index_in_params = False + + def __len__(self) -> int: + """ + Returns the total length of the dataset. + + Returns: + int: The number of items in the dataset. + """ + return self.length + + def __getitem__(self, index: int) -> Any: + """ + Retrieves an item at a specific index from the dataset by calling the function `fn`. + Passes the index to `fn` if `fn` is designed to accept an index. + + Args: + index (int): The index of the item to retrieve. + + Returns: + Any: The item returned by the function `fn`. + """ + if self.is_index_in_params: + return self.fn(index) # Call fn with index if it accepts an index parameter. + return self.fn() # Call fn without any parameters if it does not accept the index. + + +class RepeatDataset(torch.utils.data.Dataset): + """ + A dataset wrapper that allows repeating access to items from an underlying dataset. + + This dataset can be used to create an artificial extension of the underlying dataset + to a specified `length`. Each item from the original dataset can be accessed + repeatedly up to `num_item` times before it loops back. + + Attributes: + length (int): The total length of the dataset to be exposed. + dataset (Dataset): The original dataset. + num_item (int): Number of times each item is repeated. + cache_item (dict): Cache to store accessed items to avoid recomputation. + """ + + def __init__(self, dataset: Dataset, length: int = MAX_LENGTH, num_item: int = 1) -> None: + """ + Initializes the RepeatDataset with a dataset, length, and number of repeats per item. + + Args: + dataset (Dataset): The dataset to repeat. + length (int): The total length of the dataset to be exposed. Defaults to MAX_LENGTH. + num_item (int): The number of times to repeat each item. Defaults to 1. + """ + self.length = length + self.dataset = dataset + self.num_item = num_item + self.cache_item = {} + + def __len__(self) -> int: + return self.length + + def __getitem__(self, index: int) -> Any: + index = index % self.num_item + if index not in self.cache_item: + self.cache_item[index] = self.dataset[index] + return self.cache_item[index] + + +class CombinedDictDataset(torch.utils.data.Dataset): + """ + A dataset that wraps multiple PyTorch datasets and returns a dictionary of data items from each dataset for a given index. + This dataset ensures that all constituent datasets have the same length by setting the length to the minimum length of the datasets provided. + + Parameters: + ----------- + **datasets : Dict[str, Dataset] + A dictionary where keys are string identifiers for the datasets and values are the datasets instances themselves. + + Attributes: + ----------- + datasets : Dict[str, Dataset] + Stores the input datasets. + max_length : int + The minimum length among all provided datasets, determining the length of this combined dataset. + + Examples: + --------- + >>> dataset1 = torch.utils.data.TensorDataset(torch.randn(100, 3, 32, 32)) + >>> dataset2 = torch.utils.data.TensorDataset(torch.randn(100, 3, 32, 32)) + >>> combined_dataset = CombinedDictDataset(dataset1=dataset1, dataset2=dataset2) + >>> print(len(combined_dataset)) + 100 + >>> data = combined_dataset[50] + >>> print(data.keys()) + dict_keys(['dataset1', 'dataset2']) + """ + + def __init__(self, **datasets: Dict[str, Dataset]) -> None: + """ + Initializes the CombinedDictDataset with multiple datasets. + + Args: + **datasets (Dict[str, Dataset]): Key-value pairs where keys are dataset names and values + are dataset instances. Each key-value pair adds a dataset + under the specified key. + """ + self.datasets = datasets + self.max_length = min([len(dataset) for dataset in datasets.values()]) + + def __len__(self) -> int: + return self.max_length + + def __getitem__(self, index: int) -> Dict[str, Any]: + """ + Retrieves an item from each dataset at the specified index, combines them into a dictionary, + and returns the dictionary. Each key in the dictionary corresponds to one of the dataset names provided + during initialization, and its value is the item from that dataset at the given index. + + Args: + index (int): The index of the items to retrieve across all datasets. + + Returns: + Dict[str, Any]: A dictionary containing data items from all datasets for the given index. + Each key corresponds to a dataset name, and its value is the data item from that dataset. + """ + data = {} + for key, dataset in self.datasets.items(): + data[key] = dataset[index] + return data diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/datasets/mock_dataset_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/datasets/mock_dataset_test.py new file mode 100644 index 0000000000000000000000000000000000000000..a4867ba7508c4c7176007083f4a727f30e8885ce --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/datasets/mock_dataset_test.py @@ -0,0 +1,59 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Usage: + pytest -s cosmos_policy/_src/imaginaire/datasets/mock_dataset_test.py +""" + +import pytest +import torch + +from cosmos_policy._src.imaginaire.datasets.mock_dataset import CombinedDictDataset, LambdaDataset, RepeatDataset +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import instantiate + + +@pytest.fixture +def cfg(): + return L(CombinedDictDataset)( + key1=L(LambdaDataset)( + length=64, + fn=lambda: torch.randn(3, 32, 32), + ), + key2=L(RepeatDataset)( + dataset=L(LambdaDataset)( + fn=lambda: torch.randn(3, 32, 32), + ), + ), + ) + + +@pytest.mark.L0 +def test_mock_dataset(cfg): + batch_size = 4 + dataset_obj = instantiate(cfg) + dataloader = torch.utils.data.DataLoader( + dataset=dataset_obj, + batch_size=batch_size, + pin_memory=True, + num_workers=1, + ) + assert len(dataset_obj) == 64 + for ith, batch in enumerate(dataloader): + assert batch["key1"].shape == (batch_size, 3, 32, 32) + assert batch["key2"].shape == (batch_size, 3, 32, 32) + if ith > 2: + break diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/flags.py b/REGEN-main/cosmos_policy/_src/imaginaire/flags.py new file mode 100644 index 0000000000000000000000000000000000000000..902528e6174b324d4ed12f7ba7f3d54cdfb52c4b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/flags.py @@ -0,0 +1,52 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""Feature flags.""" + +import os +from dataclasses import dataclass + + +def _parse_bool(value: str) -> bool: + """Parse string to a boolean.""" + return value.lower() in ["true", "1", "yes", "y"] + + +INTERNAL = _parse_bool(os.environ.get("COSMOS_INTERNAL", "0")) +"""Whether to enable internal (nvidia-only) features.""" + +SMOKE = _parse_bool(os.environ.get("COSMOS_SMOKE", "0")) +"""Whether to enable smoke test. + +Disables expensive operations such as checkpoint loading. +""" + +VERBOSE = _parse_bool(os.environ.get("COSMOS_VERBOSE", "0")) +"""Whether to enable verbose output.""" + +EXPERIMENTAL_CHECKPOINTS = _parse_bool(os.environ.get("COSMOS_EXPERIMENTAL_CHECKPOINTS", "0")) +"""Whether to enable experimental checkpoints.""" + + +@dataclass +class Flags: + internal: bool = INTERNAL + smoke: bool = SMOKE + verbose: bool = VERBOSE + experimental_checkpoints: bool = EXPERIMENTAL_CHECKPOINTS + + +FLAGS = Flags() +"""Convenience object for accessing flags.""" diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/functional/batch_ops.py b/REGEN-main/cosmos_policy/_src/imaginaire/functional/batch_ops.py new file mode 100644 index 0000000000000000000000000000000000000000..a72b24097f7cc9e7e6a8b324919455131bf84d47 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/functional/batch_ops.py @@ -0,0 +1,61 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Functions for performing operations with broadcasting to the right axis +# +# Example +# input1: tensor of size (N1, N2) +# input2: tensor of size (N1, N2, N3, N4) +# batch_mul(input1, input2) = input1[:, :, None, None] * input2 +# +# If the common dimensions don't match, we raise an assertion error. + +from torch import Tensor + + +def common_broadcast(x: Tensor, y: Tensor) -> tuple[Tensor, Tensor]: + ndims1 = x.ndim + ndims2 = y.ndim + + common_ndims = min(ndims1, ndims2) + for axis in range(common_ndims): + assert x.shape[axis] == y.shape[axis], "Dimensions not equal at axis {}".format(axis) + + if ndims1 < ndims2: + x = x.reshape(x.shape + (1,) * (ndims2 - ndims1)) + elif ndims2 < ndims1: + y = y.reshape(y.shape + (1,) * (ndims1 - ndims2)) + + return x, y + + +def batch_add(x: Tensor, y: Tensor) -> Tensor: + x, y = common_broadcast(x, y) + return x + y + + +def batch_mul(x: Tensor, y: Tensor) -> Tensor: + x, y = common_broadcast(x, y) + return x * y + + +def batch_sub(x: Tensor, y: Tensor) -> Tensor: + x, y = common_broadcast(x, y) + return x - y + + +def batch_div(x: Tensor, y: Tensor) -> Tensor: + x, y = common_broadcast(x, y) + return x / y diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/functional/lr_scheduler.py b/REGEN-main/cosmos_policy/_src/imaginaire/functional/lr_scheduler.py new file mode 100644 index 0000000000000000000000000000000000000000..7f3d78b66ad6a7f09cdaf28185395f37ee49e0a0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/functional/lr_scheduler.py @@ -0,0 +1,178 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Optional + +import numpy as np + +from cosmos_policy._src.imaginaire.utils import distributed, log + + +class TeroPolyScheduler: + def __init__( + self, + total_Mimg: int, + batch_size: int, + ref_Mimg: Optional[int] = None, + ref_batches: float = 70e3 / 1024, + max_lr_ratio: Optional[float] = 1.0, + min_lr_ratio: Optional[float] = None, + rampup_Mimg: float = 0, + rampdown_Mimg: int = 0, + verbosity_interval: int = 0, + formula: str = "poly", + poly_exp: float = 0.5, + ): + self.total_Mimg = total_Mimg + self.batch_size = batch_size * distributed.get_world_size() + self.ref_Mimg = ref_Mimg or ref_batches * batch_size / 1e6 + self.ref_batches = ref_batches + self.max_lr_ratio = max_lr_ratio + self.min_lr_ratio = min_lr_ratio + self.rampup_Mimg = rampup_Mimg + self.rampdown_Mimg = rampdown_Mimg + self.verbosity_interval = verbosity_interval + self.formula = formula + self.poly_exp = poly_exp + + self._model = None + + @property + def model(self): + return self._model + + @model.setter + def model(self, model): + self._model = model + + def schedule(self, n, **kwargs): + cur_Mimg = getattr(self.model, "sample_counter", 0) / 1e6 + + if self.formula == "constant": + lr = 1.0 + elif self.formula == "poly": + lr = max(cur_Mimg / self.ref_Mimg, 1e-8) ** -self.poly_exp + else: + raise ValueError(f'Invalid learning rate formula "{self.formula}"') + + if self.max_lr_ratio is not None: + lr = min(lr, self.max_lr_ratio) + if self.min_lr_ratio is not None: + lr = max(lr, self.min_lr_ratio) + + if self.rampup_Mimg > 0 and cur_Mimg < self.rampup_Mimg: + lr *= cur_Mimg / self.rampup_Mimg + if self.rampdown_Mimg > 0 and cur_Mimg > self.total_Mimg - self.rampdown_Mimg: + lr *= (self.total_Mimg - cur_Mimg) / self.rampdown_Mimg + + return lr + + def __call__(self, n, **kwargs): + return self.schedule(n, **kwargs) + + +class LambdaWarmUpCosineScheduler: + """ + A learning rate scheduler that combines warm-up with a cosine decay schedule for multiple cycles. + It supports different configurations for each cycle, including the number of warm-up steps, minimum + and maximum scaling factors for the learning rate. + + The scheduler is intended to be used with a base learning rate of 1.0, where the actual learning + rate at any step is the base learning rate multiplied by the scaling factor computed by the scheduler. + + Parameters: + warm_up_steps (list[int]): List of integers where each element represents the number of warm-up + steps for the corresponding cycle. + f_min (list[float]): List of the minimum scaling factors for each cycle after warm-up. + f_max (list[float]): List of the maximum scaling factors at the start and end of each cosine cycle. + f_start (list[float]): List of starting scaling factors for each warm-up phase. + cycle_lengths (list[int]): List of the total lengths of each cycle, including warm-up steps. + verbosity_interval (int, optional): Interval of training steps at which to print current step and + scaling factor information. Set to 0 by default to disable verbosity. + + Examples: + >>> scheduler = LambdaWarmUpCosineScheduler2( + warm_up_steps=[10, 10], + f_min=[0.1, 0.1], + f_max=[1.0, 1.0], + f_start=[0.01, 0.01], + cycle_lengths=[50, 50], + verbosity_interval=10) + >>> for step in range(100): + >>> lr_multiplier = scheduler(step) + >>> print(f"Step {step}: LR Multiplier = {lr_multiplier}") + """ + + def __init__(self, warm_up_steps, f_min, f_max, f_start, cycle_lengths, verbosity_interval=0): + assert len(warm_up_steps) == len(f_min) == len(f_max) == len(f_start) == len(cycle_lengths) + self.lr_warm_up_steps = warm_up_steps + self.f_start = f_start + self.f_min = f_min + self.f_max = f_max + self.cycle_lengths = cycle_lengths + self.cum_cycles = np.cumsum([0] + list(self.cycle_lengths)) + self.last_f = 0.0 + self.verbosity_interval = verbosity_interval + + def find_in_interval(self, n): + interval = 0 + for cl in self.cum_cycles[1:]: + if n <= cl: + return interval + interval += 1 + + def schedule(self, n, **kwargs): + cycle = self.find_in_interval(n) + n = n - self.cum_cycles[cycle] + if self.verbosity_interval > 0: + if n % self.verbosity_interval == 0: + log.info(f"current step: {n}, recent lr-multiplier: {self.last_f}, current cycle {cycle}") + if n < self.lr_warm_up_steps[cycle]: + f = (self.f_max[cycle] - self.f_start[cycle]) / self.lr_warm_up_steps[cycle] * n + self.f_start[cycle] + self.last_f = f + return f + else: + t = (n - self.lr_warm_up_steps[cycle]) / (self.cycle_lengths[cycle] - self.lr_warm_up_steps[cycle]) + t = min(t, 1.0) + f = self.f_min[cycle] + 0.5 * (self.f_max[cycle] - self.f_min[cycle]) * (1 + np.cos(t * np.pi)) + self.last_f = f + return f + + def __call__(self, n, **kwargs): + return self.schedule(n, **kwargs) + + +class LambdaLinearScheduler(LambdaWarmUpCosineScheduler): + """ + Linear instead of cosine decay for the main part of the cycle. + """ + + def schedule(self, n, **kwargs): + cycle = self.find_in_interval(n) + n = n - self.cum_cycles[cycle] + if self.verbosity_interval > 0: + if n % self.verbosity_interval == 0: + log.info(f"current step: {n}, recent lr-multiplier: {self.last_f}, current cycle {cycle}") + + if n < self.lr_warm_up_steps[cycle]: + f = (self.f_max[cycle] - self.f_start[cycle]) / self.lr_warm_up_steps[cycle] * n + self.f_start[cycle] + self.last_f = f + return f + else: + f = self.f_min[cycle] + (self.f_max[cycle] - self.f_min[cycle]) * (self.cycle_lengths[cycle] - n) / ( + self.cycle_lengths[cycle] - self.lr_warm_up_steps[cycle] + ) + self.last_f = f + return f diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/functional/multi_step.py b/REGEN-main/cosmos_policy/_src/imaginaire/functional/multi_step.py new file mode 100644 index 0000000000000000000000000000000000000000..3dbc61be2cada614363f6b5dc8ee192c3605b998 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/functional/multi_step.py @@ -0,0 +1,60 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Impl of multistep methods to solve the ODE in the diffusion model. +""" + +from typing import Callable, List, Tuple + +import torch + +from cosmos_policy._src.imaginaire.functional.runge_kutta import reg_x0_euler_step, res_x0_rk2_step + + +def order2_fn( + x_s: torch.Tensor, s: torch.Tensor, t: torch.Tensor, x0_s: torch.Tensor, x0_preds: torch.Tensor +) -> Tuple[torch.Tensor, List[torch.Tensor]]: + """ + impl the second order multistep method in https://arxiv.org/pdf/2308.02157 + Adams Bashforth approach! + """ + if x0_preds: + x0_s1, s1 = x0_preds[0] + x_t = res_x0_rk2_step(x_s, t, s, x0_s, s1, x0_s1) + else: + x_t = reg_x0_euler_step(x_s, s, t, x0_s)[0] + return x_t, [(x0_s, s)] + + +# key: method name, value: method function +# key: order + algorithm name +MULTISTEP_FNs = { + "2ab": order2_fn, +} + + +def get_multi_step_fn(name: str) -> Callable: + if name in MULTISTEP_FNs: + return MULTISTEP_FNs[name] + methods = "\n\t".join(MULTISTEP_FNs.keys()) + raise RuntimeError("Only support multistep method\n" + methods) + + +def is_multi_step_fn_supported(name: str) -> bool: + """ + Check if the multistep method is supported. + """ + return name in MULTISTEP_FNs diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/functional/runge_kutta.py b/REGEN-main/cosmos_policy/_src/imaginaire/functional/runge_kutta.py new file mode 100644 index 0000000000000000000000000000000000000000..e95297ec8af14f037446f8ece44e84103e69c250 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/functional/runge_kutta.py @@ -0,0 +1,333 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Callable, Tuple + +import torch + +from cosmos_policy._src.imaginaire.functional.batch_ops import batch_mul + + +def phi1(t: torch.Tensor) -> torch.Tensor: + """ + Compute the first order phi function: (exp(t) - 1) / t. + + Args: + t: Input tensor. + + Returns: + Tensor: Result of phi1 function. + """ + input_dtype = t.dtype + t = t.to(dtype=torch.float64) + return (torch.expm1(t) / t).to(dtype=input_dtype) + + +def phi2(t: torch.Tensor) -> torch.Tensor: + """ + Compute the second order phi function: (phi1(t) - 1) / t. + + Args: + t: Input tensor. + + Returns: + Tensor: Result of phi2 function. + """ + input_dtype = t.dtype + t = t.to(dtype=torch.float64) + return ((phi1(t) - 1.0) / t).to(dtype=input_dtype) + + +def res_x0_rk2_step( + x_s: torch.Tensor, + t: torch.Tensor, + s: torch.Tensor, + x0_s: torch.Tensor, + s1: torch.Tensor, + x0_s1: torch.Tensor, +) -> torch.Tensor: + """ + Perform a residual-based 2nd order Runge-Kutta step. + + Args: + x_s: Current state tensor. + t: Target time tensor. + s: Current time tensor. + x0_s: Prediction at current time. + s1: Intermediate time tensor. + x0_s1: Prediction at intermediate time. + + Returns: + Tensor: Updated state tensor. + + Raises: + AssertionError: If step size is too small. + """ + s = -torch.log(s) + t = -torch.log(t) + m = -torch.log(s1) + + dt = t - s + assert not torch.any(torch.isclose(dt, torch.zeros_like(dt), atol=1e-6)), "Step size is too small" + assert not torch.any(torch.isclose(m - s, torch.zeros_like(dt), atol=1e-6)), "Step size is too small" + + c2 = (m - s) / dt + phi1_val, phi2_val = phi1(-dt), phi2(-dt) + + # Handle edge case where t = s = m + b1 = torch.nan_to_num(phi1_val - 1.0 / c2 * phi2_val, nan=0.0) + b2 = torch.nan_to_num(1.0 / c2 * phi2_val, nan=0.0) + + return batch_mul(torch.exp(-dt), x_s) + batch_mul(dt, batch_mul(b1, x0_s) + batch_mul(b2, x0_s1)) + + +def reg_x0_euler_step( + x_s: torch.Tensor, + s: torch.Tensor, + t: torch.Tensor, + x0_s: torch.Tensor, +) -> Tuple[torch.Tensor, torch.Tensor]: + """ + Perform a regularized Euler step based on x0 prediction. + + Args: + x_s: Current state tensor. + s: Current time tensor. + t: Target time tensor. + x0_s: Prediction at current time. + + Returns: + Tuple[Tensor, Tensor]: Updated state tensor and current prediction. + """ + coef_x0 = (s - t) / s + coef_xs = t / s + return batch_mul(coef_x0, x0_s) + batch_mul(coef_xs, x_s), x0_s + + +def reg_eps_euler_step( + x_s: torch.Tensor, s: torch.Tensor, t: torch.Tensor, eps_s: torch.Tensor +) -> Tuple[torch.Tensor, torch.Tensor]: + """ + Perform a regularized Euler step based on epsilon prediction. + + Args: + x_s: Current state tensor. + s: Current time tensor. + t: Target time tensor. + eps_s: Epsilon prediction at current time. + + Returns: + Tuple[Tensor, Tensor]: Updated state tensor and current x0 prediction. + """ + return x_s + batch_mul(eps_s, t - s), x_s + batch_mul(eps_s, 0 - s) + + +def rk1_euler( + x_s: torch.Tensor, s: torch.Tensor, t: torch.Tensor, x0_fn: Callable +) -> Tuple[torch.Tensor, torch.Tensor]: + """ + Perform a first-order Runge-Kutta (Euler) step. + + Recommended for diffusion models with guidance or model undertrained + Usually more stable at the cost of a bit slower convergence. + + Args: + x_s: Current state tensor. + s: Current time tensor. + t: Target time tensor. + x0_fn: Function to compute x0 prediction. + + Returns: + Tuple[Tensor, Tensor]: Updated state tensor and x0 prediction. + """ + x0_s = x0_fn(x_s, s) + return reg_x0_euler_step(x_s, s, t, x0_s) + + +def rk2_mid_stable( + x_s: torch.Tensor, s: torch.Tensor, t: torch.Tensor, x0_fn: Callable +) -> Tuple[torch.Tensor, torch.Tensor]: + """ + Perform a stable second-order Runge-Kutta (midpoint) step. + + Args: + x_s: Current state tensor. + s: Current time tensor. + t: Target time tensor. + x0_fn: Function to compute x0 prediction. + + Returns: + Tuple[Tensor, Tensor]: Updated state tensor and x0 prediction. + """ + s1 = torch.sqrt(s * t) + x_s1, _ = rk1_euler(x_s, s, s1, x0_fn) + + x0_s1 = x0_fn(x_s1, s1) + return reg_x0_euler_step(x_s, s, t, x0_s1) + + +def rk2_mid(x_s: torch.Tensor, s: torch.Tensor, t: torch.Tensor, x0_fn: Callable) -> Tuple[torch.Tensor, torch.Tensor]: + """ + Perform a second-order Runge-Kutta (midpoint) step. + + Args: + x_s: Current state tensor. + s: Current time tensor. + t: Target time tensor. + x0_fn: Function to compute x0 prediction. + + Returns: + Tuple[Tensor, Tensor]: Updated state tensor and x0 prediction. + """ + s1 = torch.sqrt(s * t) + x_s1, x0_s = rk1_euler(x_s, s, s1, x0_fn) + + x0_s1 = x0_fn(x_s1, s1) + + return res_x0_rk2_step(x_s, t, s, x0_s, s1, x0_s1), x0_s1 + + +def rk_2heun_naive( + x_s: torch.Tensor, s: torch.Tensor, t: torch.Tensor, x0_fn: Callable +) -> Tuple[torch.Tensor, torch.Tensor]: + """ + Perform a naive second-order Runge-Kutta (Heun's method) step. + Impl based on rho-rk-deis solvers, https://github.com/qsh-zh/deis + Recommended for diffusion models without guidance and relative large NFE + + Args: + x_s: Current state tensor. + s: Current time tensor. + t: Target time tensor. + x0_fn: Function to compute x0 prediction. + + Returns: + Tuple[Tensor, Tensor]: Updated state tensor and current state. + """ + x_t, x0_s = rk1_euler(x_s, s, t, x0_fn) + eps_s = batch_mul(1.0 / s, x_t - x0_s) + x0_t = x0_fn(x_t, t) + eps_t = batch_mul(1.0 / t, x_t - x0_t) + + avg_eps = (eps_s + eps_t) / 2 + + return reg_eps_euler_step(x_s, s, t, avg_eps) + + +def rk_2heun_edm( + x_s: torch.Tensor, s: torch.Tensor, t: torch.Tensor, x0_fn: Callable +) -> Tuple[torch.Tensor, torch.Tensor]: + """ + Perform a naive second-order Runge-Kutta (Heun's method) step. + Impl based no EDM second order Heun method + + Args: + x_s: Current state tensor. + s: Current time tensor. + t: Target time tensor. + x0_fn: Function to compute x0 prediction. + + Returns: + Tuple[Tensor, Tensor]: Updated state tensor and current state. + """ + x_t, x0_s = rk1_euler(x_s, s, t, x0_fn) + x0_t = x0_fn(x_t, t) + + avg_x0 = (x0_s + x0_t) / 2 + + return reg_x0_euler_step(x_s, s, t, avg_x0) + + +def rk_3kutta_naive( + x_s: torch.Tensor, s: torch.Tensor, t: torch.Tensor, x0_fn: Callable +) -> Tuple[torch.Tensor, torch.Tensor]: + """ + Perform a naive third-order Runge-Kutta step. + Impl based on rho-rk-deis solvers, https://github.com/qsh-zh/deis + Recommended for diffusion models without guidance and relative large NFE + + Args: + x_s: Current state tensor. + s: Current time tensor. + t: Target time tensor. + x0_fn: Function to compute x0 prediction. + + Returns: + Tuple[Tensor, Tensor]: Updated state tensor and current state. + """ + c2, c3 = 0.5, 1.0 + a31, a32 = -1.0, 2.0 + b1, b2, b3 = 1.0 / 6, 4.0 / 6, 1.0 / 6 + + delta = t - s + + s1 = c2 * delta + s + s2 = c3 * delta + s + x_s1, x0_s = rk1_euler(x_s, s, s1, x0_fn) + eps_s = batch_mul(1.0 / s, x_s - x0_s) + x0_s1 = x0_fn(x_s1, s1) + eps_s1 = batch_mul(1.0 / s1, x_s1 - x0_s1) + + _eps = a31 * eps_s + a32 * eps_s1 + x_s2, _ = reg_eps_euler_step(x_s, s, s2, _eps) + + x0_s2 = x0_fn(x_s2, s2) + eps_s2 = batch_mul(1.0 / s2, x_s2 - x0_s2) + + avg_eps = b1 * eps_s + b2 * eps_s1 + b3 * eps_s2 + return reg_eps_euler_step(x_s, s, t, avg_eps) + + +# key : order + name +RK_FNs = { + "1euler": rk1_euler, + "2mid": rk2_mid, + "2mid_stable": rk2_mid_stable, + "2heun_edm": rk_2heun_edm, + "2heun_naive": rk_2heun_naive, + "3kutta_naive": rk_3kutta_naive, +} + + +def get_runge_kutta_fn(name: str) -> Callable: + """ + Get the specified Runge-Kutta function. + + Args: + name: Name of the Runge-Kutta method. + + Returns: + Callable: The specified Runge-Kutta function. + + Raises: + RuntimeError: If the specified method is not supported. + """ + if name in RK_FNs: + return RK_FNs[name] + methods = "\n\t".join(RK_FNs.keys()) + raise RuntimeError(f"Only support the following Runge-Kutta methods:\n\t{methods}") + + +def is_runge_kutta_fn_supported(name: str) -> bool: + """ + Check if the specified Runge-Kutta function is supported. + + Args: + name: Name of the Runge-Kutta method. + + Returns: + bool: True if the method is supported, False otherwise. + """ + return name in RK_FNs diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..52acf51249366289769e95b511f0b38d08233943 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/__init__.py @@ -0,0 +1,65 @@ +# Copyright (c) Facebook, Inc. and its affiliates. +import os + +from omegaconf import DictConfig, OmegaConf + +from cosmos_policy._src.imaginaire.lazy_config.instantiate import instantiate +from cosmos_policy._src.imaginaire.lazy_config.lazy import LazyCall, LazyConfig +from cosmos_policy._src.imaginaire.lazy_config.omegaconf_patch import to_object + +OmegaConf.to_object = to_object + +PLACEHOLDER = None + + +class LazyDict(DictConfig): # NOTE: to differentiate between LazyDict & DictConfig + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + + +__all__ = ["instantiate", "LazyCall", "LazyConfig", "PLACEHOLDER", "LazyDict"] + + +DOC_BUILDING = os.getenv("_DOC_BUILDING", False) # set in docs/conf.py + + +def fixup_module_metadata(module_name, namespace, keys=None): + """ + Fix the __qualname__ of module members to be their exported api name, so + when they are referenced in docs, sphinx can find them. Reference: + https://github.com/python-trio/trio/blob/6754c74eacfad9cc5c92d5c24727a2f3b620624e/trio/_util.py#L216-L241 + """ + if not DOC_BUILDING: + return + seen_ids = set() + + def fix_one(qualname, name, obj): + # avoid infinite recursion (relevant when using + # typing.Generic, for example) + if id(obj) in seen_ids: + return + seen_ids.add(id(obj)) + + mod = getattr(obj, "__module__", None) + if mod is not None and (mod.startswith(module_name) or mod.startswith("fvcore.")): + obj.__module__ = module_name + # Modules, unlike everything else in Python, put fully-qualitied + # names into their __name__ attribute. We check for "." to avoid + # rewriting these. + if hasattr(obj, "__name__") and "." not in obj.__name__: + obj.__name__ = name + obj.__qualname__ = qualname + if isinstance(obj, type): + for attr_name, attr_value in obj.__dict__.items(): + fix_one(objname + "." + attr_name, attr_name, attr_value) + + if keys is None: + keys = namespace.keys() + for objname in keys: + if not objname.startswith("_"): + obj = namespace[objname] + fix_one(objname, objname, obj) + + +fixup_module_metadata(__name__, globals(), __all__) +del fixup_module_metadata diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/file_io.py b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/file_io.py new file mode 100644 index 0000000000000000000000000000000000000000..0c6693f43e4d0d1f7ac84774f588c47059207b56 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/file_io.py @@ -0,0 +1,25 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +from iopath.common.file_io import HTTPURLHandler, OneDrivePathHandler, PathHandler +from iopath.common.file_io import PathManager as PathManagerBase + +__all__ = ["PathManager", "PathHandler"] + + +PathManager = PathManagerBase() +PathManager.register_handler(HTTPURLHandler()) +PathManager.register_handler(OneDrivePathHandler()) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/instantiate.py b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/instantiate.py new file mode 100644 index 0000000000000000000000000000000000000000..33e77635c200113675af1dab9277c2a4aea3085b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/instantiate.py @@ -0,0 +1,120 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import collections.abc as abc +import dataclasses +import logging +from typing import Any + +import attrs + +from cosmos_policy._src.imaginaire.lazy_config.registry import _convert_target_to_string, locate + +__all__ = ["dump_dataclass", "instantiate"] + + +def is_dataclass_or_attrs(target): + return dataclasses.is_dataclass(target) or attrs.has(target) + + +def dump_dataclass(obj: Any): + """ + Dump a dataclass recursively into a dict that can be later instantiated. + + Args: + obj: a dataclass object + + Returns: + dict + """ + assert dataclasses.is_dataclass(obj) and not isinstance(obj, type), ( + "dump_dataclass() requires an instance of a dataclass." + ) + ret = {"_target_": _convert_target_to_string(type(obj))} + for f in dataclasses.fields(obj): + v = getattr(obj, f.name) + if dataclasses.is_dataclass(v): + v = dump_dataclass(v) + if isinstance(v, (list, tuple)): + v = [dump_dataclass(x) if dataclasses.is_dataclass(x) else x for x in v] + ret[f.name] = v + return ret + + +def instantiate(cfg, *args, **kwargs): + """ + Recursively instantiate objects defined in dictionaries by + "_target_" and arguments. + + Args: + cfg: a dict-like object with "_target_" that defines the caller, and + other keys that define the arguments + args: Optional positional parameters pass-through. + kwargs: Optional named parameters pass-through. + + Returns: + object instantiated by cfg + """ + from omegaconf import DictConfig, ListConfig, OmegaConf + + if isinstance(cfg, ListConfig): + lst = [instantiate(x) for x in cfg] + return ListConfig(lst, flags={"allow_objects": True}) + if isinstance(cfg, list): + # Specialize for list, because many classes take + # list[objects] as arguments, such as ResNet, DatasetMapper + return [instantiate(x) for x in cfg] + + # If input is a DictConfig backed by dataclasses (i.e. omegaconf's structured config), + # instantiate it to the actual dataclass. + if isinstance(cfg, DictConfig) and is_dataclass_or_attrs(cfg._metadata.object_type): + return OmegaConf.to_object(cfg) + + if isinstance(cfg, abc.Mapping) and "_target_" in cfg: + # conceptually equivalent to hydra.utils.instantiate(cfg) with _convert_=all, + # but faster: https://github.com/facebookresearch/hydra/issues/1200 + is_recursive = getattr(cfg, "_recursive_", True) + if is_recursive: + cfg = {k: instantiate(v) for k, v in cfg.items()} + else: + cfg = {k: v for k, v in cfg.items()} + # pop the _recursive_ key to avoid passing it as a parameter + if "_recursive_" in cfg: + cfg.pop("_recursive_") + cls = cfg.pop("_target_") + cls = instantiate(cls) + + if isinstance(cls, str): + cls_name = cls + cls = locate(cls_name) + assert cls is not None, cls_name + else: + try: + cls_name = cls.__module__ + "." + cls.__qualname__ + except Exception: + # target could be anything, so the above could fail + cls_name = str(cls) + assert callable(cls), f"_target_ {cls} does not define a callable object" + try: + # override config with kwargs + instantiate_kwargs = {} + instantiate_kwargs.update(cfg) + instantiate_kwargs.update(kwargs) + return cls(*args, **instantiate_kwargs) + except TypeError: + logger = logging.getLogger(__name__) + logger.error(f"Error when instantiating {cls_name}!") + raise + return cfg # return as-is if don't know what to do diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/lazy.py b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/lazy.py new file mode 100644 index 0000000000000000000000000000000000000000..7b60ea93e0ee571967aa3db23923bb4a67f03d74 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/lazy.py @@ -0,0 +1,430 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import ast +import builtins +import collections.abc as abc +import importlib +import inspect +import logging +import os +import pickle +import uuid +from collections import OrderedDict +from contextlib import contextmanager +from copy import deepcopy +from dataclasses import is_dataclass +from typing import Any, Dict, List, Tuple, Union + +import attrs +import yaml +from omegaconf import DictConfig, ListConfig, OmegaConf + +try: + import dill as dill_pickle +except ImportError: + dill_pickle = None + +try: + import cloudpickle +except ImportError: + cloudpickle = None + +from cosmos_policy._src.imaginaire.lazy_config.file_io import PathManager +from cosmos_policy._src.imaginaire.lazy_config.registry import _convert_target_to_string + +__all__ = ["LazyCall", "LazyConfig"] + + +def sort_dict(d: Dict[str, Any]) -> OrderedDict[str, Any]: + return OrderedDict(sorted(d.items(), key=lambda x: x[0])) + + +def dict_representer(dumper: yaml.Dumper, data: OrderedDict[str, Any]) -> yaml.nodes.MappingNode: + return dumper.represent_mapping("tag:yaml.org,2002:map", data.items()) + + +def sort_recursive(obj: Union[Dict[str, Any], List[Any], Any]) -> Union[OrderedDict[str, Any], List[Any], Any]: + if isinstance(obj, dict): + return sort_dict({k: sort_recursive(v) for k, v in obj.items()}) + elif isinstance(obj, list): + return [sort_recursive(item) for item in obj] + return obj + + +yaml.add_representer(OrderedDict, dict_representer) + +OmegaConf.register_new_resolver("add", lambda *vals: sum(vals)) +OmegaConf.register_new_resolver("subtract", lambda *vals: vals[0] - sum(vals[1:])) + + +def get_default_params(cls_or_func): + if callable(cls_or_func): + # inspect signature for function + signature = inspect.signature(cls_or_func) + else: + # inspect signature for class + signature = inspect.signature(cls_or_func.__init__) + params = signature.parameters + default_params = { + name: param.default for name, param in params.items() if param.default is not inspect.Parameter.empty + } + return default_params + + +class LazyCall: + """ + Wrap a callable so that when it's called, the call will not be executed, + but returns a dict that describes the call. + + LazyCall object has to be called with only keyword arguments. Positional + arguments are not yet supported. + + Examples: + :: + from detectron2.config import instantiate, LazyCall + + layer_cfg = LazyCall(nn.Conv2d)(in_channels=32, out_channels=32) + layer_cfg.out_channels = 64 # can edit it afterwards + layer = instantiate(layer_cfg) + """ + + def __init__(self, target): + if not (callable(target) or isinstance(target, (str, abc.Mapping))): + raise TypeError(f"target of LazyCall must be a callable or defines a callable! Got {target}") + self._target = target + + def __call__(self, **kwargs): + if is_dataclass(self._target) or attrs.has(self._target): + # omegaconf object cannot hold dataclass type + # https://github.com/omry/omegaconf/issues/784 + target = _convert_target_to_string(self._target) + else: + target = self._target + kwargs["_target_"] = target + + _final_params = get_default_params(self._target) + _final_params.update(kwargs) + + return DictConfig(content=_final_params, flags={"allow_objects": True}) + + +def _visit_dict_config(cfg, func): + """ + Apply func recursively to all DictConfig in cfg. + """ + if isinstance(cfg, DictConfig): + func(cfg) + for v in cfg.values(): + _visit_dict_config(v, func) + elif isinstance(cfg, ListConfig): + for v in cfg: + _visit_dict_config(v, func) + + +def _validate_py_syntax(filename): + # see also https://github.com/open-mmlab/mmcv/blob/master/mmcv/utils/config.py + with PathManager.open(filename, "r") as f: + content = f.read() + try: + ast.parse(content) + except SyntaxError as e: + raise SyntaxError(f"Config file {filename} has syntax error!") from e + + +def _cast_to_config(obj): + # if given a dict, return DictConfig instead + if isinstance(obj, dict): + return DictConfig(obj, flags={"allow_objects": True}) + return obj + + +_CFG_PACKAGE_NAME = "detectron2._cfg_loader" +""" +A namespace to put all imported config into. +""" + + +def _random_package_name(filename): + # generate a random package name when loading config files + return _CFG_PACKAGE_NAME + str(uuid.uuid4())[:4] + "." + os.path.basename(filename) + + +@contextmanager +def _patch_import(): + """ + Enhance relative import statements in config files, so that they: + 1. locate files purely based on relative location, regardless of packages. + e.g. you can import file without having __init__ + 2. do not cache modules globally; modifications of module states has no side effect + 3. support other storage system through PathManager, so config files can be in the cloud + 4. imported dict are turned into omegaconf.DictConfig automatically + """ + old_import = builtins.__import__ + + def find_relative_file(original_file, relative_import_path, level): + # NOTE: "from . import x" is not handled. Because then it's unclear + # if such import should produce `x` as a python module or DictConfig. + # This can be discussed further if needed. + relative_import_err = """ +Relative import of directories is not allowed within config files. +Within a config file, relative import can only import other config files. +""".replace("\n", " ") + if not len(relative_import_path): + raise ImportError(relative_import_err) + + cur_file = os.path.dirname(original_file) + for _ in range(level - 1): + cur_file = os.path.dirname(cur_file) + cur_name = relative_import_path.lstrip(".") + for part in cur_name.split("."): + cur_file = os.path.join(cur_file, part) + if not cur_file.endswith(".py"): + cur_file += ".py" + if not PathManager.isfile(cur_file): + cur_file_no_suffix = cur_file[: -len(".py")] + if PathManager.isdir(cur_file_no_suffix): + raise ImportError(f"Cannot import from {cur_file_no_suffix}." + relative_import_err) + else: + raise ImportError( + f"Cannot import name {relative_import_path} from {original_file}: {cur_file} does not exist." + ) + return cur_file + + def new_import(name, globals=None, locals=None, fromlist=(), level=0): + if ( + # Only deal with relative imports inside config files + level != 0 and globals is not None and (globals.get("__package__", "") or "").startswith(_CFG_PACKAGE_NAME) + ): + cur_file = find_relative_file(globals["__file__"], name, level) + _validate_py_syntax(cur_file) + spec = importlib.machinery.ModuleSpec(_random_package_name(cur_file), None, origin=cur_file) + module = importlib.util.module_from_spec(spec) + module.__file__ = cur_file + with PathManager.open(cur_file) as f: + content = f.read() + exec(compile(content, cur_file, "exec"), module.__dict__) + for name in fromlist: # turn imported dict into DictConfig automatically + val = _cast_to_config(module.__dict__[name]) + module.__dict__[name] = val + return module + return old_import(name, globals, locals, fromlist=fromlist, level=level) + + builtins.__import__ = new_import + yield new_import + builtins.__import__ = old_import + + +class LazyConfig: + """ + Provide methods to save, load, and overrides an omegaconf config object + which may contain definition of lazily-constructed objects. + """ + + @staticmethod + def load_rel(filename: str, keys: Union[None, str, Tuple[str, ...]] = None): + """ + Similar to :meth:`load()`, but load path relative to the caller's + source file. + + This has the same functionality as a relative import, except that this method + accepts filename as a string, so more characters are allowed in the filename. + """ + caller_frame = inspect.stack()[1] + caller_fname = caller_frame[0].f_code.co_filename + assert caller_fname != "", "load_rel Unable to find caller" + caller_dir = os.path.dirname(caller_fname) + filename = os.path.join(caller_dir, filename) + return LazyConfig.load(filename, keys) + + @staticmethod + def load(filename: str, keys: Union[None, str, Tuple[str, ...]] = None): + """ + Load a config file. + + Args: + filename: absolute path or relative path w.r.t. the current working directory + keys: keys to load and return. If not given, return all keys + (whose values are config objects) in a dict. + """ + has_keys = keys is not None + filename = filename.replace("/./", "/") # redundant + if os.path.splitext(filename)[1] not in [".py", ".yaml", ".yml"]: + raise ValueError(f"Config file {filename} has to be a python or yaml file.") + if filename.endswith(".py"): + _validate_py_syntax(filename) + + with _patch_import(): + # Record the filename + module_namespace = { + "__file__": filename, + "__package__": _random_package_name(filename), + } + with PathManager.open(filename) as f: + content = f.read() + # Compile first with filename to: + # 1. make filename appears in stacktrace + # 2. make load_rel able to find its parent's (possibly remote) location + exec(compile(content, filename, "exec"), module_namespace) + + ret = module_namespace + else: + with PathManager.open(filename) as f: + obj = yaml.unsafe_load(f) + ret = OmegaConf.create(obj, flags={"allow_objects": True}) + + if has_keys: + if isinstance(keys, str): + return _cast_to_config(ret[keys]) + else: + return tuple(_cast_to_config(ret[a]) for a in keys) + else: + if filename.endswith(".py"): + # when not specified, only load those that are config objects + ret = DictConfig( + { + name: _cast_to_config(value) + for name, value in ret.items() + if isinstance(value, (DictConfig, ListConfig, dict)) and not name.startswith("_") + }, + flags={"allow_objects": True}, + ) + return ret + + @staticmethod + def save_pkl(cfg, filename: str) -> str: + """ + Saves a Config object to a file using pickle serialization. This method is typically used + when the configuration object contains complex objects, such as lambdas, that are not supported by + simpler serialization methods like YAML. The function attempts to create a deep copy of the configuration + object before serialization to ensure that the original object remains unmodified. + + Args: + cfg: A Config object to be serialized and saved. + filename: The path and name of the file where the configuration should be saved. The function + assumes the file extension indicates a pickle format (e.g., .pkl). + + Returns: + str: The filename to which the configuration was saved. This can be used to verify the file location + or log the outcome. + + Notes: + - The function logs a warning if the configuration is successfully saved using pickle. + - If saving fails, an error is logged with the exception details. + """ + logger = logging.getLogger(__name__) + try: + cfg = deepcopy(cfg) + except Exception: + pass + + try: + with PathManager.open(filename, "wb") as f: + pickle.dump(cfg, f) + logger.warning(f"Config is saved using pickle at {filename}.") + except Exception as e: + logger.error(f"Failed to save config to {filename}: {e}. Trying dill or cloudpickle instead") + if dill_pickle: + try: + with PathManager.open(filename, "wb") as f: + pickle.dump(dill_pickle.dumps(cfg, recurse=True), f) + logger.warning(f"Config is saved using dill at {filename}.") + except Exception as e: + logger.error(f"Failed to save config to {filename}: {e}.") + if cloudpickle: + try: + with PathManager.open(filename, "wb") as f: + pickle.dump(cloudpickle.dumps(cfg), f) + logger.warning(f"Config is saved using cloudpickle at {filename}.") + except Exception as e: + logger.error(f"Failed to save config to {filename}: {e}.") + else: + logger.error("cloudpickle is not available. Cannot save the config.") + raise e + + return filename + + @staticmethod + def save_yaml(cfg, filename: str) -> str: + """ + Saves a Config object to a file using YAML serialization. This method is beneficial when the configuration object's content needs to be human-readable and easily editable. YAML is suitable for configurations that do not contain complex types like lambdas, which must be handled differently. The function converts unserializable items to strings before saving to ensure compatibility with YAML serialization. + + Args: + cfg: A Config object to be serialized and saved. It handles both DictConfig and ListConfig types. + filename: The path and name of the file where the configuration should be saved. The function does not require a specific file extension but typically uses '.yaml'. + + Returns: + str: The filename to which the configuration was saved. This can be used to verify the file location or log the outcome. + + Notes: + - The function logs a warning if the configuration is successfully saved using YAML. + - If saving fails, an error is logged with the exception details. + """ + logger = logging.getLogger(__name__) + try: + cfg = deepcopy(cfg) + except Exception: + pass + + # Define a function to check if an item is serializable to YAML + def is_serializable(item): + try: + OmegaConf.to_yaml(item) + return True + except Exception: + return False + + # Function to convert unserializable items to strings + def serialize_config(config): + if isinstance(config, DictConfig): + for key, value in config.items(): + if isinstance(value, (DictConfig, ListConfig)): + try: + if "_target_" in value: + default_params = get_default_params(value["_target_"]) + for default_key, default_v in default_params.items(): + if default_key not in value: + value[default_key] = default_v + except Exception as e: + logger.error(f"Failed to add default argument values: {e}") + + serialize_config(value) + else: + if not is_serializable(value) and value is not None: + config[key] = str(value) + elif isinstance(config, ListConfig): + for i, item in enumerate(config): + if isinstance(item, (DictConfig, ListConfig)): + serialize_config(item) + else: + if not is_serializable(item) and item is not None: + config[i] = str(item) + else: + raise NotImplementedError("Input config must be a DictConfig or ListConfig.") + return config + + # Convert Config object to a DictConfig object. + config_dict = attrs.asdict(cfg) + config_omegaconf = DictConfig(content=config_dict, flags={"allow_objects": True}) + + # Serialize the DictConfig object by converting non-serializable objects to strings. + config_omegaconf = serialize_config(config_omegaconf) + + config_dict: Dict[str, Any] = OmegaConf.to_container(config_omegaconf, resolve=True) + sorted_config: OrderedDict[str, Any] = sort_recursive(config_dict) + with open(filename, "w") as f: + yaml.dump(sorted_config, f, default_flow_style=False) + logger.warning(f"Config is saved using omegaconf at {filename}.") + return filename diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/omegaconf_patch.py b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/omegaconf_patch.py new file mode 100644 index 0000000000000000000000000000000000000000..39dca42a0a71383de919b750cedf2606faae206d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/omegaconf_patch.py @@ -0,0 +1,65 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any, Dict, List, Union + +from omegaconf import OmegaConf +from omegaconf.base import DictKeyType, SCMode +from omegaconf.dictconfig import DictConfig # pragma: no cover + + +def to_object(cfg: Any) -> Union[Dict[DictKeyType, Any], List[Any], None, str, Any]: + """ + Converts an OmegaConf configuration object to a native Python container (dict or list), unless + the configuration is specifically created by LazyCall, in which case the original configuration + is returned directly. + + This function serves as a modification of the original `to_object` method from OmegaConf, + preventing DictConfig objects created by LazyCall from being automatically converted to Python + dictionaries. This ensures that configurations meant to be lazily evaluated retain their intended + structure and behavior. + + Differences from OmegaConf's original `to_object`: + - Adds a check at the beginning to return the configuration unchanged if it is created by LazyCall. + + Reference: + - Original OmegaConf `to_object` method: https://github.com/omry/omegaconf/blob/master/omegaconf/omegaconf.py#L595 + + Args: + cfg (Any): The OmegaConf configuration object to convert. + + Returns: + Union[Dict[DictKeyType, Any], List[Any], None, str, Any]: The converted Python container if + `cfg` is not a LazyCall created configuration, otherwise the unchanged `cfg`. + + Examples: + >>> cfg = DictConfig({"key": "value", "_target_": "Model"}) + >>> to_object(cfg) + DictConfig({"key": "value", "_target_": "Model"}) + + >>> cfg = DictConfig({"list": [1, 2, 3]}) + >>> to_object(cfg) + {'list': [1, 2, 3]} + """ + if isinstance(cfg, DictConfig) and "_target_" in cfg.keys(): + return cfg + + return OmegaConf.to_container( + cfg=cfg, + resolve=True, + throw_on_missing=True, + enum_to_str=False, + structured_config_mode=SCMode.INSTANTIATE, + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/registry.py b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/registry.py new file mode 100644 index 0000000000000000000000000000000000000000..f58085e67b8ca749e85126ac2f516f55de9bfed7 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/lazy_config/registry.py @@ -0,0 +1,75 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import pydoc +from typing import Any + +from fvcore.common.registry import Registry # for backward compatibility. + +""" +``Registry`` and `locate` provide ways to map a string (typically found +in config files) to callable objects. +""" + +__all__ = ["Registry", "locate"] + + +def _convert_target_to_string(t: Any) -> str: + """ + Inverse of ``locate()``. + + Args: + t: any object with ``__module__`` and ``__qualname__`` + """ + module, qualname = t.__module__, t.__qualname__ + + # Compress the path to this object, e.g. ``module.submodule._impl.class`` + # may become ``module.submodule.class``, if the later also resolves to the same + # object. This simplifies the string, and also is less affected by moving the + # class implementation. + module_parts = module.split(".") + for k in range(1, len(module_parts)): + prefix = ".".join(module_parts[:k]) + candidate = f"{prefix}.{qualname}" + try: + if locate(candidate) is t: + return candidate + except ImportError: + pass + return f"{module}.{qualname}" + + +def locate(name: str) -> Any: + """ + Locate and return an object ``x`` using an input string ``{x.__module__}.{x.__qualname__}``, + such as "module.submodule.class_name". + + Raise Exception if it cannot be found. + """ + obj = pydoc.locate(name) + + # Some cases (e.g. torch.optim.sgd.SGD) not handled correctly + # by pydoc.locate. Try a private function from hydra. + if obj is None: + try: + # from hydra.utils import get_method - will print many errors + + from hydra.utils import _locate + except ImportError as e: + raise ImportError(f"Cannot dynamically locate object {name}!") from e + else: + obj = _locate(name) # it raises if fails + + return obj diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/model.py b/REGEN-main/cosmos_policy/_src/imaginaire/model.py new file mode 100644 index 0000000000000000000000000000000000000000..9ceb077b9cded61aab3229c283b43c1a38292db6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/model.py @@ -0,0 +1,129 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any + +import torch + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict, instantiate + + +class ImaginaireModel(torch.nn.Module): + """The base model class of Imaginaire. It is inherited from torch.nn.Module. + + All models in Imaginaire should inherit ImaginaireModel. It should include the implementions for all the + computation graphs. All inheriting child classes should implement the following methods: + - training_step(): The training step of the model, including the loss computation. + - validation_step(): The validation step of the model, including the loss computation. + - forward(): The computation graph for model inference. + The following methods have default implementations in ImaginaireModel: + - init_optimizer_scheduler(): Creates the optimizer and scheduler for the model. + """ + + def __init__(self) -> None: + super().__init__() + + def init_optimizer_scheduler( + self, optimizer_config: LazyDict, scheduler_config: LazyDict + ) -> tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LRScheduler]: + """Creates the optimizer and scheduler for the model. + + Args: + config_model (ModelConfig): The config object for the model. + + Returns: + optimizer (torch.optim.Optimizer): The model optimizer. + scheduler (torch.optim.lr_scheduler.LRScheduler): The optimization scheduler. + """ + optimizer_config.params = self.parameters() + optimizer = instantiate(optimizer_config) + scheduler_config.optimizer = optimizer + scheduler = instantiate(scheduler_config) + return optimizer, scheduler + + def training_step( + self, data_batch: dict[str, torch.Tensor], iteration: int + ) -> tuple[dict[str, torch.Tensor], torch.Tensor]: + """The training step of the model, including the loss computation. + + Args: + data (dict[str, torch.Tensor]): Data batch (dictionary of tensors). + iteration (int): Current iteration number. + + Returns: + output_batch (dict[str, torch.Tensor]): Auxiliary model output from the training batch. + loss (torch.Tensor): The total loss for backprop (weighted sum of various losses). + """ + raise NotImplementedError + + @torch.no_grad() + def validation_step( + self, data_batch: dict[str, torch.Tensor], iteration: int + ) -> tuple[dict[str, torch.Tensor], torch.Tensor]: + """The validation step of the model, including the loss computation. + + Args: + data (dict[str, torch.Tensor]): Data batch (dictionary of tensors). + iteration (int): Current iteration number. + + Returns: + output_batch (dict[str, torch.Tensor]): Auxiliary model output from the validation batch. + loss (torch.Tensor): The total loss (weighted sum of various losses). + """ + raise NotImplementedError + + @torch.inference_mode() + def forward(self, *args: Any, **kwargs: Any) -> Any: + """The computation graph for model inference. + + Args: + *args: Whatever you decide to pass into the forward method. + **kwargs: Keyword arguments are also possible. + + Return: + Your model's output. + """ + raise NotImplementedError + + def on_train_start(self, memory_format: torch.memory_format = torch.preserve_format) -> None: + """The model preparation before the training is launched + + Args: + memory_format (torch.memory_format): Memory format of the model. + """ + pass + + def on_before_zero_grad( + self, optimizer: torch.optim.Optimizer, scheduler: torch.optim.lr_scheduler.LRScheduler, iteration: int + ) -> None: + """Hook before zero_grad() is called. + + Args: + optimizer (torch.optim.Optimizer): The model optimizer. + scheduler (torch.optim.lr_scheduler.LRScheduler): The optimization scheduler. + iteration (int): Current iteration number. + """ + pass + + def on_after_backward(self, iteration: int = 0) -> None: + """Hook after loss.backward() is called. + + This method is called immediately after the backward pass, allowing for custom operations + or modifications to be performed on the gradients before the optimizer step. + + Args: + iteration (int): Current iteration number. + """ + pass diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/models/abstract_emb_model.py b/REGEN-main/cosmos_policy/_src/imaginaire/models/abstract_emb_model.py new file mode 100644 index 0000000000000000000000000000000000000000..c3dca78caa345634a269a9d4a38c9f6374703e6a --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/models/abstract_emb_model.py @@ -0,0 +1,104 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +from typing import Optional, Union + +import torch +import torch.nn as nn + +from cosmos_policy._src.imaginaire.functional.batch_ops import batch_mul +from cosmos_policy._src.imaginaire.utils.count_params import count_params + + +class AbstractEmbModel(nn.Module): + def __init__(self) -> None: + super().__init__() + + self._is_trainable = None + self._dropout_rate = None + self._input_key = None + self._return_dict = False + + @property + def is_trainable(self) -> bool: + return self._is_trainable + + @property + def dropout_rate(self) -> Union[float, torch.Tensor]: + return self._dropout_rate + + @property + def input_key(self) -> str: + return self._input_key + + @property + def is_return_dict(self) -> bool: + return self._return_dict + + @is_trainable.setter + def is_trainable(self, value: bool) -> None: + self._is_trainable = value + + @dropout_rate.setter + def dropout_rate(self, value: Union[float, torch.Tensor]) -> None: + self._dropout_rate = value + + @input_key.setter + def input_key(self, value: str) -> None: + self._input_key = value + + @is_return_dict.setter + def is_return_dict(self, value: bool) -> None: + self._return_dict = value + + @is_trainable.deleter + def is_trainable(self) -> None: + del self._is_trainable + + @dropout_rate.deleter + def dropout_rate(self) -> None: + del self._dropout_rate + + @input_key.deleter + def input_key(self) -> None: + del self._input_key + + @is_return_dict.deleter + def is_return_dict(self) -> None: + del self._return_dict + + def random_dropout_input( + self, in_tensor: torch.Tensor, dropout_rate: Optional[float] = None, key: Optional[str] = None + ) -> torch.Tensor: + del key + dropout_rate = dropout_rate if dropout_rate is not None else self.dropout_rate + return batch_mul( + torch.bernoulli((1.0 - dropout_rate) * torch.ones(in_tensor.shape[0])).type_as(in_tensor), + in_tensor, + ) + + def details(self) -> str: + return "" + + def summary(self) -> str: + input_key = self.input_key if self.input_key is not None else getattr(self, "input_keys", None) + return ( + f"{self.__class__.__name__} \n\tinput key: {input_key}" + f"\n\tParam count: {count_params(self, False)} \n\tTrainable: {self.is_trainable}" + f"\n\tDropout rate: {self.dropout_rate}" + f"\n\t{self.details()}" + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/camera.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/camera.py new file mode 100644 index 0000000000000000000000000000000000000000..d7754b711e59bf3cbe1a388998d0d88edf56aa97 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/camera.py @@ -0,0 +1,660 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import numpy as np +import torch + + +def _recursive_to_numpy(x): + if isinstance(x, torch.Tensor): + return x.detach().cpu().numpy() + if isinstance(x, (list, tuple)): + return type(x)(_recursive_to_numpy(v) for v in x) + if isinstance(x, dict): + return {k: _recursive_to_numpy(v) for k, v in x.items()} + return x + + +def supports_numpy(arg_names, use_no_grad: bool = True): + """Decorator to transparently support numpy inputs. + + - Converts the specified named args from numpy arrays to torch tensors on entry + - Runs the wrapped function (optionally under no_grad) + - Converts returns back to numpy iff the FIRST targeted arg was a numpy array + """ + import functools + import inspect + + def decorator(fn): + sig = inspect.signature(fn) + + @functools.wraps(fn) + def wrapper(*args, **kwargs): + bound = sig.bind(*args, **kwargs) + bound.apply_defaults() + first_is_numpy = False + for idx, name in enumerate(arg_names): + if name in bound.arguments: + val = bound.arguments[name] + # Handle direct ndarray + if isinstance(val, np.ndarray): + if idx == 0: + first_is_numpy = True + bound.arguments[name] = torch.from_numpy(val) + continue + # Handle list/tuple of ndarrays -> list/tuple of tensors + if isinstance(val, (list, tuple)): + saw_numpy_first_elem = False + converted = [] + for i, el in enumerate(val): + if isinstance(el, np.ndarray): + if i == 0: + saw_numpy_first_elem = True + converted.append(torch.from_numpy(el)) + else: + converted.append(el) + if idx == 0 and saw_numpy_first_elem: + first_is_numpy = True + bound.arguments[name] = type(val)(converted) + + ctx = torch.no_grad() if use_no_grad else torch.enable_grad() + with ctx: + out = fn(*bound.args, **bound.kwargs) + return _recursive_to_numpy(out) if first_is_numpy else out + + return wrapper + + return decorator + + +class Camera: + """A class with a collection of common ops for camera transformations (Pytorch tensors). + + All poses are expected to have shape [...,3,4], where (...) indicates batch sizes of various ranks. + The last two dimensions (of size (3,4)) correspond to the extrinsic matrix [R|t] in OpenCV format. + + Convention: cam_pose is always a world-to-camera transform (world2cam): x_cam = R @ x_world + t. + This module operates on row-vector points with homogeneous coordinates on the right, so we apply + transformations as: points_hom @ cam_pose^T. + """ + + @staticmethod + @supports_numpy(["cam_pose"], use_no_grad=True) + def _check_valid_pose(cam_pose: torch.Tensor | np.ndarray) -> None: + """Checks whether the input tensor is a valid camera pose. + + Args: + cam_pose (torch.Tensor [...,3,4]): Input camera pose in world2cam [R|t] (OpenCV) format. + """ + assert cam_pose.shape[-2:] == (3, 4), "Camera pose is not of shape (3,4)." + R = cam_pose[..., :3] + # Compute determinant in float32 for numerical stability and allow dtype-dependent tolerance. + det_R = torch.linalg.det(R.to(torch.float32)) + one = torch.tensor(1.0, dtype=torch.float32, device=cam_pose.device) + if cam_pose.dtype in (torch.bfloat16, torch.float16): + rtol, atol = 1e-2, 1e-2 + else: + rtol, atol = 1e-4, 1e-6 + finite = bool(torch.isfinite(det_R).all()) + close = torch.allclose(det_R, one, rtol=rtol, atol=atol) + assert finite and close, ( + f"Rotation component in camera pose is invalid (det != 1 within tol). " + f"dtype={cam_pose.dtype}, rtol={rtol}, atol={atol}, det_mean={det_R.mean().item():.6f}" + ) + + @staticmethod + @supports_numpy(["cam_pose"], use_no_grad=True) + def invert_pose(cam_pose: torch.Tensor | np.ndarray) -> torch.Tensor | np.ndarray: + """Invert a camera pose. + + Args: + cam_pose (torch.Tensor/np.ndarray [...,3,4]): Input camera pose (world2cam [R|t]). + + Returns: + cam_pose_inv (torch.Tensor/np.ndarray [...,3,4]): The inverted camera pose (cam2world [R|t]). + """ + Camera._check_valid_pose(cam_pose) + in_dtype = cam_pose.dtype if isinstance(cam_pose, torch.Tensor) else torch.float32 + R, t = cam_pose[..., :3], cam_pose[..., 3:] + # Compute in float32 for numerical stability, cast back at the end + R32 = R.to(torch.float32) + t32 = t.to(torch.float32) + # For rotation matrices, inverse equals transpose; prefer transpose for stability and speed + R_inv32 = R32.transpose(-1, -2) + t_inv32 = -R_inv32 @ t32 + cam_pose_inv32 = torch.cat([R_inv32, t_inv32], dim=-1) + return cam_pose_inv32.to(in_dtype) + + @staticmethod + @supports_numpy(["cam_poses"], use_no_grad=True) + def compose_poses(cam_poses: list[torch.Tensor | np.ndarray]) -> torch.Tensor | np.ndarray: + """Compose a sequence of camera transformations together. + + pose_new = compose_poses([pose_1, pose_2, ... pose_N]) + pose_new(x) = pose_N o ... o pose_2 o pose_1(x) + + Args: + cam_poses (list[torch.Tensor/np.ndarray [...,3,4]]): Sequence of rigid transforms [R|t]. + When used as camera extrinsics in this module, each pose is assumed to be world2cam. + The composition follows the same row-vector convention: points_hom @ pose^T. + List items may be numpy arrays; they will be converted to torch internally. + + Returns: + cam_pose_new (torch.Tensor/np.ndarray [...,3,4]): The composed transformation [R|t]. + """ + cam_pose_new = cam_poses[0] + Camera._check_valid_pose(cam_pose_new) + out_dtype = cam_pose_new.dtype if isinstance(cam_pose_new, torch.Tensor) else torch.float32 + R_new, t_new = cam_pose_new[..., :3].to(torch.float32), cam_pose_new[..., 3:].to(torch.float32) + for cam_pose in cam_poses[1:]: + Camera._check_valid_pose(cam_pose) + # pose_new(x) = pose o pose_new(x) + R, t = cam_pose[..., :3].to(torch.float32), cam_pose[..., 3:].to(torch.float32) + R_new = R @ R_new + t_new = R @ t_new + t + cam_pose_new32 = torch.cat([R_new, t_new], dim=-1) + return cam_pose_new32.to(out_dtype) + + @staticmethod + @supports_numpy(["cam_pose", "cam_intr"], use_no_grad=True) + def get_camera_rays( + cam_pose: torch.Tensor | np.ndarray, + cam_intr: torch.Tensor | np.ndarray, + image_size: tuple[int, int], + ) -> torch.Tensor | np.ndarray: + """Get unit-norm camera rays in world coordinates for each pixel center. + + Args: + cam_pose (torch.Tensor/np.ndarray [...,3,4]): Camera pose (world2cam [R|t]). + cam_intr (torch.Tensor/np.ndarray [...,3,3]): Camera intrinsics. + image_size (Tuple[int, int]): Image size (height, width). + + Returns: + rays_world (torch.Tensor/np.ndarray [...,HW,3]): Unit direction rays from camera center through pixel centers, flattened over pixels. + """ + H, W = image_size + with torch.no_grad(): + # Compute image coordinate grid (in float32 for stability). + y_range = torch.arange(H, dtype=torch.float32, device=cam_pose.device).add_(0.5) + x_range = torch.arange(W, dtype=torch.float32, device=cam_pose.device).add_(0.5) + y_grid, x_grid = torch.meshgrid(y_range, x_range, indexing="ij") # [H,W] + xy_grid = torch.stack([x_grid, y_grid], dim=-1).view(-1, 2) # [HW,2] + xy_grid = xy_grid.repeat(*cam_pose.shape[:-2], 1, 1) # [...,HW,2] + # Pixel centers in camera coordinates at depth 1 (flattened HW) + grid_camera = Camera.image2camera(Camera.to_homogeneous(xy_grid), cam_intr) # [...,HW,3] + # Transform sample points and center to world + grid_world = Camera.camera2world(grid_camera, cam_pose) # [...,HW,3] + center_world = Camera.get_camera_center(cam_pose).unsqueeze(-2).expand_as(grid_world) # [...,HW,3] + rays_world = grid_world - center_world # [...,HW,3] + # Normalize to unit vectors + eps = 1e-8 + if cam_pose.dtype in (torch.bfloat16, torch.float16): + eps = 1e-2 + norms32 = rays_world.to(torch.float32).norm(dim=-1, keepdim=True).clamp_min(eps) + rays_world = rays_world / norms32.to(rays_world.dtype) + # Cast back to input dtype for consistency + rays_world = rays_world.to(cam_pose.dtype) + # Keep flattened shape [...,HW,3] + return rays_world + + @staticmethod + @supports_numpy(["cam_pose", "cam_intr"], use_no_grad=True) + def get_plucker_rays( + cam_pose: torch.Tensor | np.ndarray, + cam_intr: torch.Tensor | np.ndarray, + image_size: tuple[int, int], + ) -> torch.Tensor | np.ndarray: + """Get Plücker coordinates (moment, direction) for each pixel center. + + Args: + cam_pose (torch.Tensor/np.ndarray [...,3,4]): Camera pose (world2cam [R|t]). + cam_intr (torch.Tensor/np.ndarray [...,3,3]): Camera intrinsics. + image_size (Tuple[int, int]): Image size (height, width). + + Returns: + plucker (torch.Tensor/np.ndarray [...,HW,6]): Plücker coordinates [m | d], where + d is a unit direction vector and m = o × d with o the camera center in world. + """ + H, W = image_size + rays_world = Camera.get_camera_rays(cam_pose, cam_intr, image_size) # [...,HW,3] + # Expand center to [...,HW,3] + center_hw = Camera.get_camera_center(cam_pose).unsqueeze(-2).expand_as(rays_world) + moment = torch.linalg.cross(center_hw, rays_world) # [...,HW,3] + plucker = torch.cat([moment, rays_world], dim=-1) # [...,HW,6] + return plucker + + @staticmethod + @supports_numpy(["cam_pose"], use_no_grad=True) + def get_relative_poses_wrt_frame0( + cam_pose: torch.Tensor | np.ndarray, + ) -> torch.Tensor | np.ndarray: + """Compute poses relative to the first frame (index 0). + + All poses are world-to-camera [R|t] with shape [...,3,4]. The returned poses are expressed + in the coordinate system of the first camera, so the first pose is identity [I|0]. For the + i-th pose: pose_rel_i = compose(pose_i, inverse(pose_ref)). + + Args: + cam_pose (torch.Tensor/np.ndarray [...,V,3,4]): World-to-camera extrinsics per view. + + Returns: + cam_pose_rel (torch.Tensor/np.ndarray [...,V,3,4]): Relative world-to-camera extrinsics in the first frame. + """ + # supports_numpy handles numpy + assert cam_pose.shape[-2:] == (3, 4), "cam_pose must have shape [..., V, 3, 4]." + # Reference pose and its inverse + pose_ref = cam_pose.select(dim=-3, index=0) # [...,3,4] + pose_ref_inv = Camera.invert_pose(pose_ref) # [...,3,4] + # Compose with broadcasting: pose_rel = pose ∘ pose_ref_inv + cam_pose_rel = Camera.compose_poses([pose_ref_inv, cam_pose]) + return cam_pose_rel + + @staticmethod + @supports_numpy(["cam_pose"], use_no_grad=True) + def get_camera_center(cam_pose: torch.Tensor | np.ndarray) -> torch.Tensor | np.ndarray: + """Get the camera center in world coordinates for a given world2cam pose. + + Args: + cam_pose (torch.Tensor/np.ndarray [...,3,4]): Camera pose (world2cam [R|t]). + + Returns: + center_world (torch.Tensor/np.ndarray [...,3]): Camera center in world coordinates. + """ + Camera._check_valid_pose(cam_pose) + R, t = cam_pose[..., :3], cam_pose[..., 3:] # [...,3,3], [...,3,1] + center_world32 = (-R.to(torch.float32).transpose(-1, -2) @ t.to(torch.float32)).squeeze(-1) + return center_world32.to(R.dtype) + + @staticmethod + @supports_numpy(["points"], use_no_grad=True) + def to_homogeneous(points: torch.Tensor | np.ndarray) -> torch.Tensor | np.ndarray: + """Get homogeneous coordinates of the input points. + + Args: + points (torch.Tensor/np.ndarray [...,K]): Input coordinates. + + Returns: + points_hom (torch.Tensor/np.ndarray [...,K+1]): Homogeneous coordinates. + """ + # Compute homogeneous coordinate in float32 for stability, then cast back + one32 = torch.ones_like( + points[..., :1], dtype=torch.float32, device=(points.device if isinstance(points, torch.Tensor) else None) + ) + points_hom = torch.cat([points, one32.to(points.dtype)], dim=-1) + return points_hom + + @staticmethod + @supports_numpy(["points", "cam_pose"], use_no_grad=True) + def world2camera( + points: torch.Tensor | np.ndarray, cam_pose: torch.Tensor | np.ndarray + ) -> torch.Tensor | np.ndarray: + """Given the camera pose, transform input 3D points from world coordinates to camera coordinates. + + Args: + points (torch.Tensor/np.ndarray [...,N,3]): Input 3D points. + cam_pose (torch.Tensor/np.ndarray [...,3,4]/[3,4]): (Batched) camera pose (world2cam [R|t]). + + Returns: + points_new (torch.Tensor/np.ndarray [...,N,3]): Transformed 3D points. + """ + points_hom = Camera.to_homogeneous(points).to(torch.float32) # [...,N,4] + points_new32 = points_hom @ cam_pose.to(torch.float32).transpose(-1, -2) # [...,N,3] + return points_new32.to(points.dtype) + + @staticmethod + @supports_numpy(["points", "cam_pose"], use_no_grad=True) + def camera2world( + points: torch.Tensor | np.ndarray, cam_pose: torch.Tensor | np.ndarray + ) -> torch.Tensor | np.ndarray: + """Given the camera pose, transform input 3D points from camera coordinates to world coordinates. + + Args: + points (torch.Tensor/np.ndarray [...,N,3]): Input 3D points. + cam_pose (torch.Tensor/np.ndarray [...,3,4]/[3,4]): (Batched) camera pose (world2cam [R|t]). + + Returns: + points_new (torch.Tensor/np.ndarray [...,N,3]): Transformed 3D points. + """ + points_hom = Camera.to_homogeneous(points).to(torch.float32) + pose_inv = Camera.invert_pose(cam_pose) + points_new32 = points_hom @ pose_inv.to(torch.float32).transpose(-1, -2) + # To reduce double-quantization error on low-precision dtypes (e.g., bf16 on CPU), + # keep high precision on output for transform back to world space. + if isinstance(points, torch.Tensor) and points.dtype in (torch.bfloat16, torch.float16): + return points_new32 + return points_new32.to(points.dtype) + + @staticmethod + @supports_numpy(["points", "cam_intr"], use_no_grad=True) + def camera2image( + points: torch.Tensor | np.ndarray, cam_intr: torch.Tensor | np.ndarray + ) -> torch.Tensor | np.ndarray: + """Given the camera intrinsics, calibrate input 3D points from camera frame to image (pixel) frame. + + Args: + points (torch.Tensor/np.ndarray [...,N,3]): Input 3D points. + cam_intr (torch.Tensor/np.ndarray [...,3,3]/[3,3]): (Batched) camera intrinsic matrix. + + Returns: + points_new (torch.Tensor/np.ndarray [...,N,3]): Transformed 3D points. + """ + points32 = points.to(torch.float32) + points_new32 = points32 @ cam_intr.to(torch.float32).transpose(-1, -2) + return points_new32.to(points.dtype) + + @staticmethod + @supports_numpy(["points", "cam_intr"], use_no_grad=True) + def image2camera( + points: torch.Tensor | np.ndarray, cam_intr: torch.Tensor | np.ndarray + ) -> torch.Tensor | np.ndarray: + """Given the camera intrinsics, calibrate input 3D points from image (pixel) frame to camera frame. + + Args: + points (torch.Tensor/np.ndarray [...,N,3]): Input 3D points. + cam_intr (torch.Tensor/np.ndarray [...,3,3]/[3,3]): (Batched) camera intrinsic matrix. + + Returns: + points_new (torch.Tensor/np.ndarray [...,N,3]): Transformed 3D points. + """ + K_inv32 = torch.linalg.inv(cam_intr.to(torch.float32)) + points32 = points.to(torch.float32) + points_new32 = points32 @ K_inv32.transpose(-1, -2) + return points_new32.to(points.dtype) + + @staticmethod + @supports_numpy(["params"], use_no_grad=True) + def intrinsic_params_to_matrices(params: torch.Tensor | np.ndarray) -> torch.Tensor | np.ndarray: + """Convert (fx, fy, cx, cy) parameters to camera intrinsic matrix/matrices. + + Args: + params (torch.Tensor/np.ndarray [...,4]): Intrinsic parameters (fx, fy, cx, cy). + + Returns: + K (torch.Tensor/np.ndarray [...,3,3]): Camera intrinsic matrices. + """ + assert params.shape[-1] == 4, "Intrinsic params must have shape (..., 4) for (fx, fy, cx, cy)." + fx, fy, cx, cy = params.unbind(dim=-1) + one = torch.ones_like(fx) + zero = torch.zeros_like(fx) + row0 = torch.stack([fx, zero, cx], dim=-1) + row1 = torch.stack([zero, fy, cy], dim=-1) + row2 = torch.stack([zero, zero, one], dim=-1) + K = torch.stack([row0, row1, row2], dim=-2) + return K + + @staticmethod + @supports_numpy(["cam_intr"], use_no_grad=True) + def intrinsic_matrices_to_params( + cam_intr: torch.Tensor | np.ndarray, atol: float = 1e-6 + ) -> torch.Tensor | np.ndarray: + """Extract (fx, fy, cx, cy) from camera intrinsic matrix/matrices. + + Args: + cam_intr (torch.Tensor/np.ndarray [...,3,3]): Camera intrinsic matrices. + atol (float): Tolerance when checking the bottom row against [0,0,1]. + + Returns: + params (torch.Tensor/np.ndarray [...,4]): Intrinsic parameters (fx, fy, cx, cy). + """ + assert cam_intr.shape[-2:] == (3, 3), "Intrinsic matrix must have shape (..., 3, 3)." + row32 = cam_intr[..., 2, :].to(torch.float32) + target32 = torch.tensor([0.0, 0.0, 1.0], dtype=torch.float32, device=cam_intr.device) + rtol = 1e-5 + atol_eff = atol + if cam_intr.dtype in (torch.bfloat16, torch.float16): + rtol = 1e-2 + atol_eff = max(atol, 1e-2) + if not torch.allclose(row32, target32, rtol=rtol, atol=atol_eff): + # Still proceed but warn via assertion message if strictness is desired. + pass + fx = cam_intr[..., 0, 0] + fy = cam_intr[..., 1, 1] + cx = cam_intr[..., 0, 2] + cy = cam_intr[..., 1, 2] + params = torch.stack([fx, fy, cx, cy], dim=-1) + return params + + @staticmethod + @supports_numpy(["qxyzw_t"], use_no_grad=True) + def extrinsic_params_to_matrices(qxyzw_t: torch.Tensor | np.ndarray) -> torch.Tensor | np.ndarray: + """Convert (x,y,z,w, tx,ty,tz) to world2cam extrinsic matrix/matrices [R|t]. + + Args: + qxyzw_t (torch.Tensor/np.ndarray [...,7]): Quaternion (xyzw) and translation stacked. + + Returns: + cam_pose (torch.Tensor/np.ndarray [...,3,4]): World-to-camera extrinsic [R|t]. + """ + assert qxyzw_t.shape[-1] == 7, "Input must have shape (..., 7) for (qx,qy,qz,qw,tx,ty,tz)." + q = qxyzw_t[..., :4] + t = qxyzw_t[..., 4:7] + # Enforce unit quaternion + Quaternion._check_valid_quaternion(q, require_normalized=True) + R = Quaternion.to_rotation_matrix(q) # [...,3,3] + cam_pose = torch.cat([R, t.unsqueeze(-1)], dim=-1) + return cam_pose + + @staticmethod + @supports_numpy(["cam_pose"], use_no_grad=True) + def extrinsic_matrices_to_params(cam_pose: torch.Tensor | np.ndarray) -> torch.Tensor | np.ndarray: + """Convert world2cam extrinsic matrix/matrices [R|t] to (x,y,z,w, tx,ty,tz). + + Args: + cam_pose (torch.Tensor/np.ndarray [...,3,4]): World-to-camera extrinsic [R|t]. + + Returns: + qxyzw_t (torch.Tensor/np.ndarray [...,7]): Quaternion (xyzw) and translation stacked. + """ + Camera._check_valid_pose(cam_pose) + R = cam_pose[..., :3] + t = cam_pose[..., 3:].squeeze(-1) + q = Quaternion.from_rotation_matrix(R) + qxyzw_t = torch.cat([q, t], dim=-1) + return qxyzw_t + + +class Quaternion: + """A collection of common quaternion operations (Pytorch tensors). + + Convention (STRICT): Quaternions are represented in (x, y, z, w) order (xyzw) and are unit-norm. + The last dimension must be size 4. + """ + + @staticmethod + @supports_numpy(["q"], use_no_grad=True) + def _check_valid_quaternion( + q: torch.Tensor | np.ndarray, require_normalized: bool = True, atol: float = 1e-5 + ) -> torch.Tensor | np.ndarray: + """Checks whether the input tensor is a valid quaternion. + + Args: + q (torch.Tensor [...,4]): Input quaternion(s) in (x, y, z, w) order. + require_normalized (bool): If True, assert unit-norm within atol. Defaults to True. + atol (float): Absolute tolerance for the unit-norm check. + """ + assert q.shape[-1] == 4, "Quaternion is not of shape (..., 4)." + if require_normalized: + norms32 = q.to(torch.float32).norm(dim=-1) + ones32 = torch.ones_like(norms32) + tol = max(atol, 1e-2) if q.dtype in (torch.bfloat16, torch.float16) else atol + assert torch.allclose(norms32, ones32, atol=tol), "Quaternion must be unit length." + return q + + @staticmethod + @supports_numpy(["q"], use_no_grad=True) + def normalize(q: torch.Tensor | np.ndarray, eps: float = 1e-8) -> torch.Tensor | np.ndarray: + """Normalize quaternion(s) to unit length. + + Args: + q (torch.Tensor [...,4]): Input quaternion(s). + eps (float): Small epsilon to avoid division by zero. + + Returns: + q_norm (torch.Tensor [...,4]): Unit quaternions. + """ + # Allow non-normalized input here, since this function normalizes + Quaternion._check_valid_quaternion(q, require_normalized=False) + eps_eff = eps + if q.dtype in (torch.bfloat16, torch.float16): + eps_eff = max(eps, 1e-2) + norm32 = q.to(torch.float32).norm(dim=-1, keepdim=True).clamp_min(eps_eff) + out32 = q.to(torch.float32) / norm32 + out = out32.to(q.dtype) + return out + + @staticmethod + @supports_numpy(["q"], use_no_grad=True) + def to_rotation_matrix(q: torch.Tensor | np.ndarray) -> torch.Tensor | np.ndarray: + """Convert quaternion(s) to rotation matrix/matrices. + + Args: + q (torch.Tensor [...,4]): Quaternion(s) (x, y, z, w). + + Returns: + R (torch.Tensor [...,3,3]): Rotation matrix/matrices. + """ + # Enforce unit quaternions for rotations + Quaternion._check_valid_quaternion(q, require_normalized=True) + q32 = q.to(torch.float32) + qx, qy, qz, qw = q32.unbind(dim=-1) + two = torch.tensor(2.0, dtype=torch.float32, device=q32.device) + + r00 = 1 - two * (qy * qy + qz * qz) + r01 = two * (qx * qy - qz * qw) + r02 = two * (qx * qz + qy * qw) + r10 = two * (qx * qy + qz * qw) + r11 = 1 - two * (qx * qx + qz * qz) + r12 = two * (qy * qz - qx * qw) + r20 = two * (qx * qz - qy * qw) + r21 = two * (qx * qw + qy * qz) + r22 = 1 - two * (qx * qx + qy * qy) + + R32 = torch.stack( + [ + torch.stack([r00, r01, r02], dim=-1), + torch.stack([r10, r11, r12], dim=-1), + torch.stack([r20, r21, r22], dim=-1), + ], + dim=-2, + ) + return R32.to(q.dtype) + + @staticmethod + @supports_numpy(["R"], use_no_grad=True) + def from_rotation_matrix(R: torch.Tensor | np.ndarray, eps: float = 1e-8) -> torch.Tensor | np.ndarray: + """Convert rotation matrix/matrices to quaternion(s). + + Args: + R (torch.Tensor [...,3,3]): Rotation matrix/matrices. + eps (float): Numerical stability epsilon. + + Returns: + q (torch.Tensor [...,4]): Quaternion(s) in (x, y, z, w) order. + """ + assert R.shape[-2:] == (3, 3), "Rotation matrix is not of shape (..., 3, 3)." + R32 = R.to(torch.float32) + m00 = R32[..., 0, 0] + m11 = R32[..., 1, 1] + m22 = R32[..., 2, 2] + trace = m00 + m11 + m22 + + q32 = torch.empty(*R32.shape[:-2], 4, dtype=torch.float32, device=R32.device) + + cond0 = trace > 0 + eps_eff = max(eps, 1e-6) + s0 = torch.sqrt(trace + 1.0 + eps_eff) * 2.0 + qw0 = 0.25 * s0 + qx0 = (R[..., 2, 1] - R[..., 1, 2]) / s0 + qy0 = (R[..., 0, 2] - R[..., 2, 0]) / s0 + qz0 = (R[..., 1, 0] - R[..., 0, 1]) / s0 + + cond1 = (~cond0) & (m00 > m11) & (m00 > m22) + s1 = torch.sqrt(1.0 + m00 - m11 - m22 + eps_eff) * 2.0 + qw1 = (R[..., 2, 1] - R[..., 1, 2]) / s1 + qx1 = 0.25 * s1 + qy1 = (R[..., 0, 1] + R[..., 1, 0]) / s1 + qz1 = (R[..., 0, 2] + R[..., 2, 0]) / s1 + + cond2 = (~cond0) & (~cond1) & (m11 > m22) + s2 = torch.sqrt(1.0 + m11 - m00 - m22 + eps_eff) * 2.0 + qw2 = (R[..., 0, 2] - R[..., 2, 0]) / s2 + qx2 = (R[..., 0, 1] + R[..., 1, 0]) / s2 + qy2 = 0.25 * s2 + qz2 = (R[..., 1, 2] + R[..., 2, 1]) / s2 + + cond3 = (~cond0) & (~cond1) & (~cond2) + s3 = torch.sqrt(1.0 + m22 - m00 - m11 + eps_eff) * 2.0 + qw3 = (R[..., 1, 0] - R[..., 0, 1]) / s3 + qx3 = (R[..., 0, 2] + R[..., 2, 0]) / s3 + qy3 = (R[..., 1, 2] + R[..., 2, 1]) / s3 + qz3 = 0.25 * s3 + + qw = torch.where(cond0, qw0, torch.where(cond1, qw1, torch.where(cond2, qw2, qw3))) + qx = torch.where(cond0, qx0, torch.where(cond1, qx1, torch.where(cond2, qx2, qx3))) + qy = torch.where(cond0, qy0, torch.where(cond1, qy1, torch.where(cond2, qy2, qy3))) + qz = torch.where(cond0, qz0, torch.where(cond1, qz1, torch.where(cond2, qz2, qz3))) + + q32[..., 0] = qx + q32[..., 1] = qy + q32[..., 2] = qz + q32[..., 3] = qw + + q_out = Quaternion.normalize(q32).to(R.dtype) + return q_out + + @staticmethod + @supports_numpy(["q"], use_no_grad=True) + def invert(q: torch.Tensor | np.ndarray, eps: float = 1e-8) -> torch.Tensor | np.ndarray: + """Inverse of quaternion(s). Equivalent to conjugate of quaternion. + + For unit quaternions, the inverse equals the conjugate. For non-unit quaternions, + q^{-1} = conjugate(q) / ||q||^2. + + Args: + q (torch.Tensor [...,4]): Input quaternion(s). + eps (float): Small epsilon to avoid division by zero. + + Returns: + q_inv (torch.Tensor [...,4]): Inverted quaternion(s). + """ + # Enforce unit quaternions; inverse equals conjugate + Quaternion._check_valid_quaternion(q, require_normalized=True) + qx, qy, qz, qw = q.unbind(dim=-1) + return torch.stack([-qx, -qy, -qz, qw], dim=-1) + + @staticmethod + @supports_numpy(["q1", "q2"], use_no_grad=True) + def multiply(q1: torch.Tensor | np.ndarray, q2: torch.Tensor | np.ndarray) -> torch.Tensor | np.ndarray: + """Hamilton product of two quaternion sets. + + Args: + q1 (torch.Tensor [...,4]): Left quaternion(s) in (x, y, z, w) order. + q2 (torch.Tensor [...,4]): Right quaternion(s) in (x, y, z, w) order. + + Returns: + q (torch.Tensor [...,4]): Product quaternion(s) q = q1 ⊗ q2 in (x, y, z, w) order. + """ + # Enforce unit inputs + Quaternion._check_valid_quaternion(q1, require_normalized=True) + Quaternion._check_valid_quaternion(q2, require_normalized=True) + q1x, q1y, q1z, q1w = q1.unbind(dim=-1) + q2x, q2y, q2z, q2w = q2.unbind(dim=-1) + qx = q1w * q2x + q2w * q1x + q1y * q2z - q1z * q2y + qy = q1w * q2y + q2w * q1y + q1z * q2x - q1x * q2z + qz = q1w * q2z + q2w * q1z + q1x * q2y - q1y * q2x + qw = q1w * q2w - (q1x * q2x + q1y * q2y + q1z * q2z) + q = torch.stack([qx, qy, qz, qw], dim=-1) + # Re-normalize to counteract numerical drift and enforce constraint + return Quaternion.normalize(q) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/camera_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/camera_test.py new file mode 100644 index 0000000000000000000000000000000000000000..bd0aa5d4f202bbcba6352a898747d99fefe7f803 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/camera_test.py @@ -0,0 +1,437 @@ +# ----------------------------------------------------------------------------- +# Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. +# All rights reserved. +# ----------------------------------------------------------------------------- + +import numpy as np +import pytest +import torch + +from cosmos_policy._src.imaginaire.modules.camera import Camera, Quaternion + + +def _make_pose(R: torch.Tensor, t: torch.Tensor) -> torch.Tensor: + return torch.cat([R, t.unsqueeze(-1)], dim=-1) + + +def _random_unit_quaternion(batch_shape=()) -> torch.Tensor: + q = torch.randn(*batch_shape, 4) + return Quaternion.normalize(q) + + +def _rand_quaternion(dtype: torch.dtype = torch.float32, device: str = "cpu") -> torch.Tensor: + q = torch.randn(4, dtype=dtype, device=device) + return Quaternion.normalize(q) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_camera_invert_pose_identity(): + R = torch.eye(3) + t = torch.zeros(3) + pose = _make_pose(R, t) + pose_inv = Camera.invert_pose(pose) + assert torch.allclose(pose, pose_inv) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_camera_invert_pose_roundtrip_points(): + # Create a valid random rotation via quaternion and random translation + q = _random_unit_quaternion() + R = Quaternion.to_rotation_matrix(q) + t = torch.tensor([0.3, -1.2, 2.5]) + pose = _make_pose(R, t) + + points = torch.randn(7, 3) + pc = Camera.world2camera(points, pose) + pw = Camera.camera2world(pc, pose) + assert torch.allclose(points, pw, atol=1e-5) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_camera_compose_poses_matches_matrix_product(): + # Build two random valid poses + q1 = _random_unit_quaternion() + q2 = _random_unit_quaternion() + R1 = Quaternion.to_rotation_matrix(q1) + R2 = Quaternion.to_rotation_matrix(q2) + t1 = torch.tensor([0.1, 0.2, -0.3]) + t2 = torch.tensor([-1.0, 0.5, 0.7]) + pose1 = _make_pose(R1, t1) + pose2 = _make_pose(R2, t2) + + # Compose using implementation + pose_comp = Camera.compose_poses([pose1, pose2]) + + # Compose in homogeneous 4x4 explicitly: H = H2 @ H1 + def to_h(T): + return torch.cat([torch.cat([T[..., :3], T[..., 3:]], dim=-1), torch.tensor([[0.0, 0.0, 0.0, 1.0]])]) + + H1 = to_h(pose1) + H2 = to_h(pose2) + Hc = H2 @ H1 + # Back to 3x4 + pose_expected = Hc[:3, :] + + assert torch.allclose(pose_comp, pose_expected, atol=1e-5) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_camera_image_camera_inversion(): + K = torch.tensor([[500.0, 0.0, 320.0], [0.0, 500.0, 240.0], [0.0, 0.0, 1.0]]) + # Random camera-space points in homogeneous depth-1 coordinates + cam_pts = torch.randn(11, 3) + img_pts = Camera.camera2image(cam_pts, K) + cam_pts_rec = Camera.image2camera(img_pts, K) + assert torch.allclose(cam_pts, cam_pts_rec, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_camera_image_camera_inversion_batched(): + # Batch of intrinsics + K = torch.stack( + [ + torch.tensor([[500.0, 0.0, 320.0], [0.0, 500.0, 240.0], [0.0, 0.0, 1.0]]), + torch.tensor([[800.0, 0.0, 100.0], [0.0, 600.0, 50.0], [0.0, 0.0, 1.0]]), + ], + dim=0, + ) # [B,3,3] + B, N = 2, 13 + cam_pts = torch.randn(B, N, 3) + img_pts = Camera.camera2image(cam_pts, K) + cam_pts_rec = Camera.image2camera(img_pts, K) + assert torch.allclose(cam_pts, cam_pts_rec, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_intrinsics_param_matrix_roundtrip(): + params = torch.tensor([500.0, 600.0, 320.0, 240.0]) + K = Camera.intrinsic_params_to_matrices(params) + params_rec = Camera.intrinsic_matrices_to_params(K) + assert torch.allclose(params, params_rec, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_intrinsics_param_matrix_roundtrip_batched(): + params = torch.tensor( + [ + [500.0, 600.0, 320.0, 240.0], + [800.0, 400.0, 100.0, 50.0], + ] + ) + K = Camera.intrinsic_params_to_matrices(params) + params_rec = Camera.intrinsic_matrices_to_params(K) + assert torch.allclose(params, params_rec, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_camera_rays_identity_intrinsics(): + # Identity extrinsics -> camera and world frames coincide + R = torch.eye(3) + t = torch.zeros(3) + pose = _make_pose(R, t) + + # Simple intrinsics + fx, fy, cx, cy = 100.0, 150.0, 2.0, 1.0 + K = torch.tensor([[fx, 0.0, cx], [0.0, fy, cy], [0.0, 0.0, 1.0]]) + + H, W = 2, 3 + rays = Camera.get_camera_rays(pose, K, (H, W)) + + # Check a couple of pixels for expected ray directions at depth=1 + def expected_ray(x, y): + xn = (x + 0.5 - cx) / fx + yn = (y + 0.5 - cy) / fy + return torch.tensor([xn, yn, 1.0]) + + # (y=0,x=0) + v = expected_ray(0, 0) + v = v / v.norm() + assert torch.allclose(rays[0], v, atol=1e-6) + # (y=1,x=2) + idx = 1 * W + 2 + v = expected_ray(2, 1) + v = v / v.norm() + assert torch.allclose(rays[idx], v, atol=1e-6) + + # All rays should be unit length + assert torch.allclose(rays.norm(dim=-1), torch.ones(H * W), atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_plucker_rays_properties(): + # Identity pose and simple intrinsics + R = torch.eye(3) + t = torch.zeros(3) + pose = _make_pose(R, t) + fx, fy, cx, cy = 100.0, 150.0, 2.0, 1.0 + K = torch.tensor([[fx, 0.0, cx], [0.0, fy, cy], [0.0, 0.0, 1.0]]) + H, W = 2, 3 + plucker = Camera.get_plucker_rays(pose, K, (H, W)) # [HW,6] + moment, direction = plucker[..., :3], plucker[..., 3:] + # Directions unit + assert torch.allclose(direction.norm(dim=-1), torch.ones(H * W), atol=1e-6) + # For camera at origin, m = o × d = 0 + assert torch.allclose(moment, torch.zeros(H * W, 3), atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_camera_get_camera_center_matches_formula(): + q = _random_unit_quaternion() + R = Quaternion.to_rotation_matrix(q) + t = torch.tensor([0.3, -1.2, 2.5]) + pose = _make_pose(R, t) + center = Camera.get_camera_center(pose) + # Center should satisfy R @ C + t = 0 -> C = -R^T t + expected = -R.T @ t + assert torch.allclose(center, expected, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_quaternion_invert_roundtrip(): + q = torch.tensor([0.2, -0.5, 0.7, 0.1]) + qn = Quaternion.normalize(q) + inv = Quaternion.invert(qn) + + # q ⊗ q^{-1} = identity and q^{-1} ⊗ q = identity + identity = torch.tensor([0.0, 0.0, 0.0, 1.0]) + prod1 = Quaternion.multiply(qn, inv) + prod2 = Quaternion.multiply(inv, qn) + assert torch.allclose(prod1, identity, atol=1e-6) + assert torch.allclose(prod2, identity, atol=1e-6) + + # invert(invert(q)) = q for unit quaternions + q_back = Quaternion.invert(inv) + assert torch.allclose(q_back, qn, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_quaternion_to_from_rotation_matrix_roundtrip(): + q = _random_unit_quaternion() + R = Quaternion.to_rotation_matrix(q) + q2 = Quaternion.from_rotation_matrix(R) + # Account for double-cover ambiguity: q == -q + err1 = (q - q2).abs().max() + err2 = (q + q2).abs().max() + assert min(err1.item(), err2.item()) < 1e-5 + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_quaternion_rotation_matrix_properties(): + q = _random_unit_quaternion() + R = Quaternion.to_rotation_matrix(q) + I = torch.eye(3) + assert torch.allclose(R.T @ R, I, atol=1e-6) + det = torch.det(R) + assert torch.allclose(det, torch.tensor(1.0), atol=1e-5) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_quaternion_multiply_matches_rotation_composition(): + q1 = _random_unit_quaternion() + q2 = _random_unit_quaternion() + q12 = Quaternion.multiply(q1, q2) + + R1 = Quaternion.to_rotation_matrix(q1) + R2 = Quaternion.to_rotation_matrix(q2) + R12 = Quaternion.to_rotation_matrix(q12) + + assert torch.allclose(R12, R1 @ R2, atol=1e-5) + + +# ----------------------------------------------------------------------------- +# bf16-oriented tests +# ----------------------------------------------------------------------------- + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_check_valid_pose_bf16_tolerance(): + device = "cpu" + dtype = torch.bfloat16 + q = _rand_quaternion(dtype=dtype, device=device) + R = Quaternion.to_rotation_matrix(q) + t = torch.zeros(3, 1, dtype=dtype, device=device) + cam_pose = torch.cat([R, t], dim=-1) + Camera._check_valid_pose(cam_pose) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_world2camera_camera2world_roundtrip_bf16(): + device = "cpu" + dtype = torch.bfloat16 + q = _rand_quaternion(dtype=dtype, device=device) + R = Quaternion.to_rotation_matrix(q) + t = torch.tensor([[0.1], [-0.2], [0.3]], dtype=dtype, device=device) + cam_pose = torch.cat([R, t], dim=-1) + + points = torch.tensor([[0.5, -0.1, 2.0], [1.2, 0.3, 4.0], [-0.7, 0.9, 1.5]], dtype=dtype, device=device) + cam = Camera.world2camera(points, cam_pose) + back = Camera.camera2world(cam, cam_pose) + assert torch.allclose(back.to(torch.float32), points.to(torch.float32), rtol=1e-3, atol=2e-2) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_image2camera_camera2image_roundtrip_bf16(): + device = "cpu" + dtype = torch.bfloat16 + fx, fy, cx, cy = 500.0, 480.0, 320.0, 240.0 + K = torch.tensor([[fx, 0.0, cx], [0.0, fy, cy], [0.0, 0.0, 1.0]], dtype=dtype, device=device) + pts_cam = torch.tensor([[-0.2, 0.1, 1.0], [0.3, -0.4, 2.0], [0.0, 0.0, 3.0]], dtype=dtype, device=device) + pix = Camera.camera2image(pts_cam, K) + rec = Camera.image2camera(pix, K) + assert torch.allclose(rec.to(torch.float32), pts_cam.to(torch.float32), rtol=1e-3, atol=2e-2) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_get_camera_rays_bf16_unit_norm(): + device = "cpu" + dtype = torch.bfloat16 + q = _rand_quaternion(dtype=dtype, device=device) + R = Quaternion.to_rotation_matrix(q) + t = torch.zeros(3, 1, dtype=dtype, device=device) + cam_pose = torch.cat([R, t], dim=-1) + fx, fy, cx, cy = 400.0, 400.0, 1.0, 1.0 + K = torch.tensor([[fx, 0.0, cx], [0.0, fy, cy], [0.0, 0.0, 1.0]], dtype=dtype, device=device) + rays = Camera.get_camera_rays(cam_pose, K, image_size=(3, 3)) + norms = rays.to(torch.float32).norm(dim=-1) + assert torch.allclose(norms, torch.ones_like(norms), rtol=1e-3, atol=2e-2) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_quaternion_roundtrip_bf16(): + device = "cpu" + dtype = torch.bfloat16 + q = _rand_quaternion(dtype=dtype, device=device) + R = Quaternion.to_rotation_matrix(q) + q2 = Quaternion.from_rotation_matrix(R) + d = torch.sum(q.to(torch.float32) * q2.to(torch.float32)) + assert torch.isfinite(d) + assert abs(float(d)) >= 0.98 + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_xyzw_t_pose_roundtrip(): + # Random unit quaternion and translation + q = _random_unit_quaternion() + t = torch.randn(3) + vec = torch.cat([q, t], dim=-1) + pose = Camera.extrinsic_params_to_matrices(vec) + vec_rec = Camera.extrinsic_matrices_to_params(pose) + # Compare quaternion up to sign + q_rec, t_rec = vec_rec[:4], vec_rec[4:] + err1 = (q - q_rec).abs().max() + err2 = (q + q_rec).abs().max() + assert min(err1.item(), err2.item()) < 1e-5 + assert torch.allclose(t, t_rec, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_xyzw_t_to_pose_matches_rotation_matrix(): + q = _random_unit_quaternion() + t = torch.tensor([0.4, -0.2, 1.1]) + pose = Camera.extrinsic_params_to_matrices(torch.cat([q, t], dim=-1)) + R = Quaternion.to_rotation_matrix(q) + assert torch.allclose(pose[:3, :3], R, atol=1e-6) + assert torch.allclose(pose[:3, 3], t, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_numpy_world_camera_roundtrip(): + q = _random_unit_quaternion() + R = Quaternion.to_rotation_matrix(q) + t = torch.randn(3) + pose = _make_pose(R, t) + points = torch.randn(9, 3) + + pose_np = pose.detach().cpu().numpy() + points_np = points.detach().cpu().numpy() + + pc_np = Camera.world2camera(points_np, pose_np) + pw_np = Camera.camera2world(pc_np, pose_np) + assert np.allclose(points_np, pw_np, atol=1e-5) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_numpy_image_camera_roundtrip(): + K = torch.tensor([[500.0, 0.0, 320.0], [0.0, 500.0, 240.0], [0.0, 0.0, 1.0]]) + cam_pts = torch.randn(7, 3) + K_np = K.detach().cpu().numpy() + cam_pts_np = cam_pts.detach().cpu().numpy() + img_pts_np = Camera.camera2image(cam_pts_np, K_np) + cam_pts_rec_np = Camera.image2camera(img_pts_np, K_np) + assert np.allclose(cam_pts_np, cam_pts_rec_np, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_numpy_intrinsic_param_matrix_roundtrip(): + params_np = np.array([500.0, 600.0, 320.0, 240.0], dtype=np.float32) + K_np = Camera.intrinsic_params_to_matrices(params_np) + params_rec_np = Camera.intrinsic_matrices_to_params(K_np) + assert np.allclose(params_np, params_rec_np, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_numpy_extrinsic_param_matrix_roundtrip(): + q = _random_unit_quaternion() + t = torch.randn(3) + vec_np = torch.cat([q, t], dim=-1).detach().cpu().numpy() + pose_np = Camera.extrinsic_params_to_matrices(vec_np) + vec_rec_np = Camera.extrinsic_matrices_to_params(pose_np) + q_np, t_np = vec_np[:4], vec_np[4:] + q_rec_np, t_rec_np = vec_rec_np[:4], vec_rec_np[4:] + err1 = np.max(np.abs(q_np - q_rec_np)) + err2 = np.max(np.abs(q_np + q_rec_np)) + assert min(err1, err2) < 1e-5 + assert np.allclose(t_np, t_rec_np, atol=1e-6) + + +@pytest.mark.L0 +@pytest.mark.GPU +def test_numpy_rays_and_plucker(): + R = torch.eye(3) + t = torch.zeros(3) + pose_np = _make_pose(R, t).detach().cpu().numpy() + fx, fy, cx, cy = 100.0, 150.0, 2.0, 1.0 + K_np = np.array([[fx, 0.0, cx], [0.0, fy, cy], [0.0, 0.0, 1.0]], dtype=np.float32) + H, W = 2, 3 + + rays_np = Camera.get_camera_rays(pose_np, K_np, (H, W)) + norms = np.linalg.norm(rays_np, axis=-1) + assert np.allclose(norms, np.ones(H * W), atol=1e-6) + + def expected_ray(x, y): + xn = (x + 0.5 - cx) / fx + yn = (y + 0.5 - cy) / fy + v = np.array([xn, yn, 1.0], dtype=np.float32) + return v / np.linalg.norm(v) + + assert np.allclose(rays_np[0], expected_ray(0, 0), atol=1e-6) + + plucker_np = Camera.get_plucker_rays(pose_np, K_np, (H, W)) + m_np, d_np = plucker_np[..., :3], plucker_np[..., 3:] + assert np.allclose(np.linalg.norm(d_np, axis=-1), np.ones(H * W), atol=1e-6) + assert np.allclose(m_np, np.zeros((H * W, 3)), atol=1e-6) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/denoiser_scaling.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/denoiser_scaling.py new file mode 100644 index 0000000000000000000000000000000000000000..6d843ef40e06ea06253ef837a4036f30d9c940c8 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/denoiser_scaling.py @@ -0,0 +1,64 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Tuple + +import torch + + +class EDMScaling: + def __init__(self, sigma_data: float = 0.5): + self.sigma_data = sigma_data + + def __call__(self, sigma: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]: + c_skip = self.sigma_data**2 / (sigma**2 + self.sigma_data**2) + c_out = sigma * self.sigma_data / (sigma**2 + self.sigma_data**2) ** 0.5 + c_in = 1 / (sigma**2 + self.sigma_data**2) ** 0.5 + c_noise = 0.25 * sigma.log() + return c_skip, c_out, c_in, c_noise + + +class RectifiedFlowScaling: + def __init__(self, sigma_data: float = 1.0, t_scaling_factor: float = 1.0, loss_weight_uniform: bool = True): + assert abs(sigma_data - 1.0) < 1e-6, "sigma_data must be 1.0 for RectifiedFlowScaling" + self.t_scaling_factor = t_scaling_factor + self.loss_weight_uniform = loss_weight_uniform + if loss_weight_uniform is False: + # using huan lin suggested one here. which put more weight on the middle of the timesteps. + self.num_steps = 1000 + t = torch.linspace(0, 1, self.num_steps) + y = torch.exp(-2 * (t - 0.5) ** 2) + shift = y - y.min() + weights = shift * (self.num_steps / shift.sum()) # make sure the avg weights is 1.0 + self.weights = weights + + def sigma_loss_weights(self, sigma: torch.Tensor) -> torch.Tensor: + if self.loss_weight_uniform: + return (1.0 + sigma) ** 2 / sigma**2 + else: + t = sigma / (sigma + 1) + index = (t * self.num_steps).round().long() + # Clamp index to valid range [0, num_steps-1] to avoid out of bounds + index = torch.clamp(index, 0, self.num_steps - 1) + weights_on_device = self.weights.to(sigma.device) + return weights_on_device[index].type_as(sigma) + + def __call__(self, sigma: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]: + t = sigma / (sigma + 1) + c_skip = 1.0 - t + c_out = -t + c_in = 1.0 - t + c_noise = t * self.t_scaling_factor + return c_skip, c_out, c_in, c_noise diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/edm_sampler.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/edm_sampler.py new file mode 100644 index 0000000000000000000000000000000000000000..a32fe7d7e1abe575ca3ea043e08148824204408c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/edm_sampler.py @@ -0,0 +1,69 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Callable + +import numpy as np +import torch +from torch import nn + + +class Sampler(nn.Module): + @torch.no_grad() + def forward( + self, + x0_fn: Callable, + x_sigma_max: torch.Tensor, + num_steps: int = 35, + sigma_min: float = 0.002, + sigma_max: float = 80, + rho: float = 7, + S_churn: float = 0, + S_min: float = 0, + S_max: float = float("inf"), + S_noise: float = 1, + ) -> torch.Tensor: + # https://github.com/NVlabs/edm/blob/62072d2612c7da05165d6233d13d17d71f213fee/generate.py#L25 + # Time step discretization. + in_dtype = x_sigma_max.dtype + _ones = torch.ones(x_sigma_max.shape[0], dtype=in_dtype, device=x_sigma_max.device) + step_indices = torch.arange(num_steps, dtype=torch.float64, device=x_sigma_max.device) + t_steps = ( + sigma_max ** (1 / rho) + step_indices / (num_steps - 1) * (sigma_min ** (1 / rho) - sigma_max ** (1 / rho)) + ) ** rho + t_steps = torch.cat([t_steps, torch.zeros_like(t_steps[:1])]) # t_N = 0 + + # Main sampling loop. + x_next = x_sigma_max.to(torch.float64) + for i, (t_cur, t_next) in enumerate(zip(t_steps[:-1], t_steps[1:], strict=False)): # 0, ..., N-1 + x_cur = x_next + + # Increase noise temporarily. + gamma = min(S_churn / num_steps, np.sqrt(2) - 1) if S_min <= t_cur <= S_max else 0 + t_hat = t_cur + gamma * t_cur + x_hat = x_cur + (t_hat**2 - t_cur**2).sqrt() * S_noise * torch.randn_like(x_cur) + + # Euler step. + denoised = x0_fn(x_hat.to(in_dtype), t_hat.to(in_dtype) * _ones).to(torch.float64) + d_cur = (x_hat - denoised) / t_hat + x_next = x_hat + (t_next - t_hat) * d_cur + + # Apply 2nd order correction. + if i < num_steps - 1: + denoised = x0_fn(x_hat.to(in_dtype), t_hat.to(in_dtype) * _ones).to(torch.float64) + d_prime = (x_next - denoised) / t_next + x_next = x_hat + (t_next - t_hat) * (0.5 * d_cur + 0.5 * d_prime) + + return x_next.to(in_dtype) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/edm_sde.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/edm_sde.py new file mode 100644 index 0000000000000000000000000000000000000000..3d08a8229f03c9fdd6a8d905ad4543fe5fe5238a --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/edm_sde.py @@ -0,0 +1,43 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from statistics import NormalDist + +import numpy as np +import torch + + +class EDMSDE: + def __init__( + self, + p_mean: float = -1.2, + p_std: float = 1.2, + sigma_max: float = 80.0, + sigma_min: float = 0.002, + ): + self.gaussian_dist = NormalDist(mu=p_mean, sigma=p_std) + self.sigma_max = sigma_max + self.sigma_min = sigma_min + + def sample_t(self, batch_size: int) -> torch.Tensor: + cdf_vals = np.random.uniform(size=(batch_size)) + samples_interval_gaussian = [self.gaussian_dist.inv_cdf(cdf_val) for cdf_val in cdf_vals] + + log_sigma = torch.tensor(samples_interval_gaussian, device="cuda") + return torch.exp(log_sigma) + + def marginal_prob(self, x0: torch.Tensor, sigma: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]: + """This is trivial in the base class, but may be used by derived classes in a more interesting way""" + return x0, sigma diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/image_embeddings.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/image_embeddings.py new file mode 100644 index 0000000000000000000000000000000000000000..a8d4b5eb1b70dd54fdc03aa386abda1f8f584d50 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/image_embeddings.py @@ -0,0 +1,766 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Forked from https://github.com/openai/CLIP/blob/main/clip/model.py +This file differs in that it exposes the prepooled patch tokens alongside the global image tokens +when calling the CLIP model +""" + +import hashlib +import os +import urllib +import warnings +from collections import OrderedDict +from typing import Any, List, Tuple, Union # noqa: F401 + +import numpy as np +import torch +import torch.nn.functional as F +from PIL import Image +from pkg_resources import packaging +from torch import nn +from torchvision.transforms import CenterCrop, Compose, Normalize, Resize, ToTensor +from tqdm import tqdm + +try: + from torchvision.transforms import InterpolationMode + + BICUBIC = InterpolationMode.BICUBIC +except ImportError: + BICUBIC = Image.BICUBIC + + +if packaging.version.parse(torch.__version__) < packaging.version.parse("1.7.1"): + warnings.warn("PyTorch version 1.7.1 or higher is recommended") + +__all__ = ["available_models", "load"] + +_MODELS = { + "RN50": "https://openaipublic.azureedge.net/clip/models/afeb0e10f9e5a86da6080e35cf09123aca3b358a0c3e3b6c78a7b63bc04b6762/RN50.pt", # noqa: E501 + "RN101": "https://openaipublic.azureedge.net/clip/models/8fa8567bab74a42d41c5915025a8e4538c3bdbe8804a470a72f30b0d94fab599/RN101.pt", # noqa: E501 + "RN50x4": "https://openaipublic.azureedge.net/clip/models/7e526bd135e493cef0776de27d5f42653e6b4c8bf9e0f653bb11773263205fdd/RN50x4.pt", # noqa: E501 + "RN50x16": "https://openaipublic.azureedge.net/clip/models/52378b407f34354e150460fe41077663dd5b39c54cd0bfd2b27167a4a06ec9aa/RN50x16.pt", # noqa: E501 + "RN50x64": "https://openaipublic.azureedge.net/clip/models/be1cfb55d75a9666199fb2206c106743da0f6468c9d327f3e0d0a543a9919d9c/RN50x64.pt", # noqa: E501 + "ViT-B/32": "https://openaipublic.azureedge.net/clip/models/40d365715913c9da98579312b702a82c18be219cc2a73407c4526f58eba950af/ViT-B-32.pt", # noqa: E501 + "ViT-B/16": "https://openaipublic.azureedge.net/clip/models/5806e77cd80f8b59890b7e101eabd078d9fb84e6937f9e85e4ecb61988df416f/ViT-B-16.pt", # noqa: E501 + "ViT-L/14": "https://openaipublic.azureedge.net/clip/models/b8cca3fd41ae0c99ba7e8951adf17d267cdb84cd88be6f7c2e0eca1737a03836/ViT-L-14.pt", # noqa: E501 + "ViT-L/14@336px": "https://openaipublic.azureedge.net/clip/models/3035c92b350959924f9f00213499208652fc7ea050643e8b385c2dac08641f02/ViT-L-14-336px.pt", # noqa: E501 +} + + +class Bottleneck(nn.Module): + """Bottleneck residual block for ResNet.""" + + expansion = 4 + + def __init__(self, inplanes, planes, stride=1): + super().__init__() + + # all conv layers have stride 1. an avgpool is performed after the second convolution when stride > 1 + self.conv1 = nn.Conv2d(inplanes, planes, 1, bias=False) + self.bn1 = nn.BatchNorm2d(planes) + self.relu1 = nn.ReLU(inplace=True) + + self.conv2 = nn.Conv2d(planes, planes, 3, padding=1, bias=False) + self.bn2 = nn.BatchNorm2d(planes) + self.relu2 = nn.ReLU(inplace=True) + + self.avgpool = nn.AvgPool2d(stride) if stride > 1 else nn.Identity() + + self.conv3 = nn.Conv2d(planes, planes * self.expansion, 1, bias=False) + self.bn3 = nn.BatchNorm2d(planes * self.expansion) + self.relu3 = nn.ReLU(inplace=True) + + self.downsample = None + self.stride = stride + + if stride > 1 or inplanes != planes * Bottleneck.expansion: + # downsampling layer is prepended with an avgpool, and the subsequent convolution has stride 1 + self.downsample = nn.Sequential( + OrderedDict( + [ + ("-1", nn.AvgPool2d(stride)), + ("0", nn.Conv2d(inplanes, planes * self.expansion, 1, stride=1, bias=False)), + ("1", nn.BatchNorm2d(planes * self.expansion)), + ] + ) + ) + + def forward(self, x: torch.Tensor): + identity = x + + out = self.relu1(self.bn1(self.conv1(x))) + out = self.relu2(self.bn2(self.conv2(out))) + out = self.avgpool(out) + out = self.bn3(self.conv3(out)) + + if self.downsample is not None: + identity = self.downsample(x) + + out += identity + out = self.relu3(out) + return out + + +class AttentionPool2d(nn.Module): + """Attention pooling layer for 2D feature maps.""" + + def __init__(self, spacial_dim: int, embed_dim: int, num_heads: int, output_dim: int = None): + super().__init__() + self.positional_embedding = nn.Parameter(torch.randn(spacial_dim**2 + 1, embed_dim) / embed_dim**0.5) + self.k_proj = nn.Linear(embed_dim, embed_dim) + self.q_proj = nn.Linear(embed_dim, embed_dim) + self.v_proj = nn.Linear(embed_dim, embed_dim) + self.c_proj = nn.Linear(embed_dim, output_dim or embed_dim) + self.num_heads = num_heads + + def forward(self, x): + x = x.reshape(x.shape[0], x.shape[1], x.shape[2] * x.shape[3]).permute(2, 0, 1) # NCHW -> (HW)NC + x = torch.cat([x.mean(dim=0, keepdim=True), x], dim=0) # (HW+1)NC + x = x + self.positional_embedding[:, None, :].to(x.dtype) # (HW+1)NC + x, _ = F.multi_head_attention_forward( + query=x, + key=x, + value=x, + embed_dim_to_check=x.shape[-1], + num_heads=self.num_heads, + q_proj_weight=self.q_proj.weight, + k_proj_weight=self.k_proj.weight, + v_proj_weight=self.v_proj.weight, + in_proj_weight=None, + in_proj_bias=torch.cat([self.q_proj.bias, self.k_proj.bias, self.v_proj.bias]), + bias_k=None, + bias_v=None, + add_zero_attn=False, + dropout_p=0, + out_proj_weight=self.c_proj.weight, + out_proj_bias=self.c_proj.bias, + use_separate_proj_weight=True, + training=self.training, + need_weights=False, + ) + + return x[0] + + +class ModifiedResNet(nn.Module): + """ + A ResNet class that is similar to torchvision's but contains the following changes: + - There are now 3 "stem" convolutions as opposed to 1, with an average pool instead of a max pool. + - Performs anti-aliasing strided convolutions, where an avgpool is prepended to convolutions with stride > 1 + - The final pooling layer is a QKV attention instead of an average pool + """ + + def __init__(self, layers, output_dim, heads, input_resolution=224, width=64): + super().__init__() + self.output_dim = output_dim + self.input_resolution = input_resolution + + # the 3-layer stem + self.conv1 = nn.Conv2d(3, width // 2, kernel_size=3, stride=2, padding=1, bias=False) + self.bn1 = nn.BatchNorm2d(width // 2) + self.relu1 = nn.ReLU(inplace=True) + self.conv2 = nn.Conv2d(width // 2, width // 2, kernel_size=3, padding=1, bias=False) + self.bn2 = nn.BatchNorm2d(width // 2) + self.relu2 = nn.ReLU(inplace=True) + self.conv3 = nn.Conv2d(width // 2, width, kernel_size=3, padding=1, bias=False) + self.bn3 = nn.BatchNorm2d(width) + self.relu3 = nn.ReLU(inplace=True) + self.avgpool = nn.AvgPool2d(2) + + # residual layers + self._inplanes = width # this is a *mutable* variable used during construction + self.layer1 = self._make_layer(width, layers[0]) + self.layer2 = self._make_layer(width * 2, layers[1], stride=2) + self.layer3 = self._make_layer(width * 4, layers[2], stride=2) + self.layer4 = self._make_layer(width * 8, layers[3], stride=2) + + embed_dim = width * 32 # the ResNet feature dimension + self.attnpool = AttentionPool2d(input_resolution // 32, embed_dim, heads, output_dim) + + def _make_layer(self, planes, blocks, stride=1): + """ + Create a layer of residual blocks. + - planes: the number of output channels for this layer + - blocks: the number of residual blocks in this layer + - stride: the stride to use for the first convolution of the layer + """ + layers = [Bottleneck(self._inplanes, planes, stride)] + + self._inplanes = planes * Bottleneck.expansion + for _ in range(1, blocks): + layers.append(Bottleneck(self._inplanes, planes)) + + return nn.Sequential(*layers) + + def forward(self, x): + def stem(x): + """ + The stem convolutions at the beginning of the network. + Performs 3 convolutions and an average pool. + """ + x = self.relu1(self.bn1(self.conv1(x))) + x = self.relu2(self.bn2(self.conv2(x))) + x = self.relu3(self.bn3(self.conv3(x))) + x = self.avgpool(x) + return x + + x = x.type(self.conv1.weight.dtype) + x = stem(x) + x = self.layer1(x) + x = self.layer2(x) + x = self.layer3(x) + x = self.layer4(x) + x = self.attnpool(x) + + return x + + +class LayerNorm(nn.LayerNorm): + """Subclass torch's LayerNorm to handle fp16.""" + + def forward(self, x: torch.Tensor): + orig_type = x.dtype + ret = super().forward(x.type(torch.float32)) + return ret.type(orig_type) + + +class QuickGELU(nn.Module): + """GELU activation function approximation""" + + def forward(self, x: torch.Tensor): + return x * torch.sigmoid(1.702 * x) + + +class ResidualAttentionBlock(nn.Module): + def __init__(self, d_model: int, n_head: int, attn_mask: torch.Tensor = None): + super().__init__() + + self.attn = nn.MultiheadAttention(d_model, n_head) + self.ln_1 = LayerNorm(d_model) + self.mlp = nn.Sequential( + OrderedDict( + [ + ("c_fc", nn.Linear(d_model, d_model * 4)), + ("gelu", QuickGELU()), + ("c_proj", nn.Linear(d_model * 4, d_model)), + ] + ) + ) + self.ln_2 = LayerNorm(d_model) + self.attn_mask = attn_mask + + def attention(self, x: torch.Tensor): + """Perform multi-head attention on the input tensor.""" + self.attn_mask = self.attn_mask.to(dtype=x.dtype, device=x.device) if self.attn_mask is not None else None + return self.attn(x, x, x, need_weights=False, attn_mask=self.attn_mask)[0] + + def forward(self, x: torch.Tensor): + x = x + self.attention(self.ln_1(x)) + x = x + self.mlp(self.ln_2(x)) + return x + + +class Transformer(nn.Module): + def __init__(self, width: int, layers: int, heads: int, attn_mask: torch.Tensor = None): + super().__init__() + self.width = width + self.layers = layers + self.resblocks = nn.Sequential(*[ResidualAttentionBlock(width, heads, attn_mask) for _ in range(layers)]) + + def forward(self, x: torch.Tensor): + return self.resblocks(x) + + +class VisionTransformer(nn.Module): + def __init__(self, input_resolution: int, patch_size: int, width: int, layers: int, heads: int, output_dim: int): + super().__init__() + self.input_resolution = input_resolution + self.output_dim = output_dim + self.conv1 = nn.Conv2d(in_channels=3, out_channels=width, kernel_size=patch_size, stride=patch_size, bias=False) + + scale = width**-0.5 + self.class_embedding = nn.Parameter(scale * torch.randn(width)) + self.positional_embedding = nn.Parameter(scale * torch.randn((input_resolution // patch_size) ** 2 + 1, width)) + self.ln_pre = LayerNorm(width) + + self.transformer = Transformer(width, layers, heads) + + self.ln_post = LayerNorm(width) + self.proj = nn.Parameter(scale * torch.randn(width, output_dim)) + + def forward(self, x: torch.Tensor): + x = self.conv1(x) # shape = [*, width, grid, grid] + x = x.reshape(x.shape[0], x.shape[1], -1) # shape = [*, width, grid ** 2] + x = x.permute(0, 2, 1) # shape = [*, grid ** 2, width] + x = torch.cat( + [ + self.class_embedding.to(x.dtype) + + torch.zeros(x.shape[0], 1, x.shape[-1], dtype=x.dtype, device=x.device), + x, + ], + dim=1, + ) + # shape = [*, grid ** 2 + 1, width] + x = x + self.positional_embedding.to(x.dtype) + x = self.ln_pre(x) + + x = x.permute(1, 0, 2) # NLD -> LND + x = self.transformer(x) + x = x.permute(1, 0, 2) # LND -> NLD + + x_pre_pooling = x + x = self.ln_post(x[:, 0, :]) + + if self.proj is not None: + x = x @ self.proj + + return x, x_pre_pooling + + +class CLIP(nn.Module): + """ + This CLIP module combines a visual encoder and a text encoder. It initializes the parameters, builds the attention mask, + and provides methods to encode images and text separately. The forward method computes the cosine similarity between + the encoded image and text features. + """ + + def __init__( + self, + embed_dim: int, + # vision + image_resolution: int, + vision_layers: Union[Tuple[int, int, int, int], int], + vision_width: int, + vision_patch_size: int, + # text + context_length: int, + vocab_size: int, + transformer_width: int, + transformer_heads: int, + transformer_layers: int, + ): + super().__init__() + + self.context_length = context_length + + if isinstance(vision_layers, (tuple, list)): + vision_heads = vision_width * 32 // 64 + self.visual = ModifiedResNet( + layers=vision_layers, + output_dim=embed_dim, + heads=vision_heads, + input_resolution=image_resolution, + width=vision_width, + ) + else: + vision_heads = vision_width // 64 + self.visual = VisionTransformer( + input_resolution=image_resolution, + patch_size=vision_patch_size, + width=vision_width, + layers=vision_layers, + heads=vision_heads, + output_dim=embed_dim, + ) + + self.transformer = Transformer( + width=transformer_width, + layers=transformer_layers, + heads=transformer_heads, + attn_mask=self.build_attention_mask(), + ) + + self.vocab_size = vocab_size + self.token_embedding = nn.Embedding(vocab_size, transformer_width) + self.positional_embedding = nn.Parameter(torch.empty(self.context_length, transformer_width)) + self.ln_final = LayerNorm(transformer_width) + + self.text_projection = nn.Parameter(torch.empty(transformer_width, embed_dim)) + self.logit_scale = nn.Parameter(torch.ones([]) * np.log(1 / 0.07)) + + self.initialize_parameters() + + def initialize_parameters(self): + """Initialize the parameters of the CLIP module.""" + nn.init.normal_(self.token_embedding.weight, std=0.02) + nn.init.normal_(self.positional_embedding, std=0.01) + + if isinstance(self.visual, ModifiedResNet): + if self.visual.attnpool is not None: + std = self.visual.attnpool.c_proj.in_features**-0.5 + nn.init.normal_(self.visual.attnpool.q_proj.weight, std=std) + nn.init.normal_(self.visual.attnpool.k_proj.weight, std=std) + nn.init.normal_(self.visual.attnpool.v_proj.weight, std=std) + nn.init.normal_(self.visual.attnpool.c_proj.weight, std=std) + + for resnet_block in [self.visual.layer1, self.visual.layer2, self.visual.layer3, self.visual.layer4]: + for name, param in resnet_block.named_parameters(): + if name.endswith("bn3.weight"): + nn.init.zeros_(param) + + proj_std = (self.transformer.width**-0.5) * ((2 * self.transformer.layers) ** -0.5) + attn_std = self.transformer.width**-0.5 + fc_std = (2 * self.transformer.width) ** -0.5 + for block in self.transformer.resblocks: + nn.init.normal_(block.attn.in_proj_weight, std=attn_std) + nn.init.normal_(block.attn.out_proj.weight, std=proj_std) + nn.init.normal_(block.mlp.c_fc.weight, std=fc_std) + nn.init.normal_(block.mlp.c_proj.weight, std=proj_std) + + if self.text_projection is not None: + nn.init.normal_(self.text_projection, std=self.transformer.width**-0.5) + + def build_attention_mask(self): + # lazily create causal attention mask, with full attention between the vision tokens + # pytorch uses additive attention mask; fill with -inf + mask = torch.empty(self.context_length, self.context_length) + mask.fill_(float("-inf")) + mask.triu_(1) # zero out the lower diagonal + return mask + + @property + def dtype(self): + return self.visual.conv1.weight.dtype + + def encode_image(self, image): + """ + Encode an image using the visual encoder. + + Args: + image (torch.Tensor): The input image. + + Returns: + torch.Tensor: The encoded image features. + """ + return self.visual(image.type(self.dtype)) + + def encode_text(self, text): + """ + Encode text using the text encoder. + + Args: + text (torch.Tensor): The input text. + + Returns: + torch.Tensor: The encoded text features. + """ + x = self.token_embedding(text).type(self.dtype) # [batch_size, n_ctx, d_model] + + x = x + self.positional_embedding.type(self.dtype) + x = x.permute(1, 0, 2) # NLD -> LND + x = self.transformer(x) + x = x.permute(1, 0, 2) # LND -> NLD + x = self.ln_final(x).type(self.dtype) + + # x.shape = [batch_size, n_ctx, transformer.width] + # take features from the eot embedding (eot_token is the highest number in each sequence) + x = x[torch.arange(x.shape[0]), text.argmax(dim=-1)] @ self.text_projection + + return x + + def forward(self, image, text): + """ + Forward pass of the CLIP module. + + Args: + image (torch.Tensor): The input image. + text (torch.Tensor): The input text. + + Returns: + tuple: A tuple containing the logits per image and logits per text. + """ + image_features, _ = self.encode_image(image) + text_features = self.encode_text(text) + + # normalized features + image_features = image_features / image_features.norm(dim=1, keepdim=True) + text_features = text_features / text_features.norm(dim=1, keepdim=True) + + # cosine similarity as logits + logit_scale = self.logit_scale.exp() + logits_per_image = logit_scale * image_features @ text_features.t() + logits_per_text = logits_per_image.t() + + # shape = [global_batch_size, global_batch_size] + return logits_per_image, logits_per_text + + +def convert_weights(model: nn.Module): + """Convert applicable model parameters to fp16""" + + def _convert_weights_to_fp16(ll): + if isinstance(ll, (nn.Conv1d, nn.Conv2d, nn.Linear)): + ll.weight.data = ll.weight.data.half() + if ll.bias is not None: + ll.bias.data = ll.bias.data.half() + + if isinstance(ll, nn.MultiheadAttention): + for attr in [*[f"{s}_proj_weight" for s in ["in", "q", "k", "v"]], "in_proj_bias", "bias_k", "bias_v"]: + tensor = getattr(ll, attr) + if tensor is not None: + tensor.data = tensor.data.half() + + for name in ["text_projection", "proj"]: + if hasattr(ll, name): + attr = getattr(ll, name) + if attr is not None: + attr.data = attr.data.half() + + model.apply(_convert_weights_to_fp16) + + +def build_model(state_dict: dict): + """Build the CLIP model from a state dictionary.""" + vit = state_dict.get("visual.proj") is not None + + if vit: + vision_width = state_dict["visual.conv1.weight"].shape[0] + vision_layers = len( + [k for k in state_dict.keys() if k.startswith("visual.") and k.endswith(".attn.in_proj_weight")] + ) + vision_patch_size = state_dict["visual.conv1.weight"].shape[-1] + grid_size = round((state_dict["visual.positional_embedding"].shape[0] - 1) ** 0.5) + image_resolution = vision_patch_size * grid_size + else: + counts: list = [ + len(set(k.split(".")[2] for k in state_dict if k.startswith(f"visual.layer{b}"))) for b in [1, 2, 3, 4] + ] + vision_layers = tuple(counts) + vision_width = state_dict["visual.layer1.0.conv1.weight"].shape[0] + output_width = round((state_dict["visual.attnpool.positional_embedding"].shape[0] - 1) ** 0.5) + vision_patch_size = None + assert output_width**2 + 1 == state_dict["visual.attnpool.positional_embedding"].shape[0] + image_resolution = output_width * 32 + + embed_dim = state_dict["text_projection"].shape[1] + context_length = state_dict["positional_embedding"].shape[0] + vocab_size = state_dict["token_embedding.weight"].shape[0] + transformer_width = state_dict["ln_final.weight"].shape[0] + transformer_heads = transformer_width // 64 + transformer_layers = len(set(k.split(".")[2] for k in state_dict if k.startswith("transformer.resblocks"))) + + model = CLIP( + embed_dim, + image_resolution, + vision_layers, + vision_width, + vision_patch_size, + context_length, + vocab_size, + transformer_width, + transformer_heads, + transformer_layers, + ) + + for key in ["input_resolution", "context_length", "vocab_size"]: + if key in state_dict: + del state_dict[key] + + convert_weights(model) + model.load_state_dict(state_dict) + return model.eval() + + +def _download(url: str, root: str): + """ + Download a file from a URL and place it in root. + + Args: + url (str): URL to download file from. + root (str): Directory to place the downloaded file. + + Returns: + str: Path to the downloaded file. + """ + os.makedirs(root, exist_ok=True) + filename = os.path.basename(url) + + expected_sha256 = url.split("/")[-2] + download_target = os.path.join(root, filename) + + if os.path.exists(download_target) and not os.path.isfile(download_target): + raise RuntimeError(f"{download_target} exists and is not a regular file") + + if os.path.isfile(download_target): + if hashlib.sha256(open(download_target, "rb").read()).hexdigest() == expected_sha256: + return download_target + else: + warnings.warn(f"{download_target} exists, but the SHA256 checksum does not match; re-downloading the file") + + with urllib.request.urlopen(url) as source, open(download_target, "wb") as output: + with tqdm( + total=int(source.info().get("Content-Length")), ncols=80, unit="iB", unit_scale=True, unit_divisor=1024 + ) as loop: + while True: + buffer = source.read(8192) + if not buffer: + break + + output.write(buffer) + loop.update(len(buffer)) + + if hashlib.sha256(open(download_target, "rb").read()).hexdigest() != expected_sha256: + raise RuntimeError("Model has been downloaded but the SHA256 checksum does not not match") + + return download_target + + +def _convert_image_to_rgb(image): + """ + Convert an image to RGB format. + + Args: + image (PIL.Image): The image to convert. + + Returns: + PIL.Image: The converted RGB image. + """ + return image.convert("RGB") + + +def _transform(n_px): + """ + Create a transformation pipeline for preprocessing images. + + Args: + n_px (int): The desired size of the transformed image. + + Returns: + Compose: A torchvision transform pipeline. + """ + return Compose( + [ + Resize(n_px, interpolation=BICUBIC), + CenterCrop(n_px), + _convert_image_to_rgb, + ToTensor(), + Normalize((0.48145466, 0.4578275, 0.40821073), (0.26862954, 0.26130258, 0.27577711)), + ] + ) + + +def available_models() -> List[str]: + """Returns the names of available CLIP models""" + return list(_MODELS.keys()) + + +def load( + name: str, + device: Union[str, torch.device] = "cuda" if torch.cuda.is_available() else "cpu", + jit: bool = False, + download_root: str = None, +): + """Load a CLIP model + + Parameters + ---------- + name : str + A model name listed by `clip.available_models()`, or the path to a model checkpoint containing the state_dict + + device : Union[str, torch.device] + The device to put the loaded model + + jit : bool + Whether to load the optimized JIT model or more hackable non-JIT model (default). + + download_root: str + path to download the model files; by default, it uses "~/.cache/clip" + + Returns + ------- + model : torch.nn.Module + The CLIP model + + preprocess : Callable[[PIL.Image], torch.Tensor] + A torchvision transform that converts a PIL image into a tensor that the returned model can take as its input + """ + if name in _MODELS: + model_path = _download(_MODELS[name], download_root or os.path.expanduser("~/.cache/clip")) + elif os.path.isfile(name): + model_path = name + else: + raise RuntimeError(f"Model {name} not found; available models = {available_models()}") + + with open(model_path, "rb") as opened_file: + try: + # loading JIT archive + model = torch.jit.load(opened_file, map_location=device if jit else "cpu").eval() + state_dict = None + except RuntimeError: + # loading saved state dict + if jit: + warnings.warn(f"File {model_path} is not a JIT archive. Loading as a state dict instead") + jit = False + state_dict = torch.load(opened_file, map_location="cpu") + + if not jit: + model = build_model(state_dict or model.state_dict()).to(device) + if str(device) == "cpu": + model.float() + return model, _transform(model.visual.input_resolution) + + # patch the device names + device_holder = torch.jit.trace(lambda: torch.ones([]).to(torch.device(device)), example_inputs=[]) + device_node = [n for n in device_holder.graph.findAllNodes("prim::Constant") if "Device" in repr(n)][-1] + + def patch_device(module): + try: + graphs = [module.graph] if hasattr(module, "graph") else [] + except RuntimeError: + graphs = [] + + if hasattr(module, "forward1"): + graphs.append(module.forward1.graph) + + for graph in graphs: + for node in graph.findAllNodes("prim::Constant"): + if "value" in node.attributeNames() and str(node["value"]).startswith("cuda"): + node.copyAttributes(device_node) + + model.apply(patch_device) + patch_device(model.encode_image) + patch_device(model.encode_text) + + # patch dtype to float32 on CPU + if str(device) == "cpu": + float_holder = torch.jit.trace(lambda: torch.ones([]).float(), example_inputs=[]) + float_input = list(float_holder.graph.findNode("aten::to").inputs())[1] + float_node = float_input.node() + + def patch_float(module): + try: + graphs = [module.graph] if hasattr(module, "graph") else [] + except RuntimeError: + graphs = [] + + if hasattr(module, "forward1"): + graphs.append(module.forward1.graph) + + for graph in graphs: + for node in graph.findAllNodes("aten::to"): + inputs = list(node.inputs()) + for i in [1, 2]: # dtype can be the second or third argument to aten::to() + if inputs[i].node()["value"] == 5: + inputs[i].node().copyAttributes(float_node) + + model.apply(patch_float) + patch_float(model.encode_image) + patch_float(model.encode_text) + + model.float() + + return model, _transform(model.input_resolution.item()) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/input_handling/utils.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/input_handling/utils.py new file mode 100644 index 0000000000000000000000000000000000000000..04c121f57a682b96625af053ffa32e23bcf5dc5e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/input_handling/utils.py @@ -0,0 +1,131 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math + +import numpy as np +import torch +import torchvision.transforms as transforms +from PIL import Image + + +def detect_aspect_ratio(img_size): + r""" + Function for detecting the closest aspect ratio. + """ + + _aspect_ratios = np.array([(16 / 9), (4 / 3), 1, (3 / 4), (9 / 16)]) + _aspect_ratio_keys = ["16,9", "4,3", "1,1", "3,4", "9,16"] + w, h = img_size + current_ratio = w / h + closest_aspect_ratio = np.argmin((_aspect_ratios - current_ratio) ** 2) + return _aspect_ratio_keys[closest_aspect_ratio] + + +def detect_resolution(img_size): + r""" + Function to detect resolution. + """ + w, h = img_size + if max(w, h) >= 1024 and min(w, h) >= 576: + resolution = 1024 + elif max(w, h) >= 256 and min(w, h) >= 144: + resolution = 256 + else: + raise ValueError("Images should be of at least 256x144 or 144x256 resolution") + return resolution + + +def resize_image_to_aspect_ratio(image, resolution, aspect_ratio, center_crop=True): + r""" + Function for resizing to a specific aspect ratio and a resolution. + """ + + # Finding the target shape based on resolution and aspect ratio + asp_ratio = aspect_ratio.split(",") + asp_ratio[0] = int(asp_ratio[0]) + asp_ratio[1] = int(asp_ratio[1]) + dim_ratio = asp_ratio[0] / asp_ratio[1] + if dim_ratio >= 1: + target_w, target_h = resolution, int(math.ceil(resolution / dim_ratio)) + else: + target_w, target_h = int(math.ceil(resolution * dim_ratio)), resolution + + if type(image) == torch.Tensor: + # Perform resizing + if center_crop: # Do aspect ratio preserving resize, then center crop + orig_h, orig_w = image.shape[-2:] + scaling_ratio = max((target_w / orig_w), (target_h / orig_h)) + resizing_shape = (int(math.ceil(scaling_ratio * orig_h)), int(math.ceil(scaling_ratio * orig_w))) + img_resized = torch.nn.functional.interpolate(image, resizing_shape, mode="bicubic") + + # Perform center crop + resize_box = [int((resizing_shape[0] - target_h) / 2), int((resizing_shape[1] - target_w) / 2), 0, 0] + resize_box[2] = resize_box[0] + target_h + resize_box[3] = resize_box[1] + target_w + img_resized = img_resized[:, :, resize_box[0] : resize_box[2], resize_box[1] : resize_box[3]] + else: # Directly resize to target aspect ratio. + img_resized = torch.nn.functional.interpolate(image, (target_h, target_w), mode="bicubic") + + else: + # Perform resizing + if center_crop: # Do aspect ratio preserving resize, then center crop + orig_w, orig_h = image.size + scaling_ratio = max((target_w / orig_w), (target_h / orig_h)) + resizing_shape = (int(math.ceil(scaling_ratio * orig_w)), int(math.ceil(scaling_ratio * orig_h))) + img_resized = image.resize(resizing_shape, Image.Resampling.BICUBIC, reducing_gap=True) + + # Perform center crop + resize_box = [int((resizing_shape[0] - target_w) / 2), int((resizing_shape[1] - target_h) / 2), 0, 0] + resize_box[2] = resize_box[0] + target_w + resize_box[3] = resize_box[1] + target_h + img_resized = img_resized.crop(resize_box) + else: # Directly resize to target aspect ratio. + img_resized = image.resize((target_w, target_h), Image.Resampling.BICUBIC, reducing_gap=True) + + return img_resized + + +def resize_batch(images, resolution): + r""" + Function for resizing a batch of images. + """ + assert isinstance(images, list), "Invalid input type. Expects a list of images as inputs" + aspect_ratio = detect_aspect_ratio(images[0].size) + images_resized = [ + resize_image_to_aspect_ratio(img, resolution=resolution, aspect_ratio=aspect_ratio) for img in images + ] + return images_resized + + +def process_input_image(input_image, resolution): + if isinstance(input_image, str): + return process_input_image(Image.open(input_image).convert("RGB"), resolution) + + if isinstance(input_image, Image.Image): + return process_input_image([input_image], resolution) + + if isinstance(input_image, list) and all(isinstance(i, Image.Image) for i in input_image): + # Perform resizing + input_image = resize_batch(input_image, resolution=resolution) + input_image = [ + (2.0 * transforms.functional.pil_to_tensor(img) / 255.0 - 1.0) for img in input_image + ] # [-1, 1] images, + return process_input_image(torch.stack(input_image), resolution) + + if isinstance(input_image, torch.Tensor): + return input_image.cuda(), input_image.shape[0] + + raise TypeError("Invalid input type. Expected one of [str, PIL.Image.Image, List[PIL.Image.Image], torch.Tensor]") diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/nlp/encodings.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/nlp/encodings.py new file mode 100644 index 0000000000000000000000000000000000000000..2a952974826a940347f65f18c1e70da8ea24516e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/nlp/encodings.py @@ -0,0 +1,209 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any, Union + +import torch +from einops import repeat + +from cosmos_policy._src.imaginaire.modules.nlp.t5xxl.t5encoder import T5Encoder + + +class TextEncoder: + """Text encoder base class.""" + + name: str + + def update_encoding_params(self, *args: Any, **kwargs: Any) -> None: + """Updates encoding params of your text encoder. + + Args: + *args: Whatever you need to update, e.g. max_len of encoding. + **kwargs: Keyword arguments are also possible. + + """ + raise NotImplementedError + + def __call__(self, input_text: Union[str, list[str]], **kwargs: Any) -> Any: + """Performs text encoding. + + Args: + input_text: A string or a list of strings to encode. + **kwargs: Keyword arguments are also possible. + + Return: + Your model's output. + """ + raise NotImplementedError + + +class T5TextEncoder(TextEncoder): + """Get T5 encoder for obtaining text encodings. + + Args: + t5_tokens_num (int): Max sequence length. + device (str): Device to load the model on to. + max_len (int): Max length of text encoded tokens to be returned. + dim (int): Dimension of each text encoded token. + """ + + def __init__( + self, + t5_tokens_num: int, + device: str = "cuda", + max_len: int = 113, + dim: int = 1024, + return_offsets_mapping: bool = False, + ): + super().__init__() + self.name = "T5XXL" + self.model = T5Encoder(max_seq_len=t5_tokens_num, device=device, return_offsets_mapping=return_offsets_mapping) + self.model = self.model.eval() + self.update_encoding_params(max_len=max_len, dim=dim, return_offsets_mapping=return_offsets_mapping) + + def update_encoding_params(self, max_len: int = 113, dim: int = 1024, return_offsets_mapping: bool = False): + self.max_len = max_len + self.dim = dim + self.return_offsets_mapping = return_offsets_mapping + if self.return_offsets_mapping: + assert self.model.return_offsets_mapping, ( + "T5TextEncoder needs to be initialized with return_offsets_mapping=True. " + + "Cannot turn it on after initialization." + ) + + @torch.no_grad() + def __call__(self, input_text: Union[str, list[str]]): + if isinstance(input_text, str): + input_text = [input_text] + if self.model is None: + out = (torch.zeros(1, self.max_len, self.dim), torch.zeros(1, self.max_len), None) + else: + self.model.half() + out = self.model.encode(input_text) + + output = { + "t5_text_embeddings": out[0].float(), + "t5_text_mask": out[1], + "mask": out[1], + } + + if self.return_offsets_mapping: + output["t5_offsets_mapping"] = out[2] + + return output + + +class CLIPTextEncoder(TextEncoder): + """Get CLIP encoder for obtaining text encodings. + + Args: + device (str): Device to load the model on to. + max_len (int): Max length of text encoded tokens to be returned. + dim (int): Dimension of each text encoded token. + """ + + def __init__( + self, + device: str = "cuda", + max_len: int = 77, + attr_max_len: int = 64, + dim: int = 1024, + return_offsets_mapping: bool = False, + ): + super().__init__() + self.name = "CLIP" + self.model = None + self.update_encoding_params( + max_len=max_len, dim=dim, attr_max_len=attr_max_len, return_offsets_mapping=return_offsets_mapping + ) + + def update_encoding_params( + self, max_len: int = 77, attr_max_len: int = 64, dim: int = 1024, return_offsets_mapping: bool = False + ): + self.max_len = max_len + self.attr_max_len = attr_max_len + self.dim = dim + self.return_offsets_mapping = return_offsets_mapping + + @torch.no_grad() + def __call__(self, input_text: Union[str, list[str]]): + raise NotImplementedError + if isinstance(input_text, str): + input_text = [input_text] + if self.model is None: + out = (torch.zeros(1, self.max_len, self.dim), torch.zeros(1, self.max_len), torch.zeros(1, self.dim), None) + else: + self.model.half() + out = self.model(input_text) + + output = {"clip_text_embeddings": out[0].float(), "clip_text_mask": out[1], "mask": out[1]} + + if self.return_offsets_mapping: + output["clip_offsets_mapping"] = out[2] + + return output + + +def repeat_embedding(embedding, batch_size): + return {k: repeat(v, "b ... -> (b n) ...", n=batch_size) for k, v in embedding.items()} + + +def get_text_embeddings( + text_input: Union[str, list[str]], + text_encoders: list[TextEncoder], + batch_size: int, + negative: bool = False, + override_masks_with_1s=True, +): + """Gets text embeddings of input text. + Args: + text_input (str or list of strs): Input text to be encoded. + text_encoders (list[TextEncoder]): list of TextEncoders to be applied to each input text str. + attr_encoder (CLIPTextEncoder or None): encoder for attributes. + batch_size (int): Batch size for replication of encodings. + negative (bool): True if negative prompt. + override_masks_with_1s (bool): True if you want all text encoding masks to be filled with 1s. + This is necessary for some edify image models. + """ + error_status = "" + embeddings = {} + + # Prepare suffix if negative prompt. + if negative: + key_suffix = "_neg" + else: + key_suffix = "" + + # Encode the text. + for encoder in text_encoders: + output = encoder(text_input) + + # When the text mask is all 1's, the number of tokens in the text prompt + # is >= the max tokens we can handle. We send a error message in this case, + if (output.pop("mask") == 0).sum().item() == 0: + if negative: + error_status = error_status + f"{encoder.name}: Negative prompt is too long" + else: + error_status = error_status + f"{encoder.name}: Text prompt is too long" + + for k, v in output.items(): + if "mask" in k and override_masks_with_1s: + v.fill_(1) + embeddings[k + key_suffix] = v + + # Return outputs. + embeddings = repeat_embedding(embeddings, batch_size) + + return embeddings, error_status diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/nlp/t5xxl/t5encoder.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/nlp/t5xxl/t5encoder.py new file mode 100644 index 0000000000000000000000000000000000000000..e7265ce53c76c2c48d231497755ed4ee7882b483 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/nlp/t5xxl/t5encoder.py @@ -0,0 +1,212 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os + +# pyrefly: ignore # import-error +import gdown +import torch +from loguru import logger as logging +from transformers import T5Config, T5EncoderModel, T5Tokenizer, T5TokenizerFast +from transformers import logging as transformers_logging + +# Suppresses a lot of unhelpful warnings from transformers. +transformers_logging.set_verbosity_error() + + +def download_file_from_google_drive(id, destination): + r"""Download a file from google drive. + + Args: + URL: GDrive file ID. + destination: Path to save the file. + + Returns: + + """ + # download_file(f"https://docs.google.com/uc?export=download&id={URL}", destination) + url = f"https://drive.google.com/uc?id={id}" + logging.info(f"Download {url}") + gdown.download(url, destination, quiet=False) + + +class T5Encoder(torch.nn.Module): + def __init__( + self, max_seq_len: int = 512, device: str = "cuda", return_offsets_mapping: bool = False, checkpoint: str = "" + ): + super().__init__() + self.max_seq_len = max_seq_len + self.return_offsets_mapping = return_offsets_mapping + self.model_seq_len = 512 + # Initializing T5 model + if return_offsets_mapping: + # We need fast tokenizer to return offsets mapping. + self.tokenizer = T5TokenizerFast.from_pretrained("t5-11b", model_max_length=self.model_seq_len) + else: + self.tokenizer = T5Tokenizer.from_pretrained("t5-11b", model_max_length=self.model_seq_len) + + if checkpoint: + logging.info(f"Checkpoint location: {checkpoint}") + hard_coded_encoder_weight_location = checkpoint + hard_coded_encoder_config_location = os.path.join( + os.path.dirname(os.path.abspath(__file__)), "t5encoder.json" + ) + else: + # Here is a version that only loads the encoder. + # https://drive.google.com/file/d/16Y5GoOIcZJBoSZklowVnYJxDKqOQEDUv + hard_coded_encoder_weight_url = "16Y5GoOIcZJBoSZklowVnYJxDKqOQEDUv" + hard_coded_encoder_weight_location = os.path.join(os.environ["TORCH_HOME"], "nlp/t5xxl/t5encoder.bin") + hard_coded_encoder_config_location = os.path.join( + os.path.dirname(os.path.abspath(__file__)), "t5encoder.json" + ) + os.makedirs(os.path.dirname(hard_coded_encoder_weight_location), exist_ok=True) + if not os.path.exists(hard_coded_encoder_weight_location): + download_file_from_google_drive(hard_coded_encoder_weight_url, hard_coded_encoder_weight_location) + self.model = T5EncoderModel.from_pretrained( + hard_coded_encoder_weight_location, + config=T5Config.from_json_file(hard_coded_encoder_config_location), + low_cpu_mem_usage=True, + ) + self.device = device + self.to(device) + # Below is the original version + # self.model = T5EncoderModel.from_pretrained("t5-11b", low_cpu_mem_usage=True) + + def encode(self, text_batch): + encoded = self.tokenizer.batch_encode_plus( + text_batch, + return_tensors="pt", + padding="max_length", + max_length=self.model_seq_len, + truncation=True, + return_offsets_mapping=self.return_offsets_mapping, + ) + # We expect all the processing is done in GPU. + input_ids = encoded.input_ids.to(self.device) + attn_mask = encoded.attention_mask.to(self.device) + + with torch.no_grad(): + output = self.model(input_ids=input_ids, attention_mask=attn_mask) + encoded_text = output.last_hidden_state.detach() + + encoded_text = encoded_text[:, 0 : self.max_seq_len] + attn_mask = attn_mask[:, 0 : self.max_seq_len] + for bnum in range(encoded_text.shape[0]): + nvalid_elem = attn_mask[bnum].sum().item() + encoded_text[bnum][nvalid_elem:] = 0 + + offsets_mapping = encoded["offset_mapping"] if self.return_offsets_mapping else None + + return encoded_text, attn_mask, offsets_mapping + + def _get_word_inds(self, text: str, word_ind: int, tokenizer: T5TokenizerFast): + """ + Args: + text (str): string in which we will search for the word + word_ind (int): index of the word inside of this string, if split by whitespace + tokenizer (T5TokenizerFast): The tokenizer to use on the prompts. + + Returns: + positions (torch.Tensor): position(s) of the given word in the tokenized version of text. + """ + words = text.split(" ") + tokens = tokenizer(text, return_offsets_mapping=True) + offset_mapping = tokens["offset_mapping"] + + word_to_token_map, current_word_tokens = [], [] + word_idx, char_start = 0, 0 + + for i, (start, end) in enumerate(offset_mapping[:-1]): + if start >= char_start and end <= char_start + len(words[word_idx]): + current_word_tokens.append(i) + if end >= char_start + len(words[word_idx]): + word_to_token_map.append(current_word_tokens) + current_word_tokens = [] + word_idx += 1 + char_start += len(words[word_idx - 1]) + 1 # Move to the next word + + positions = torch.tensor(word_to_token_map[word_ind]) + return positions + + def _get_replacement_mapper(self, x: str, y: str, tokenizer: T5TokenizerFast, max_len=512): + """ + Args: + x (str): Source prompt. + y (str): Target prompt. + tokenizer (T5TokenizerFast): The tokenizer to use on the prompts. Since we need offset maps, + can only proceed if using T5TokenizerFast type. + + Returns: + mapper (torch.Tensor): 2d tensor that aligns token indices from two input prompts + """ + words_x = x.split(" ") + words_y = y.split(" ") + if len(words_x) != len(words_y): + raise ValueError( + f"attention replacement edit can only be applied on prompts with the same length" + f" but prompt A has {len(words_x)} words and prompt B has {len(words_y)} words." + ) + inds_replace = [i for i in range(len(words_y)) if words_y[i] != words_x[i]] + inds_source = [ + self._get_word_inds(x, i, tokenizer) for i in inds_replace + ] # position(s) in a tokenized version of x + inds_target = [self._get_word_inds(y, i, tokenizer) for i in inds_replace] + mapper = torch.zeros((max_len, max_len)) + + i = j = 0 + cur_inds = 0 + while i < max_len and j < max_len and inds_source: + if cur_inds < len(inds_source) and inds_source[cur_inds][0] == i: + inds_source_, inds_target_ = inds_source[cur_inds], inds_target[cur_inds] + if len(inds_source_) == len(inds_target_): + mapper[inds_source_, inds_target_] = 1 + else: # if a target token maps to multiple source tokens, divide its attention + ratio = 1 / len(inds_target_) + for i_t in inds_target_: + mapper[inds_source_, i_t] = ratio + cur_inds += 1 + i += len(inds_source_) + j += len(inds_target_) + else: + mapper[i, j] = 1 + i += 1 + j += 1 + + return mapper.float().to(self.device) + + def get_replacement_mapper(self, prompts: list[str], max_len: int = 512): + """ + Creates a mapping tensor that aligns token indices from two input prompts, + indicating which tokens in prompt x should be replaced by tokens in prompt y. + + The resulting mapping tensor is used to transfer attention from the source prompt (x) + to the target prompt (y) in a model that uses cross-attention. + + Eventually, this should be able to handle many target prompts. For now, we have limited to 1. + + If unexpected arguments, should return an empty map of 0s. + + Args: + prompts (list[str]): Accepts the first as the source prompt and every subsequent as a target prompt. + max_len (int): The max prompt length, in tokens, accepted; default dimension for mapper. + + Returns: + mapper (torch.Tensor): 2d tensor that aligns token indices from two input prompts + """ + if len(prompts) != 2 or not isinstance(self.tokenizer, T5TokenizerFast): + return torch.zeros((max_len, max_len)) + src_seq, tgt_seq = prompts + mapper = self._get_replacement_mapper(src_seq, tgt_seq, self.tokenizer, max_len) + return mapper diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/res_sampler.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/res_sampler.py new file mode 100644 index 0000000000000000000000000000000000000000..6b0e7bdda7a3e58681bb21c09758828ac3bbe572 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/res_sampler.py @@ -0,0 +1,287 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +A general framework for various sampling algorithm from a diffusion model. +Impl based on +* Refined Exponential Solver (RES) in https://arxiv.org/pdf/2308.02157 +* also clude other impl, DDIM, DEIS, DPM-Solver, EDM sampler. +Most of sampling algorihtm, Runge-Kutta, Multi-step, etc, can be impl in this framework by \ + adding new step function in get_runge_kutta_fn or get_multi_step_fn. +""" + +import math +from typing import Any, Callable, List, Literal, Optional, Tuple, Union + +import attrs +import torch + +from cosmos_policy._src.imaginaire.config import make_freezable +from cosmos_policy._src.imaginaire.functional.multi_step import get_multi_step_fn, is_multi_step_fn_supported +from cosmos_policy._src.imaginaire.functional.runge_kutta import get_runge_kutta_fn, is_runge_kutta_fn_supported +from cosmos_policy._src.imaginaire.utils import log + +COMMON_SOLVER_OPTIONS = Literal["2ab", "2mid", "1euler"] + + +@make_freezable +@attrs.define(slots=False) +class SolverConfig: + is_multi: bool = False + rk: str = "2mid" + multistep: str = "2ab" + # following parameters control stochasticity, see EDM paper + # BY default, we use deterministic with no stochasticity + s_churn: float = 0.0 + s_t_max: float = float("inf") + s_t_min: float = 0.05 + s_noise: float = 1.0 + + +@make_freezable +@attrs.define(slots=False) +class SolverTimestampConfig: + nfe: int = 50 + t_min: float = 0.002 + t_max: float = 80.0 + order: float = 7.0 + is_forward: bool = False # whether generate forward or backward timestamps + + +@make_freezable +@attrs.define(slots=False) +class SamplerConfig: + solver: SolverConfig = attrs.field(factory=SolverConfig) + timestamps: SolverTimestampConfig = attrs.field(factory=SolverTimestampConfig) + sample_clean: bool = True # whether run one last step to generate clean image + + +def get_rev_ts( + t_min: float, t_max: float, num_steps: int, ts_order: Union[int, float], is_forward: bool = False +) -> torch.Tensor: + """ + Generate a sequence of reverse time steps. + + Args: + t_min (float): The minimum time value. + t_max (float): The maximum time value. + num_steps (int): The number of time steps to generate. + ts_order (Union[int, float]): The order of the time step progression. + is_forward (bool, optional): If True, returns the sequence in forward order. Defaults to False. + + Returns: + torch.Tensor: A tensor containing the generated time steps in reverse or forward order. + + Raises: + ValueError: If `t_min` is not less than `t_max`. + TypeError: If `ts_order` is not an integer or float. + """ + if t_min >= t_max: + raise ValueError("t_min must be less than t_max") + + if not isinstance(ts_order, (int, float)): + raise TypeError("ts_order must be an integer or float") + + step_indices = torch.arange(num_steps + 1, dtype=torch.float64) + time_steps = ( + t_max ** (1 / ts_order) + step_indices / num_steps * (t_min ** (1 / ts_order) - t_max ** (1 / ts_order)) + ) ** ts_order + + if is_forward: + return time_steps.flip(dims=(0,)) + + return time_steps + + +class Sampler(torch.nn.Module): + def __init__(self, cfg: Optional[SamplerConfig] = None): + super().__init__() + if cfg is None: + cfg = SamplerConfig() + self.cfg = cfg + + @torch.no_grad() + def forward( + self, + x0_fn: Callable, + x_sigma_max: torch.Tensor, + num_steps: int = 35, + sigma_min: float = 0.002, + sigma_max: float = 80, + rho: float = 7, + S_churn: float = 0, + S_min: float = 0, + S_max: float = float("inf"), + S_noise: float = 1, + solver_option: str = "2ab", + ) -> torch.Tensor: + in_dtype = x_sigma_max.dtype + + def float64_x0_fn(x_B_StateShape: torch.Tensor, t_B: torch.Tensor) -> torch.Tensor: + return x0_fn(x_B_StateShape.to(in_dtype), t_B.to(in_dtype)).to(torch.float64) + + is_multistep = is_multi_step_fn_supported(solver_option) + is_rk = is_runge_kutta_fn_supported(solver_option) + assert is_multistep or is_rk, f"Only support multistep or Runge-Kutta method, got {solver_option}" + + solver_cfg = SolverConfig( + s_churn=S_churn, + s_t_max=S_max, + s_t_min=S_min, + s_noise=S_noise, + is_multi=is_multistep, + rk=solver_option, + multistep=solver_option, + ) + timestamps_cfg = SolverTimestampConfig(nfe=num_steps, t_min=sigma_min, t_max=sigma_max, order=rho) + sampler_cfg = SamplerConfig(solver=solver_cfg, timestamps=timestamps_cfg, sample_clean=True) + + return self._forward_impl(float64_x0_fn, x_sigma_max, sampler_cfg).to(in_dtype) + + @torch.no_grad() + def _forward_impl( + self, + denoiser_fn: Callable[[torch.Tensor, torch.Tensor], torch.Tensor], + noisy_input_B_StateShape: torch.Tensor, + sampler_cfg: Optional[SamplerConfig] = None, + callback_fns: Optional[List[Callable]] = None, + ) -> torch.Tensor: + """ + Internal implementation of the forward pass. + + Args: + denoiser_fn: Function to denoise the input. + noisy_input_B_StateShape: Input tensor with noise. + sampler_cfg: Configuration for the sampler. + callback_fns: List of callback functions to be called during sampling. + + Returns: + torch.Tensor: Denoised output tensor. + """ + sampler_cfg = self.cfg if sampler_cfg is None else sampler_cfg + solver_order = 1 if sampler_cfg.solver.is_multi else int(sampler_cfg.solver.rk[0]) + num_timestamps = sampler_cfg.timestamps.nfe // solver_order + + sigmas_L = get_rev_ts( + sampler_cfg.timestamps.t_min, sampler_cfg.timestamps.t_max, num_timestamps, sampler_cfg.timestamps.order + ).to(noisy_input_B_StateShape.device) + + denoised_output = differential_equation_solver( + denoiser_fn, sigmas_L, sampler_cfg.solver, callback_fns=callback_fns + )(noisy_input_B_StateShape) + + if sampler_cfg.sample_clean: + # Override denoised_output with fully denoised version + ones = torch.ones(denoised_output.size(0), device=denoised_output.device, dtype=denoised_output.dtype) + denoised_output = denoiser_fn(denoised_output, sigmas_L[-1] * ones) + + return denoised_output + + +def fori_loop(lower: int, upper: int, body_fun: Callable[[int, Any], Any], init_val: Any) -> Any: + """ + Implements a for loop with a function. + + Args: + lower: Lower bound of the loop (inclusive). + upper: Upper bound of the loop (exclusive). + body_fun: Function to be applied in each iteration. + init_val: Initial value for the loop. + + Returns: + The final result after all iterations. + """ + val = init_val + for i in range(lower, upper): + # Add log during sampling to meet APS job health requirement of one log every 2mins + if i % 10 == 0: + log.info(f"fori_loop: {i}") + val = body_fun(i, val) + return val + + +def differential_equation_solver( + x0_fn: Callable[[torch.Tensor, torch.Tensor], torch.Tensor], + sigmas_L: torch.Tensor, + solver_cfg: SolverConfig, + callback_fns: Optional[List[Callable]] = None, +) -> Callable[[torch.Tensor], torch.Tensor]: + """ + Creates a differential equation solver function. + + Args: + x0_fn: Function to compute x0 prediction. + sigmas_L: Tensor of sigma values with shape [L,]. + solver_cfg: Configuration for the solver. + callback_fns: Optional list of callback functions. + + Returns: + A function that solves the differential equation. + """ + num_step = len(sigmas_L) - 1 + + if solver_cfg.is_multi: + update_step_fn = get_multi_step_fn(solver_cfg.multistep) + else: + update_step_fn = get_runge_kutta_fn(solver_cfg.rk) + + eta = min(solver_cfg.s_churn / (num_step + 1), math.sqrt(1.2) - 1) + + def sample_fn(input_xT_B_StateShape: torch.Tensor) -> torch.Tensor: + """ + Samples from the differential equation. + + Args: + input_xT_B_StateShape: Input tensor with shape [B, StateShape]. + + Returns: + Output tensor with shape [B, StateShape]. + """ + ones_B = torch.ones(input_xT_B_StateShape.size(0), device=input_xT_B_StateShape.device, dtype=torch.float64) + + def step_fn( + i_th: int, state: Tuple[torch.Tensor, Optional[List[torch.Tensor]]] + ) -> Tuple[torch.Tensor, Optional[List[torch.Tensor]]]: + input_x_B_StateShape, x0_preds = state + sigma_cur_0, sigma_next_0 = sigmas_L[i_th], sigmas_L[i_th + 1] + + # algorithm 2: line 4-6 + if solver_cfg.s_t_min < sigma_cur_0 < solver_cfg.s_t_max: + hat_sigma_cur_0 = sigma_cur_0 + eta * sigma_cur_0 + input_x_B_StateShape = input_x_B_StateShape + ( + hat_sigma_cur_0**2 - sigma_cur_0**2 + ).sqrt() * solver_cfg.s_noise * torch.randn_like(input_x_B_StateShape) + sigma_cur_0 = hat_sigma_cur_0 + + if solver_cfg.is_multi: + x0_pred_B_StateShape = x0_fn(input_x_B_StateShape, sigma_cur_0 * ones_B) + output_x_B_StateShape, x0_preds = update_step_fn( + input_x_B_StateShape, sigma_cur_0 * ones_B, sigma_next_0 * ones_B, x0_pred_B_StateShape, x0_preds + ) + else: + output_x_B_StateShape, x0_preds = update_step_fn( + input_x_B_StateShape, sigma_cur_0 * ones_B, sigma_next_0 * ones_B, x0_fn + ) + + if callback_fns: + for callback_fn in callback_fns: + callback_fn(**locals()) + + return output_x_B_StateShape, x0_preds + + x_at_eps, _ = fori_loop(0, num_step, step_fn, [input_xT_B_StateShape, None]) + return x_at_eps + + return sample_fn diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/modules/volume_render.py b/REGEN-main/cosmos_policy/_src/imaginaire/modules/volume_render.py new file mode 100644 index 0000000000000000000000000000000000000000..1de7ce6e5781fc862b183ba430f8b9172e3cbc90 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/modules/volume_render.py @@ -0,0 +1,142 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import torch + + +def volume_render_rays( + nerf: torch.nn.Module, + center: torch.Tensor, + ray_unit: torch.Tensor, + near: torch.Tensor, + far: torch.Tensor, + num_samples: int, + stratified: bool = False, + solid_background: bool = False, +) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]: + """Given a NeRF, volume render the color and density of the rays within the near/far distance bounds. + + Args: + nerf (torch.nn.Module): A neural field predicting the color and density from input 3D points and rays. + center (torch.Tensor [...,3]): Center of rays. + ray_unit (torch.Tensor [...,3]): Direction of rays (should be of unit norm). + near (torch.Tensor [...,1]): The near bound of the range to volume render the rays. + far (torch.Tensor [...,1]): The far bound of the range to volume render the rays. + num_samples (int): Number of sampled points. + stratified (bool): Whether to enable stratified sampling. + solid_background (bool): Whether the background is assumed to be solid. Enabling this would make the sum of + alphas along the ray to be 1. This should not be enabled if the background would be modeled separately. + + Returns: + rgb (torch.Tensor [...,3]): The volme rendered rgb values. + opacity (torch.Tensor [...,1]): The volme rendered opacity values. + weights (torch.Tensor [...,N,1]): The weights for compositing the samples. + points (torch.Tensor [...,N,3]): The sampled point locations. + dists (torch.Tensor [...,N]): The distance of the sampled points to the camera center. + """ + # Sample 3D points within the near/far range for all rays. + with torch.no_grad(): + dists = sample_dists(near, far, num_samples, stratified=stratified) # [...,N] + points = center[..., None, :] + ray_unit[..., None, :] * dists[..., None] # [...,N,3] + rays_unit = ray_unit[..., None, :].expand_as(points).contiguous() # [...,N,3] + # Feed-forward pass on the neural field. + rgbs, densities = nerf(points, rays_unit) # [...,N,3],[...,N,1] + # Volume rendering. + dist_far = None if solid_background else far[..., None] + alphas = volume_rendering_alphas(densities, dists[..., None], dist_far=dist_far) # [...,N,1] + weights = alpha_compositing_weights(alphas) # [...,N,1] + opacity = composite(1.0, weights) # [...,1] # type: ignore + rgb = composite(rgbs, weights) # [...,3] + return rgb, opacity, weights, points, dists + + +def volume_rendering_alphas( + densities: torch.Tensor, dists: torch.Tensor, dist_far: torch.Tensor | None = None +) -> torch.Tensor: + """Computes the alpha weights for volume rendering (density-based). + + Args: + densities (torch.Tensor [...,samples,1]): Density values. + dists (torch.Tensor [...,samples,1]): Distance from sampled point to camera center. + dist_far (torch.Tensor [...,1,1] | None): Farthest distance of the volume rendering range. Defaults to a + large value if not provided, which equivalently assumes full opacity at the farthest end, making + sum(alphas) = 1. (default: None) + + Returns: + alphas (torch.Tensor [...,samples,1]): The opacity of each sampled point (in [0,1]). + """ + if dist_far is None: + dist_far = torch.empty_like(dists[..., :1, :]).fill_(1e10) # [...,1,1] + dists = torch.cat([dists, dist_far], dim=-2) # [...,N+1,1] + # Volume rendering: compute rendering weights (using quadrature). + dist_intvs = dists[..., 1:, :] - dists[..., :-1, :] # [...,N,1] + sigma_delta = densities * dist_intvs # [...,N,1] + alphas = 1 - (-sigma_delta).exp_() # [...,N,1] + return alphas + + +def alpha_compositing_weights(alphas: torch.Tensor) -> torch.Tensor: + """Alpha compositing to compute the blending weights. + + Args: + alphas (torch.Tensor [...,samples,1]): The opacity of each sampled point (in [0,1]). + + Returns: + weights (torch.Tensor [...,samples,1]): The compositing weights (in [0,1]). + """ + alphas_front = torch.cat([torch.zeros_like(alphas[..., :1, :]), alphas[..., :-1, :]], dim=2) # [...,N,1] + with torch.amp.autocast("cuda", enabled=False): # Half precision may cause numerical instability. + visibility = (1 - alphas_front).cumprod(dim=-2) # [...,N,1] + weights = alphas * visibility # [...,N,1] + return weights + + +def composite(quantities: torch.Tensor, weights: torch.Tensor) -> torch.Tensor: + """Composite the samples to render the corresponding pixels. + + Args: + quantities (torch.Tensor [...,samples,channels]): The quantity to compute the weighted sum. + weights (torch.Tensor [...,samples,1]): The compositing weights (in [0,1]). + + Returns: + quantity (torch.Tensor [...,channels]): The expected (rendered) quantity. + """ + # Integrate RGB and depth weighted by probability. + quantity = (quantities * weights).sum(dim=-2) # [...,K] + return quantity + + +@torch.no_grad() +def sample_dists(near: torch.Tensor, far: torch.Tensor, num_samples: int, stratified: bool = False) -> torch.Tensor: + """Sample points along view rays given the near/far bounds. + + Args: + near (torch.Tensor [...,1]): The near bound of the range to sample the points from. + far (torch.Tensor [...,1]): The far bound of the range to sample the points from. + num_samples (int): Number of sampled points. + stratified (bool): Whether to use stratified sampling (uniform within each interval bin); otherwise, + sample at the midpoint of each interval (default: False). + + Returns: + dists (torch.Tensor [...,N]): The distance of the sampled points to the camera center. + """ + if stratified: + rands = torch.rand(*near.shape[:-1], num_samples, dtype=near.dtype, device=near.device) # [...,N] + else: + rands = torch.empty(*near.shape[:-1], num_samples, dtype=near.dtype, device=near.device).fill_(0.5) # [...,N] + base = torch.arange(num_samples, dtype=near.dtype, device=near.device).repeat(*near.shape[:-1], 1) # [...,N] + rands = (rands + base) / num_samples # [...,N] + dists = rands * (far - near) + near # [...,N] + return dists diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/serialization.py b/REGEN-main/cosmos_policy/_src/imaginaire/serialization.py new file mode 100644 index 0000000000000000000000000000000000000000..01ae90589b12bc8264880caa88730210ffb400ce --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/serialization.py @@ -0,0 +1,379 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import abc +import importlib +import json +import os +from collections.abc import Callable as Callable2 +from dataclasses import fields, is_dataclass +from types import UnionType +from typing import Any, List, Optional, TypeVar, Union, get_args, get_origin + +import attrs +import torch +import yaml +from omegaconf import DictConfig, ListConfig, OmegaConf + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall, LazyDict, instantiate +from cosmos_policy._src.imaginaire.lazy_config.lazy import get_default_params + +T = TypeVar("T") + + +def from_dict( + x: dict, clazz: str | type | None = None, force_construct_target: bool | None = None, field_name: str = "" +) -> T: ... +def to_dict(x: T, field_name: str = "", hydra_compat: bool = True) -> dict: ... +def from_yaml(path: str | None = None, clazz: type | None = None, file_like_or_str=None) -> T: + if path: + assert os.path.exists(path), f"{path} does not exist" + with open(path) as in_f: + return from_dict(yaml.safe_load(in_f), clazz=clazz) + elif file_like_or_str: + return from_dict(yaml.safe_load(file_like_or_str), clazz=clazz) + else: + raise ValueError("expected file_like_or_str or path to not be None") + + +def to_yaml(config: T, out_path: str | None = None) -> dict | None: + config_dict = to_dict(config) + if out_path is not None: + with open(out_path, "w") as f: + yaml.dump(config_dict, f) + else: + return yaml.dump(config_dict) + + +def load_callable(name: str) -> Callable2 | None: + if not name: + return None + + idx = name.rfind(".") + assert idx != -1, "expected ." + module_name = name[0:idx] + fn_name = name[idx + 1 :] + mod = importlib.import_module(module_name) + return getattr(mod, fn_name) + + +def maybe_load_callable(name: str | Callable2 | None) -> Callable2 | None: + if isinstance(name, str): + return load_callable(name) + + return name + + +def maybe_idx(x: Any, idx: int) -> Any: + if idx < 0 or idx >= len(x): + return None + return x[idx] + + +def is_attrs(x: Any) -> bool: + return hasattr(x, "__attrs_attrs__") + + +def to_qualitified_name(x: Any) -> str: + result = "" + if x.__module__: + result += x.__module__ + "." + result += x.__qualname__ + return result + + +def is_optional(x: type) -> bool: + origin = get_origin(x) + args = get_args(x) + return origin is Optional or (origin in (Union, UnionType) and len(args) == 2 and type(None) in args) + + +def _to_dict_value(x: T, field_type: type, metadata: dict, field_name: str = ""): + t = type(x) + + # attrs specific + if x is attrs.NOTHING or x is None: + return None + # torch specifics + elif field_type in (torch.memory_format, torch.dtype): + return str(x) + # i4 specific types + elif field_type == LazyCall: + result = _to_dict_value(x, field_type._target, metadata, field_name) + return result + elif field_type in (DictConfig, LazyDict): + if "_target_" in x: + default_params = get_default_params(x["_target_"]) + for default_key, default_v in default_params.items(): + if default_key not in x: + x[default_key] = default_v + result = _to_dict_value(x, dict, metadata, field_name) + object_type = getattr(x._metadata, "object_type", None) + if object_type and (is_dataclass(object_type) or is_attrs(object_type)): + result.setdefault("_target_", to_qualitified_name(object_type)) + return result + elif field_type == ListConfig: + return _to_dict_value(x, list, metadata, field_name) + # general python types + dataclasses + attrs + # * meta types + elif field_type == type or field_type == abc.ABCMeta: + return to_qualitified_name(x) + elif get_origin(field_type) is type: + return to_qualitified_name(x) + elif callable(x) or get_origin(field_type) is Callable2: + if callable(x): + return to_qualitified_name(x) + else: + assert isinstance(x, str), f"{x.__class__=}" + return x + elif is_dataclass(t) or is_attrs(t): + return to_dict(x, field_name=field_name) + # * built-in composites types + elif is_optional(field_type): + return _to_dict_value(x, get_args(field_type)[0], metadata) + elif get_origin(field_type) in (Union, UnionType): + raise AssertionError("unions are not implemented yet!") + # * primitives + elif t in (dict,) or field_type in (dict,) or get_origin(field_type) in (dict,): + return { + _to_dict_value( + k, + maybe_idx(get_args(field_type), 0) or type(k), + metadata, + field_name=f"{field_name}.{k}.key", + ): _to_dict_value( + v, + maybe_idx(get_args(field_type), 1) or type(v), + metadata, + field_name=f"{field_name}.{k}", + ) + for k, v in x.items() + } + elif ( + t + in ( + tuple, + list, + ) + or field_type + in ( + tuple, + list, + ) + or get_origin(field_type) in (tuple, list) + ): + if field_type is None or field_type not in ( + tuple, + list, + ): + field_type = list + + return field_type( + [ + _to_dict_value(xx, maybe_idx(get_args(field_type), 0) or type(xx), metadata, field_name + f"[{i}]") + for i, xx in enumerate(x) + ] + ) + elif field_type in (int, str, float, bool): + result = field_type(x) + return result + else: # catch all for everything else + return x + + +def to_dict(x: T, field_name: str = "", hydra_compat: bool = True) -> dict: + if is_dataclass(x): + result = {} + if hydra_compat: + result["_target_"] = to_qualitified_name(x.__class__) + for f in fields(x): + # NOTE: defaults are unnecessary to encode + if hydra_compat and f.name == "defaults": + continue + result[f.name] = _to_dict_value( + x.__dict__[f.name], + f.type, + f.metadata, + field_name=field_name + f".{f.name}" if field_name else f.name, + ) + return result + elif is_attrs(x): + # references: + # - https://github.com/python-attrs/attrs/blob/main/src/attr/_funcs.py + attrs.resolve_types(x.__class__) + + result = {} + if hydra_compat: + result["_target_"] = to_qualitified_name(x.__class__) + for f in attrs.fields(x.__class__): + # NOTE: defaults are unnecessary to encode + if hydra_compat and f.name == "defaults": + continue + result[f.name] = _to_dict_value( + getattr(x, f.name), + f.type, + f.metadata, + field_name=field_name + f".{f.name}" if field_name else f.name, + ) + return result + + +def _from_dict_value( + x: T, + field_type: type, + concrete_type: type, + field_name: str, + force_construct_target: bool | None = None, +): + is_dc_type = is_dataclass(field_type) + is_attrs_type = is_attrs(field_type) + origin = get_origin(field_type) or field_type + args = get_args(field_type) + + if x is None: + return None + elif field_type in (torch.memory_format, torch.dtype): + return maybe_load_callable(x) + elif field_type == LazyCall: + return _from_dict_value(x, field_type._target, concrete_type, field_name=field_name) + elif is_dc_type or is_attrs_type: + if concrete_type == str: + assert isinstance(x, str) + if x.endswith(".json"): + json_value = json.loads(x) + return from_dict( + json_value, field_type, force_construct_target=force_construct_target, field_name=field_name + ) + elif x.endswith(".yaml"): + yaml_value = yaml.safe_load(x) + return from_dict( + yaml_value, field_type, force_construct_target=force_construct_target, field_name=field_name + ) + else: + raise AssertionError(f"unexpected string: {x}") + else: + assert not isinstance(x, str) + return from_dict(x, field_type, field_name=field_name) + elif field_type in (DictConfig, LazyDict) or origin in (dict,): + # NOTE: _recursive_ is the name of the flag for this behaviour + construct_target = x.get("_recursive_", field_type == DictConfig) + if force_construct_target is not None: + construct_target = force_construct_target + + target_value = x.get("_target_") + target_cls = maybe_load_callable(target_value) + + if target_value and construct_target and (is_dataclass(target_cls) or is_attrs(target_cls)): + result = from_dict(x, target_cls, force_construct_target=force_construct_target, field_name=field_name) + else: + result = { + _from_dict_value( + k, + maybe_idx(get_args(field_type), 0) or type(k), + type(k), + field_name=f"{field_name}.{k}.key", + force_construct_target=construct_target, + ): _from_dict_value( + v, + maybe_idx(get_args(field_type), 1) or type(v), + type(v), + field_name=f"{field_name}.{k}", + force_construct_target=construct_target, + ) + for k, v in x.items() + } + if field_type in (DictConfig, LazyDict): + result = OmegaConf.structured(result, flags={"allow_objects": True}) + if construct_target: + result = instantiate(result) + if "_target_" in result: + result["_target_"] = maybe_load_callable(result["_target_"]) + elif construct_target and target_cls: # instantiate a regular class from a dict + special_keys = { + "_target_", + "_recursive_", + "_convert_", + "_args_", + "_kwargs_", + } + constructable_items = { + k: v for k, v in result.items() if not (isinstance(k, str) and k in special_keys) + } + result = target_cls(**constructable_items) + return result + elif field_type is ListConfig or origin in ( + list, + List, + ): + return [ + _from_dict_value( + xx, maybe_idx(get_args(field_type), 0) or type(xx), type(xx), field_name=f"{field_type}[{i}]" + ) + for i, xx in enumerate(x) + ] + elif is_optional(field_type): + return _from_dict_value(x, args[0], type(x), field_name=field_name) + elif origin in (Union, UnionType): + raise AssertionError("unions are not implemented yet!") + elif origin is Callable2 or origin is type: + return maybe_load_callable(x) + elif field_type in (int, float, str, bool): + return x + elif field_type is type(None) or field_type == Any: # no typing + return x + else: + raise TypeError( + f"unexpected type: {field_type} (origin={origin}, concrete_type={concrete_type}, args={args}, x={x})" + ) + + +def from_dict( + x: dict, clazz: type | None = None, force_construct_target: bool | None = None, field_name: str = "" +) -> T: + if clazz is None: + assert "_target_" in x + clazz = maybe_load_callable(x["_target_"]) + + assert is_dataclass(clazz) or is_attrs(clazz), f"{clazz} is not a dataclass or attrs" + if is_dataclass(clazz): + construct_args = {} + for f in fields(clazz): + if f.name in x: + construct_args[f.name] = _from_dict_value( + x[f.name], + f.type, + type(x[f.name]), + field_name=field_name + "." + f.name if field_name else f.name, + force_construct_target=force_construct_target, + ) + elif is_optional(f.type): + construct_args[f.name] = None + return clazz(**construct_args) + elif is_attrs(clazz): + attrs.resolve_types(clazz) + + construct_args = {} + for f in attrs.fields(clazz): + if f.name in x: + construct_args[f.name] = _from_dict_value( + x[f.name], + f.type, + type(x[f.name]), + field_name=field_name + "." + f.name if field_name else f.name, + force_construct_target=force_construct_target, + ) + elif is_optional(f.type): + construct_args[f.name] = None + return clazz(**construct_args) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/trainer.py b/REGEN-main/cosmos_policy/_src/imaginaire/trainer.py new file mode 100644 index 0000000000000000000000000000000000000000..3dddedf6c6d350595274a57d8693b3fc3e14af9f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/trainer.py @@ -0,0 +1,353 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import functools +import inspect +import os +import signal + +import torch +import torch.distributed as dist +import torch.utils.data + +from cosmos_policy._src.imaginaire.flags import INTERNAL +from cosmos_policy._src.imaginaire.utils.context_managers import distributed_init +from cosmos_policy._src.imaginaire.utils.profiling import maybe_enable_memory_snapshot, maybe_enable_profiling + +try: + from megatron.core import parallel_state + + USE_MEGATRON = True +except ImportError: + USE_MEGATRON = False + print("Megatron-core is not installed.") + + +from cosmos_policy._src.imaginaire.lazy_config import LazyConfig, instantiate +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import callback, distributed, ema, log, misc +from cosmos_policy._src.imaginaire.utils.checkpointer import Checkpointer +from cosmos_policy._src.imaginaire.utils.misc import StragglerDetectorV2 + + +class ImaginaireTrainer: + """The base trainer class of Imaginaire. + + All trainers in Imaginaire should inherit ImaginaireTrainer. It contains the basic functionality for model training + (particularly suited for large-scale training), including data parallel (DDP/FSDP), model weight average (EMA), + mixed-precision training (fp16/bf16). + + Attributes: + checkpointer (Checkpointer): checkpointer object to save/load model weights and optimizer states. + training_timer (misc.Timer): Timer object to time code blocks and functions. + """ + + def __init__(self, config): + """Constructor of the trainer. + + Args: + config (Config): The config object for the Imaginaire codebase. + """ + super().__init__() + self.config = config + # Set up the distributed computing environment. + with distributed_init(): + distributed.init() + # Set up parallel states. + if hasattr(config.model, "context_parallel_size"): + if config.model_parallel.context_parallel_size > 1: + raise ValueError( + "Both config.model.context_parallel_size and config.model_parallel.context_parallel_size are set. " + "config.model.context_parallel_size is deprecated. Please only set config.model_parallel.context_parallel_size." + ) + else: + log.critical( + "Using deprecated config.model.context_parallel_size. Please use config.model_parallel.context_parallel_size instead." + ) + config.model_parallel.context_parallel_size = config.model.context_parallel_size + if USE_MEGATRON: + if ( + "create_gloo_process_groups" + in inspect.signature(parallel_state.initialize_model_parallel).parameters + ): + parallel_state.initialize_model_parallel( + pipeline_model_parallel_size=config.model_parallel.pipeline_model_parallel_size, + tensor_model_parallel_size=config.model_parallel.tensor_model_parallel_size, + context_parallel_size=config.model_parallel.context_parallel_size, + create_gloo_process_groups=False, + ) + else: + parallel_state.initialize_model_parallel( + pipeline_model_parallel_size=config.model_parallel.pipeline_model_parallel_size, + tensor_model_parallel_size=config.model_parallel.tensor_model_parallel_size, + context_parallel_size=config.model_parallel.context_parallel_size, + ) + # `config.model_parallel.sequence_parallel` is a bool that indicates whether to use sequence parallelism. + # It is not part of the original `parallel_state` API, so we need to set it manually. + parallel_state.sequence_parallel = config.model_parallel.sequence_parallel + if parallel_state.sequence_parallel: + os.environ["CUDA_DEVICE_MAX_CONNECTIONS"] = "1" + + # Create the local job directory, save the config file, and pipe to a local log. + if distributed.is_rank0(): + os.makedirs(config.job.path_local, exist_ok=True) + # Save the config as .pkl for reproducibility. + LazyConfig.save_pkl(config, f"{config.job.path_local}/config.pkl") + # Save the config as .yaml for reading or parsing experiment hyperparameters. + LazyConfig.save_yaml(config, f"{config.job.path_local}/config.yaml") + dist.barrier() + if INTERNAL: + log.init_loguru_file(f"{config.job.path_local}/stdout.log") + if distributed.is_rank0(): + # Print important environment variables and the effective config. + log.info("Config:\n" + config.pretty_print(use_color=True)) + misc.print_environ_variables(["TORCH_HOME", "IMAGINAIRE_OUTPUT_ROOT", "ENABLE_ONELOGGER"]) + else: + misc.print_environ_variables(["HF_HOME", "IMAGINAIRE_OUTPUT_ROOT"]) + # Set the random seed. If multi-GPU, different ranks are set with different seeds. + misc.set_random_seed(seed=config.trainer.seed, by_rank=True) + # Initialize cuDNN. + torch.backends.cudnn.deterministic = config.trainer.cudnn.deterministic + torch.backends.cudnn.benchmark = config.trainer.cudnn.benchmark + # Floating-point precision settings. + torch.backends.cudnn.allow_tf32 = torch.backends.cuda.matmul.allow_tf32 = True + # Initialize the callback functions. + self.callbacks = callback.CallBackGroup(config=config, trainer=self) + # Initialize the model checkpointer. + if config.checkpoint.type is None: + self.checkpointer = Checkpointer(config.checkpoint, config.job, callbacks=self.callbacks) + else: + self.checkpointer: Checkpointer = instantiate( + config.checkpoint.type, config.checkpoint, config.job, callbacks=self.callbacks + ) + # Initialize the timer for speed benchmarking. + self.training_timer = misc.TrainingTimer() + # Initialize Straggler Detection + self.straggler_detector = StragglerDetectorV2( + enabled=self.config.trainer.straggler_detection.enabled, + report_freq=self.config.trainer.straggler_detection.report_freq, + profile_freq=self.config.trainer.straggler_detection.profile_freq, + max_diff=self.config.trainer.straggler_detection.max_diff, + raise_error=self.config.trainer.straggler_detection.raise_error, + ) + self.straggler_detector.initialize() + # Send a TimeoutError if a training step takes over timeout_period seconds. + signal.signal(signal.SIGALRM, functools.partial(misc.timeout_handler, config.trainer.timeout_period)) # type: ignore + + def train( + self, + model: ImaginaireModel, + dataloader_train: torch.utils.data.DataLoader, + dataloader_val: torch.utils.data.DataLoader, + ) -> None: + """The training function. + + Args: + model (ImaginaireModel): The PyTorch model. + dataloader_train (torch.utils.data.DataLoader): The training data loader. + dataloader_val (torch.utils.data.DataLoader): The validation data loader. + """ + # Leaving this for backward compability for now, but we can think about moving this to model.on_train_start for all models. + model = model.to("cuda", memory_format=self.config.trainer.memory_format) # type: ignore + model.on_train_start(self.config.trainer.memory_format) + + # Initialize the optimizer, scheduler, and grad_scaler. + self.callbacks.on_optimizer_init_start() + optimizer, scheduler = model.init_optimizer_scheduler(self.config.optimizer, self.config.scheduler) + grad_scaler = torch.amp.GradScaler("cuda", **self.config.trainer.grad_scaler_args) + self.callbacks.on_optimizer_init_end() + # Load the model checkpoint and get the starting iteration number. + iteration = self.checkpointer.load(model, optimizer, scheduler, grad_scaler) + grad_accum_iter = 0 + log.critical(f"Distributed parallelism mode: {self.config.trainer.distributed_parallelism}") + if self.config.trainer.distributed_parallelism == "ddp": + # Create a DDP model wrapper. + model_ddp = distributed.parallel_model_wrapper(self.config.trainer.ddp, model) + elif self.config.trainer.distributed_parallelism == "fsdp": + model_ddp = model + else: + raise ValueError(f"Unknown distributed parallelism mode: {self.config.trainer.distributed_parallelism}") + + log.info("Starting training...") + self.callbacks.on_train_start(model, iteration=iteration) + # Initial validation. + if self.config.trainer.run_validation and iteration == 0 and self.config.trainer.run_validation_on_start: + self.validate(model, dataloader_val, iteration=iteration) + _end_training = False + with ( + maybe_enable_profiling(self.config, global_step=iteration) as torch_profiler, + maybe_enable_memory_snapshot(self.config, global_step=iteration) as memory_profiler, + ): + while True: + dataloader_train_iter = iter(dataloader_train) + while True: + self.callbacks.on_before_dataloading(iteration) + try: + with ( + self.training_timer("dataloader_train"), + self.straggler_detector.profile_section( + "dataloading", + self.config.trainer.straggler_detection.analyze_dataloading, + profile_cuda=False, + ), + ): + data_batch = next(dataloader_train_iter) + except StopIteration: + break + finally: + self.callbacks.on_after_dataloading(iteration) + # If max_iter is reached, exit the training loop. + if iteration >= self.config.trainer.max_iter: + _end_training = True + break + # Move all tensors in the data batch to GPU device. + data_batch = misc.to(data_batch, device="cuda") + # The actual training step. + self.callbacks.on_training_step_start(model, data_batch, iteration=iteration) + self.callbacks.on_training_step_batch_start(model, data_batch, iteration=iteration) + if not model.training: + model_ddp.train() + assert model_ddp.training, "model_ddp is not in training mode." + assert model.training, "model is not in training mode." + output_batch, loss, grad_accum_iter = self.training_step( + model_ddp, + optimizer, + scheduler, + grad_scaler, + data_batch, + iteration=iteration, + grad_accum_iter=grad_accum_iter, + ) + self.callbacks.on_training_step_batch_end( + model, data_batch, output_batch, loss, iteration=iteration + ) + # If the gradients are still being accumulated, continue to load the next training batch. + if grad_accum_iter != 0: + continue + # Do the following when an actual optimizer (update) step has been made. + iteration += 1 + # Save checkpoint. + if iteration % self.config.checkpoint.save_iter == 0: + self.checkpointer.save(model, optimizer, scheduler, grad_scaler, iteration=iteration) + self.callbacks.on_training_step_end(model, data_batch, output_batch, loss, iteration=iteration) + # Validation. + if self.config.trainer.run_validation and iteration % self.config.trainer.validation_iter == 0: + self.validate(model, dataloader_val, iteration=iteration) + # This iteration is successful; reset the timeout signal. + signal.alarm(self.config.trainer.timeout_period) + self.straggler_detector.generate_report(iteration) + if torch_profiler: + torch_profiler.step() + if memory_profiler: + memory_profiler.step() + if _end_training: + break + log.success("Done with training.") + if iteration % self.config.checkpoint.save_iter != 0: + self.checkpointer.save(model, optimizer, scheduler, grad_scaler, iteration=iteration) + self.callbacks.on_train_end(model, iteration=iteration) + self.checkpointer.finalize() + distributed.barrier() + self.callbacks.on_app_end() + + def training_step( + self, + model_ddp: torch.nn.Module | distributed.DistributedDataParallel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + data: dict[str, torch.Tensor], + iteration: int = 0, + grad_accum_iter: int = 0, + ) -> tuple[dict[str, torch.Tensor], torch.Tensor, int]: + """The training step. + + Args: + model_ddp (torch.nn.Module | distributed.DistributedDataParallel): The model with a DDP wrapper or, the bare + module, depending on whether distributed training is enabled or not. + optimizer (torch.optim.Optimizer): The model optimizer. + scheduler (torch.optim.lr_scheduler.LRScheduler): The optimization scheduler. + grad_scaler (torch.amp.GradScaler): The gradient scaler (for mixed precision training). + data (dict[str, torch.Tensor]): Data batch (dictionary of tensors). + iteration (int): Current iteration number. + grad_accum_iter (int): Number of gradient accumulation iterations. + + Returns: + output (dict[str, torch.Tensor]): The model output from the training data batch (dictionary of tensors). + loss (torch.Tensor): The total loss of the training data batch. + """ + # Only let DDP sync gradient at the last iteration of the gradient accumulation window + with distributed.ddp_sync_grad(model_ddp, grad_accum_iter == self.config.trainer.grad_accum_iter - 1): + self.callbacks.on_before_forward(iteration=iteration) + with self.training_timer("forward"): + with self.straggler_detector.profile_section( + "fwd", self.config.trainer.straggler_detection.analyze_forward + ): + output_batch, loss = model_ddp.training_step(data, iteration) + self.callbacks.on_after_forward(iteration=iteration) + self.callbacks.on_before_backward(model_ddp, loss, iteration=iteration) + with self.training_timer("backward"): + with self.straggler_detector.profile_section( + "bwd", self.config.trainer.straggler_detection.analyze_backward + ): + loss_scaled = grad_scaler.scale(loss / self.config.trainer.grad_accum_iter) + loss_scaled.backward() + if self.config.trainer.distributed_parallelism == "ddp": + model_ddp.module.on_after_backward() + else: + model_ddp.on_after_backward() + self.callbacks.on_after_backward(model_ddp, iteration=iteration) + grad_accum_iter += 1 + if grad_accum_iter == self.config.trainer.grad_accum_iter: + with self.training_timer("optimizer_step"): + with self.straggler_detector.profile_section( + "opt", self.config.trainer.straggler_detection.analyze_optimizer + ): + self.callbacks.on_before_optimizer_step( + model_ddp, optimizer, scheduler, grad_scaler, iteration=iteration + ) + grad_scaler.step(optimizer) + grad_scaler.update() + scheduler.step() + self.callbacks.on_before_zero_grad(model_ddp, optimizer, scheduler, iteration=iteration) + if self.config.trainer.distributed_parallelism == "ddp": + model_ddp.module.on_before_zero_grad(optimizer, scheduler, iteration=iteration) + else: + model_ddp.on_before_zero_grad(optimizer, scheduler, iteration=iteration) + optimizer.zero_grad(set_to_none=True) + grad_accum_iter = 0 + return output_batch, loss, grad_accum_iter + + @torch.no_grad() + def validate(self, model: ImaginaireModel, dataloader_val: torch.utils.data.DataLoader, iteration: int = 0) -> None: + """Validate on the full validation dataset. + + Args: + model (ImaginaireModel): The PyTorch model. + dataloader_val (torch.utils.data.DataLoader): The validation data loader. + iteration (int): Current iteration number. + """ + self.callbacks.on_validation_start(model, dataloader_val, iteration=iteration) + model.eval() + # Evaluate on the full validation set. + with ema.ema_scope(model, enabled=model.config.ema.enabled): + for val_iter, data_batch in enumerate(dataloader_val): + if self.config.trainer.max_val_iter is not None and val_iter >= self.config.trainer.max_val_iter: + break + data_batch = misc.to(data_batch, device="cuda") + self.callbacks.on_validation_step_start(model, data_batch, iteration=iteration) + output_batch, loss = model.validation_step(data_batch, iteration) + self.callbacks.on_validation_step_end(model, data_batch, output_batch, loss, iteration=iteration) + self.callbacks.on_validation_end(model, iteration=iteration) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/callback.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/callback.py new file mode 100644 index 0000000000000000000000000000000000000000..1207457c3be59e1091a3cf7d2bb3a1c85ed2883e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/callback.py @@ -0,0 +1,606 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import time +import warnings +from typing import TYPE_CHECKING, Any, Callable, Optional + +import omegaconf +import torch +import torch.distributed as dist +import torch.utils.data +import tqdm +import wandb + +from cosmos_policy._src.imaginaire.lazy_config import instantiate +from cosmos_policy._src.imaginaire.utils import distributed, log, misc, wandb_util +from cosmos_policy._src.imaginaire.utils.misc import get_local_tensor_if_DTensor + +try: + from megatron.core import parallel_state +except ImportError: + parallel_state = None + print("Megatron-core is not installed.") + + +if TYPE_CHECKING: + from cosmos_policy._src.imaginaire.config import Config + from cosmos_policy._src.imaginaire.model import ImaginaireModel + from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer + + +class CallBackGroup: + """A class for hosting a collection of callback objects. + + It is used to execute callback functions of multiple callback objects with the same method name. + When callbackgroup.func(args) is executed, internally it loops through the objects in self._callbacks and runs + self._callbacks[0].func(args), self._callbacks[1].func(args), etc. The method name and arguments should match. + + Attributes: + _callbacks (list[Callback]): List of callback objects. + """ + + def __init__(self, config: Config, trainer: ImaginaireTrainer) -> None: + """Initializes the list of callback objects. + + Args: + config (Config): The config object for the Imaginaire codebase. + trainer (ImaginaireTrainer): The main trainer. + """ + self._callbacks = [] + callback_configs = config.trainer.callbacks + if callback_configs: + if isinstance(callback_configs, list) or isinstance(callback_configs, omegaconf.listconfig.ListConfig): + warnings.warn( + "The 'config.trainer.callbacks' parameter should be a dict instead of a list. " + "Please update your code", + DeprecationWarning, + stacklevel=2, + ) + callback_configs = {f"callback_{i}": v for i, v in enumerate(callback_configs)} + for callback_name, current_callback_cfg in callback_configs.items(): + if "_target_" not in current_callback_cfg: + log.critical( + f"Callback {callback_name} is missing the '_target_' field. \n SKip {current_callback_cfg}" + ) + continue + log.critical(f"Instantiating callback {callback_name}: {current_callback_cfg}") + _callback = instantiate(current_callback_cfg) + assert isinstance(_callback, Callback), f"{current_callback_cfg} is not a valid callback." + _callback.config = config + _callback.trainer = trainer + self._callbacks.append(_callback) + + def __getattr__(self, method_name: str) -> Callable: + """Loops through the callback objects to call the corresponding callback function. + + Args: + method_name (str): Callback method name. + """ + + def multi_callback_wrapper(*args, **kwargs) -> None: + for callback in self._callbacks: + assert hasattr(callback, method_name) + method = getattr(callback, method_name) + assert callable(method) + _ = method(*args, **kwargs) + + return multi_callback_wrapper + + +class Callback: + """The base class for all callbacks. + + All callbacks should inherit from this class and adhere to the established method names and signatures. + """ + + def __init__(self, config: Optional["Config"] = None, trainer: Optional["ImaginaireTrainer"] = None): + """Initializes a Callback object. + + Args: + config (Optional[Config]): The configuration object for the Imaginaire codebase, if available. + trainer (Optional[ImaginaireTrainer]): The main trainer handling the training loop, if available. + + Notes: + The config and trainer parameters are optional to maintain backward compatibility. + In future releases, these parameters will be removed. Upon using these parameters, a deprecation + warning will be issued. + + """ + if config is not None or trainer is not None: + warnings.warn( + "The 'config' and 'trainer' parameters are deprecated and will be removed in a future release. " + "Please update your code to create Callback instances without these parameters.", + DeprecationWarning, + stacklevel=2, + ) + del config, trainer + + def on_train_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + pass + + def on_training_step_start(self, model: ImaginaireModel, data: dict[str, torch.Tensor], iteration: int = 0) -> None: + """ + Called before the training step, for each batch. This is paired with on_training_step_end() but note that + when using gradient accumulation, while on_training_step_end() is only called when the optimizer is updated, + this function is called for every batch. + Use on_training_step_batch_start and on_training_step_batch_end if you need callbacks that are called + for every batch, albeit with the same iteration number. + FIXME - should this either be deprecated, or called only when a new training step is started after having updated + the optimizer? + """ + pass + + def on_training_step_batch_start( + self, model: ImaginaireModel, data: dict[str, torch.Tensor], iteration: int = 0 + ) -> None: + """ + Called before the training step, for each batch, similarly to on_training_step_start(). This function is paired with + on_training_step_batch_end(), and both functions are called for every batch even when using gradient accumulation. + Note that the iteration is only updated when the optimizer is updated, and therefore it may be the same for multiple invocations. + """ + pass + + def on_before_forward(self, iteration: int = 0) -> None: + pass + + def on_after_forward(self, iteration: int = 0) -> None: + pass + + def on_before_backward( + self, model_ddp: distributed.DistributedDataParallel, loss: torch.Tensor, iteration: int = 0 + ) -> None: + pass + + def on_after_backward(self, model_ddp: distributed.DistributedDataParallel, iteration: int = 0) -> None: + pass + + def on_before_dataloading(self, iteration: int = 0) -> None: + pass + + def on_after_dataloading(self, iteration: int = 0) -> None: + pass + + def on_optimizer_init_start(self) -> None: + pass + + def on_optimizer_init_end(self) -> None: + pass + + def on_before_optimizer_step( + self, + model_ddp: distributed.DistributedDataParallel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int = 0, + ) -> None: + pass + + def on_before_zero_grad( + self, + model_ddp: distributed.DistributedDataParallel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + iteration: int = 0, + ) -> None: + pass + + def on_training_step_batch_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + """ + Called at the end of a training step for every batch even when using gradient accumulation. + This is paired with on_training_step_batch_start(). Note that the iteration is only updated when the optimizer is updated, + and therefore it may be the same for multiple batches. + """ + pass + + def on_training_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + """ + Called at the end of a training step, but note that when using gradient accumulation, this is only called + when the optimizer is updated, and the iteration incremented, whereas on_training_step_start is called every time. + Use on_training_step_batch_start and on_training_step_batch_end if you need callbacks that are called + for every batch. + """ + pass + + def on_validation_start( + self, model: ImaginaireModel, dataloader_val: torch.utils.data.DataLoader, iteration: int = 0 + ) -> None: + pass + + def on_validation_step_start( + self, model: ImaginaireModel, data: dict[str, torch.Tensor], iteration: int = 0 + ) -> None: + pass + + def on_validation_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + pass + + def on_validation_end(self, model: ImaginaireModel, iteration: int = 0) -> None: + pass + + def on_load_checkpoint_start(self, model: ImaginaireModel) -> None: + pass + + def on_load_checkpoint_end( + self, model: ImaginaireModel, iteration: int = 0, checkpoint_path: Optional[str] = None + ) -> None: + pass + + def on_load_checkpoint(self, model: ImaginaireModel, state_dict: dict[Any]) -> None: + """ + Called when checkpoint loading is about to start, but after on_save_checkpoint_start(). + FIXME - why do we need this callback, can't we just use on_save_checkpoint_start()? + """ + pass + + def on_save_checkpoint_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + """ + Called when checkpoint saving is about to start. + """ + pass + + def on_save_checkpoint_end(self, model: ImaginaireModel, iteration: int = 0) -> None: + """ + Called when the synchronous part of checkpointing is finished, this function can be used + along with on_save_checkpoint_start() to measure the exposed (synchronous) checkpoint time. + Note that for asynchronous checkpoint, the checkpoint may still be ongoing, so this function + does not mean the checkpoint is finished for the asynchronous case, use on_save_checkpoint_success() + for that. + """ + pass + + def on_save_checkpoint_success(self, iteration: int = 0, elapsed_time: float = 0) -> None: + """ + Called when checkpoint saving is fully finished, and succeeded. Not called if checkpoint failed. + For synchronous checkpoint, it is called at the same time as on_save_checkpoint_end(), but for asynchronous + checkpoint, it is called after the asynchronous part has also finished. For checkpointers with out-of-process + checkpointing, this function is called as soon as the notification is received from the checkpointer process, + which may not be immediately after the checkpoint has completed but later on. Therefore, if you need to measure + the full checkpoint duration for the asynchronous part, use the elapsed_time parameter, do not measure it directly + as this would be a significant overestimate. + """ + pass + + def on_save_checkpoint(self, model: ImaginaireModel, state_dict: dict[Any]) -> None: + pass + + def on_train_end(self, model: ImaginaireModel, iteration: int = 0) -> None: + pass + + def on_app_end(self) -> None: + pass + + +class EMAModelCallback(Callback): + """The callback class for tracking EMA model weights.""" + + def on_train_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + # Set up the EMA model weight tracker. + if model.config.ema.enabled: + assert hasattr(model, "ema"), "EMA should be initialized from ImaginaireModel" + # EMA model must be kept in FP32 precision. + model.ema = model.ema.to(dtype=torch.float32) + else: + assert not hasattr(model, "ema"), "There should be no EMA initialized." + + def on_training_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + # Update the EMA model with the new regular weights. + if model.config.ema.enabled: + model.ema.update_average(model, iteration) + + +class ProgressBarCallback(Callback): + """The callback class for visualizing the training/validation progress bar in the console.""" + + @distributed.rank0_only + def on_train_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + self.train_pbar = tqdm.trange(self.config.trainer.max_iter, initial=iteration, desc="Training") + + @distributed.rank0_only + def on_training_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + self.train_pbar.update() + + @distributed.rank0_only + def on_validation_start( + self, model: ImaginaireModel, dataloader_val: torch.utils.data.DataLoader, iteration: int = 0 + ) -> None: + if self.config.trainer.max_val_iter is not None: + num_iter = self.config.trainer.max_val_iter + else: + num_iter = len(dataloader_val) + assert num_iter is not None and num_iter > 0, f"Invalid number of validation iterations: {num_iter}" + self.val_pbar = tqdm.trange(num_iter, desc="Validating", position=1, leave=False) + + @distributed.rank0_only + def on_validation_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + self.val_pbar.update() + + @distributed.rank0_only + def on_validation_end(self, model: ImaginaireModel, iteration: int = 0) -> None: + self.val_pbar.close() + + @distributed.rank0_only + def on_train_end(self, model: ImaginaireModel, iteration: int = 0) -> None: + self.trainer.checkpointer.finalize() + self.train_pbar.close() + + +class IterationLoggerCallback(Callback): + """The callback class for visualizing the training/validation progress bar in the console.""" + + @distributed.rank0_only + def on_train_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + # self.train_pbar = tqdm.trange(self.config.trainer.max_iter, initial=iteration, desc="Training") + self.start_iteration_time = time.time() + self.elapsed_iteration_time = 0 + + @distributed.rank0_only + def on_training_step_start(self, model: ImaginaireModel, data: dict[str, torch.Tensor], iteration: int = 0) -> None: + self.start_iteration_time = time.time() + + @distributed.rank0_only + def on_training_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + # but this is only called when the optimizer is updated, so it's only the time for the last batch. + self.elapsed_iteration_time += time.time() - self.start_iteration_time + + if iteration % self.config.trainer.logging_iter == 0: + avg_time = self.elapsed_iteration_time / self.config.trainer.logging_iter + log.info(f"Iteration: {iteration}, average iter time: {avg_time:2f}, total loss {loss.item():4f}") + + self.elapsed_iteration_time = 0 + + +class WandBCallback(Callback): + """The callback class for logging to Weights and Biases (W&B). + + By default, WandBCallback logs the following training stats to W&B every config.trainer.logging_iter: + - iteration: The current iteration number (useful for visualizing the training progress over time). + - train/loss: The computed overall loss in the training batch. + - optim/lr: The current learning rate. + - timer/*: The averaged timing results of each code block recorded by trainer.training_timer. + For validation, WandBCallback logs: + - val/loss: The computed overall loss in the validation dataset. + """ + + def on_train_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + wandb_util.init_wandb(self.config, model=model) + + def on_before_optimizer_step( + self, + model_ddp: distributed.DistributedDataParallel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int = 0, + ) -> None: # Log the curent learning rate. + if iteration % self.config.trainer.logging_iter == 0 and distributed.is_rank0(): + wandb.log({"optim/lr": scheduler.get_last_lr()[0]}, step=iteration) + wandb.log({"optim/grad_scale": grad_scaler.get_scale()}, step=iteration) + + def on_training_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: # Log the timing results (over a number of iterations) and the training loss. + if iteration % self.config.trainer.logging_iter == 0: + timer_results = self.trainer.training_timer.compute_average_results() + if distributed.is_rank0(): + wandb.log({f"timer/{key}": value for key, value in timer_results.items()}, step=iteration) + wandb.log({"train/loss": loss}, step=iteration) + wandb.log({"iteration": iteration}, step=iteration) + self.trainer.training_timer.reset() + + def on_validation_start( + self, model: ImaginaireModel, dataloader_val: torch.utils.data.DataLoader, iteration: int = 0 + ) -> None: + # Cache for collecting data/output batches. + self._val_cache: dict[str, Any] = dict( + data_batches=[], + output_batches=[], + loss=torch.tensor(0.0, device="cuda"), + sample_size=torch.tensor(0, device="cuda"), + ) + + def on_validation_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: # Collect the validation batch and aggregate the overall loss. + # Collect the validation batch and aggregate the overall loss. + batch_size = misc.get_data_batch_size(data_batch) + self._val_cache["loss"] += loss * batch_size + self._val_cache["sample_size"] += batch_size + + def on_validation_end(self, model: ImaginaireModel, iteration: int = 0) -> None: + # Compute the average validation loss across all devices. + dist.all_reduce(self._val_cache["loss"], op=dist.ReduceOp.SUM) + dist.all_reduce(self._val_cache["sample_size"], op=dist.ReduceOp.SUM) + loss = self._val_cache["loss"].item() / self._val_cache["sample_size"] + # Log data/stats of validation set to W&B. + if distributed.is_rank0(): + log.info(f"Validation loss (iteration {iteration}): {loss:4f}") + wandb.log({"val/loss": loss}, step=iteration) + + def on_train_end(self, model: ImaginaireModel, iteration: int = 0) -> None: + wandb.finish() + + +class LowPrecisionCallback(Callback): + """The callback class handling low precision training""" + + def __init__(self, config: Config, trainer: ImaginaireTrainer, update_iter: int): + self.update_iter = update_iter + + def on_train_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + if model.precision == torch.float32: + log.critical("Using fp32. We should disable master weights update.") + self.update_iter = sys.maxsize # noqa: F821 + else: + assert model.precision in [ + torch.bfloat16, + torch.float16, + torch.half, + ], "LowPrecisionCallback must use a low precision dtype." + self.precision_type = model.precision + + def on_training_step_start(self, model: ImaginaireModel, data: dict[str, torch.Tensor], iteration: int = 0) -> None: + for k, v in data.items(): + if isinstance(v, torch.Tensor) and torch.is_floating_point(data[k]): + data[k] = v.to(dtype=self.precision_type) + + def on_validation_step_start( + self, model: ImaginaireModel, data: dict[str, torch.Tensor], iteration: int = 0 + ) -> None: + for k, v in data.items(): + if isinstance(v, torch.Tensor) and torch.is_floating_point(data[k]): + data[k] = v.to(dtype=self.precision_type) + + def on_before_zero_grad( + self, + model_ddp: distributed.DistributedDataParallel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + iteration: int = 0, + ) -> None: + if iteration % self.update_iter == 0: + if getattr(optimizer, "master_weights", False): + params, master_params = [], [] + for group, group_master in zip(optimizer.param_groups, optimizer.param_groups_master): + for p, p_master in zip(group["params"], group_master["params"]): + params.append(get_local_tensor_if_DTensor(p.data)) + master_params.append(p_master.data) + torch._foreach_copy_(params, master_params) + + +class NVTXCallback(Callback): + """The callback for creating NVTX ranges""" + + def __init__( + self, + synchronize: bool = False, + config: Optional["Config"] = None, + trainer: Optional["ImaginaireTrainer"] = None, + ): + super().__init__(config, trainer) + self.synchronize = synchronize + + def on_before_forward(self, iteration: int = 0) -> None: + if self.synchronize: + torch.cuda.synchronize() + torch.cuda.nvtx.range_push("forward") + + def on_after_forward(self, iteration: int = 0) -> None: + if self.synchronize: + torch.cuda.synchronize() + torch.cuda.nvtx.range_pop() + + def on_before_backward( + self, model_ddp: distributed.DistributedDataParallel, loss: torch.Tensor, iteration: int = 0 + ) -> None: + if self.synchronize: + torch.cuda.synchronize() + torch.cuda.nvtx.range_push("backward") + + def on_after_backward(self, model_ddp: distributed.DistributedDataParallel, iteration: int = 0) -> None: + if self.synchronize: + torch.cuda.synchronize() + torch.cuda.nvtx.range_pop() + + def on_before_optimizer_step( + self, + model_ddp: distributed.DistributedDataParallel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int = 0, + ) -> None: + if self.synchronize: + torch.cuda.synchronize() + torch.cuda.nvtx.range_push("optimizer_step") + + def on_before_zero_grad( + self, + model_ddp: distributed.DistributedDataParallel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + iteration: int = 0, + ) -> None: + if self.synchronize: + torch.cuda.synchronize() + torch.cuda.nvtx.range_pop() + + def on_before_dataloading(self, iteration: int = 0) -> None: + torch.cuda.nvtx.range_push("dataloading") + + def on_after_dataloading(self, iteration: int = 0) -> None: + torch.cuda.nvtx.range_pop() diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/checkpoint_db.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/checkpoint_db.py new file mode 100644 index 0000000000000000000000000000000000000000..d7d0785328447ccec6f894a9a27ef600ac54de3f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/checkpoint_db.py @@ -0,0 +1,955 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""Database of released checkpoints.""" + +import functools +import os +from functools import cached_property +from typing import Annotated + +import pydantic +from huggingface_hub import hf_hub_download, snapshot_download +from typing_extensions import override + +from cosmos_policy._src.imaginaire.flags import EXPERIMENTAL_CHECKPOINTS, INTERNAL +from cosmos_policy._src.imaginaire.utils import log + + +class _CheckpointUri(pydantic.BaseModel): + """Config for checkpoint file/directory.""" + + model_config = pydantic.ConfigDict(extra="forbid", frozen=True) + + metadata: dict = pydantic.Field(default_factory=dict) + """File metadata. + + Only used for debugging. + """ + + def _download(self) -> str: + raise NotImplementedError("Download method not implemented.") + + @cached_property + def path(self) -> str: + """Return S3 URI or local path.""" + return self._download() + + +def is_s3_uri(uri: str) -> str: + if not uri.startswith("s3://"): + raise ValueError(f"Invalid S3 URI: {uri}. Must start with 's3://'") + return uri.rstrip("/") + + +S3Uri = Annotated[str, pydantic.AfterValidator(is_s3_uri)] + + +class _CheckpointS3(_CheckpointUri): + """Config for checkpoint on S3.""" + + uri: S3Uri + """S3 URI.""" + + +class CheckpointFileS3(_CheckpointS3): + """Config for checkpoint file on S3.""" + + +class CheckpointDirS3(_CheckpointS3): + """Config for checkpoint directory on S3.""" + + +class _CheckpointHf(_CheckpointUri): + """Config for checkpoint on Hugging Face.""" + + repository: str + """Repository id (organization/repository).""" + revision: str + """Git revision id which can be a branch name, a tag, or a commit hash.""" + + +class CheckpointFileHf(_CheckpointHf): + """Config for checkpoint file on Hugging Face.""" + + filename: str + """File name.""" + + @override + def _download(self) -> str: + """Download checkpoint and return the local path.""" + download_kwargs = dict( + repo_id=self.repository, repo_type="model", revision=self.revision, filename=self.filename + ) + log.info(f"Downloading checkpoint file from Hugging Face with {download_kwargs}") + path = hf_hub_download(**download_kwargs) + assert os.path.exists(path), path + return path + + +class CheckpointDirHf(_CheckpointHf): + """Config for checkpoint directory on Hugging Face.""" + + subdirectory: str = "" + """Repository subdirectory.""" + include: tuple[str, ...] = () + """Include patterns. + + See https://huggingface.co/docs/huggingface_hub/en/guides/download#filter-files-to-download + """ + exclude: tuple[str, ...] = () + """Exclude patterns. + + See https://huggingface.co/docs/huggingface_hub/en/guides/download#filter-files-to-download + """ + + @override + def _download(self) -> str: + """Download checkpoint and return the local path.""" + patterns: dict[str, list[str]] = {} + if self.include: + patterns["allow_patterns"] = list(self.include) + else: + patterns["allow_patterns"] = ["*"] + if self.exclude: + patterns["ignore_patterns"] = list(self.exclude) + if self.subdirectory: + patterns = {key: [os.path.join(self.subdirectory, x) for x in val] for key, val in patterns.items()} + download_kwargs = dict(repo_id=self.repository, repo_type="model", revision=self.revision) | patterns + log.info(f"Downloading checkpoint from Hugging Face with {download_kwargs}") + path = snapshot_download(**download_kwargs) + if self.subdirectory: + path = os.path.join(path, self.subdirectory) + assert os.path.exists(path), path + return path + + +class CheckpointConfig(pydantic.BaseModel): + """Config for checkpoint.""" + + model_config = pydantic.ConfigDict(extra="forbid", frozen=True) + + uuid: str + """Checkpoint UUID.""" + name: str + """Checkpoint name. + + Only used for debugging. + """ + metadata: dict = pydantic.Field(default_factory=dict) + """Checkpoint metadata. + + Only used for debugging. + """ + experiment: str | None = None + """Experiment name.""" + + s3: CheckpointFileS3 | CheckpointDirS3 | None = None + """Config for checkpoint on S3.""" + hf: CheckpointFileHf | CheckpointDirHf | None = None + """Config for checkpoint on Hugging Face.""" + + @cached_property + def path(self) -> str: + """Return S3 URI or local path.""" + if INTERNAL and self.s3 is not None: + return self.s3.uri + if self.hf is None: + raise ValueError(f"Checkpoint {self.name}({self.uuid}) is not available on Hugging Face.") + log.info(f"Downloading checkpoint {self.name}({self.uuid})") + return self.hf.path + + +_CHECKPOINTS_BY_UUID: dict[str, CheckpointConfig] = {} +_CHECKPOINTS_BY_S3: dict[str, CheckpointConfig] = {} + + +def _register_checkpoint(checkpoint_config: CheckpointConfig): + if checkpoint_config.uuid in _CHECKPOINTS_BY_UUID: + raise ValueError(f"Checkpoint UUID {checkpoint_config.uuid} already registered.") + _CHECKPOINTS_BY_UUID[checkpoint_config.uuid] = checkpoint_config + if checkpoint_config.s3 is not None: + uri = checkpoint_config.s3.uri + if uri in _CHECKPOINTS_BY_S3: + raise ValueError(f"Checkpoint S3 {uri} already registered.") + _CHECKPOINTS_BY_S3[uri] = checkpoint_config + + +_register_checkpoint( + CheckpointConfig( + uuid="4dbf13c6-1d30-4b02-99d6-75780dd8b744", + name="google-t5/t5-11b", + hf=CheckpointDirHf( + repository="google-t5/t5-11b", + revision="90f37703b3334dfe9d2b009bfcbfbf1ac9d28ea3", + exclude=("tf_model.h5",), + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="a2944743-cf8d-427e-a6fc-b3c03d807064", + name="meta-llama/Llama-Guard-3-8B", + hf=CheckpointDirHf( + repository="meta-llama/Llama-Guard-3-8B", + revision="7327bd9f6efbbe6101dc6cc4736302b3cbb6e425", + exclude=("original/*",), + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="9c7b7da4-2d95-45bb-9cb8-2eed954e9736", + name="nvidia/Cosmos-Guardrail1", + hf=CheckpointDirHf( + repository="nvidia/Cosmos-Guardrail1", + revision="d6d4bfa899a71454a700907664f3e88f503950cf", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="7219c6c7-f878-4137-bbdb-76842ea85e70", + name="Qwen/Qwen2.5-VL-7B-Instruct", + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_reasoning1/pretrained/Qwen_tokenizer/Qwen/Qwen2.5-VL-7B-Instruct", + ), + hf=CheckpointDirHf( + repository="nvidia/Cosmos-Experimental", + revision="736a20b6cfbc38e42ba3f7e7d8efa1d886c20db1", + subdirectory="7219c6c7-f878-4137-bbdb-76842ea85e70", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointDirHf( + repository="nvidia/Cosmos-Reason1-7B", + revision="3210bec0495fdc7a8d3dbb8d58da5711eab4b423", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="685afcaa-4de2-42fe-b7b9-69f7a2dee4d8", + name="Wan2.1/vae", + s3=CheckpointFileS3( + uri="s3://bucket/cosmos_diffusion_v2/pretrain_weights/tokenizer/wan2pt1/Wan2.1_VAE.pth", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="736a20b6cfbc38e42ba3f7e7d8efa1d886c20db1", + filename="685afcaa-4de2-42fe-b7b9-69f7a2dee4d8.pth", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Predict2.5-2B", + revision="6787e176dce74a101d922174a95dba29fa5f0c55", + filename="tokenizer.pth", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="cb3e3ffa-7b08-4c34-822d-61c7aa31a14f", + name="nvidia/Cosmos-Reason1.1-7B", + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="736a20b6cfbc38e42ba3f7e7d8efa1d886c20db1", + filename="cb3e3ffa-7b08-4c34-822d-61c7aa31a14f/model.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointDirHf( + repository="nvidia/Cosmos-Reason1-7B", + revision="3210bec0495fdc7a8d3dbb8d58da5711eab4b423", + ), + ), +) + +# ----------------------------------------------------------------------------- +# Cosmos-Predict2.5-2B +# ----------------------------------------------------------------------------- +_register_checkpoint( + CheckpointConfig( + uuid="d20b7120-df3e-4911-919d-db6e08bad31c", + name="nvidia/Cosmos-Predict2.5-2B/base/pre-trained", + experiment="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_resume2", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_resume2/checkpoints/iter_000023000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="d20b7120-df3e-4911-919d-db6e08bad31c/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Predict2.5-2B", + revision="15a82a2ec231bc318692aa0456a36537c806e7d4", + filename="base/pre-trained/d20b7120-df3e-4911-919d-db6e08bad31c_ema_bf16.pt", + ), + ), +) + +checkpoint_hf = CheckpointDirHf( + repository="nvidia/Cosmos-Experimental", + revision="eda2f0ca1db6281c9a960908bb6bf14607a0fea0", + subdirectory="308eb96c-c4c0-4a06-9cc1-103a43beff28", +) + +_register_checkpoint( + CheckpointConfig( + uuid="308eb96c-c4c0-4a06-9cc1-103a43beff28", + name="nvidia/Cosmos-Predict2.5-2B/base/pre-trained", + experiment="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + ), + hf=checkpoint_hf, + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="7bbc8d06-2bc9-448d-94ee-b48b4ab7189c", + name="nvidia/Cosmos-Predict2.5-2B/interactive", + experiment="cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_sf_warmup", + s3=CheckpointFileS3( + uri="s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes/checkpoints/iter_000014000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="2b5e9a99b58d5a61259ca99962c4c74127481006", + filename="7bbc8d06-2bc9-448d-94ee-b48b4ab7189c/model_ema_bf16.pt", + ), + ), +) +_register_checkpoint( + CheckpointConfig( + uuid="bedc35da-1a54-4144-83db-6072c29b0fd9", + name="nvidia/Cosmos-Predict2.5-2B/interactive", + experiment="cosmos_predict2p5_2B_action_gr00t_gr1_warmup", + s3=CheckpointFileS3( + uri="s3://bucket/cosmos_predict2_action_conditioned/interactive_warmup/gr1/checkpoints/iter_000020000/model" + ), + hf=CheckpointDirHf( + repository="nvidia/Cosmos-Experimental", + revision="ded876a5b2e19aef64cd9d1100c03e5b05cf2f9c", + subdirectory="bedc35da-1a54-4144-83db-6072c29b0fd9", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="81edfebe-bd6a-4039-8c1d-737df1a790bf", + name="nvidia/Cosmos-Predict2.5-2B/base/post-trained", + experiment="Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointFileS3( + uri="s3://bucket/cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_GRPO-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-posttrain_data-HQ_V7_RF_MERGE_LOCAL_ag_every2_guidance0_scorekeyoverall_reward_databeta0.01_mincon0/checkpoints/iter_000000288/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="81edfebe-bd6a-4039-8c1d-737df1a790bf/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Predict2.5-2B", + revision="15a82a2ec231bc318692aa0456a36537c806e7d4", + filename="base/post-trained/81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="524af350-2e43-496c-8590-3646ae1325da", + name="nvidia/Cosmos-Predict2.5-2B/auto/multiview", + experiment="buttercup_predict2p5_2b_7views_res720p_fps30_t8_joint_alpamayo1capviewprefix_allcapsviewprefix_29frames_nofps_uniform_dropoutt0", + metadata={ + "resolution": "720p", + "fps": 30, + "views": 7, + "frames": 29, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_predict2_multiview/cosmos2_mv/buttercup_predict2p5_2b_7views_res720p_fps30_t8_joint_alpamayo1capviewprefix_allcapsviewprefix_29frames_nofps_uniform_dropoutt0-0/checkpoints/iter_000012000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Predict2.5-2B", + revision="865baf084d4c9e850eac59a021277d5a9b9e8b63", + filename="auto/multiview/524af350-2e43-496c-8590-3646ae1325da_ema_bf16.pt", + ), + ) +) + +_register_checkpoint( + CheckpointConfig( + uuid="6b9d7548-33bb-4517-b5e8-60caf47edba7", + name="nvidia/Cosmos-Predict2.5-2B/auto/multiview", + experiment="buttercup_predict2p5_2b_7views_res720p_fps30_t8_from48kfps30mv_condprobs0442_joint_alpamayo1capnoviewprefix_allcapsviewprefix_29frames_nofps", + metadata={ + "resolution": "720p", + "fps": 30, + "views": 7, + "frames": 29, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_predict2_multiview/cosmos2_mv/buttercup_predict2p5_2b_7views_res720p_fps30_t8_from48kfps30mv_condprobs0442_joint_alpamayo1capnoviewprefix_allcapsviewprefix_29frames_nofps-0/checkpoints/iter_000005000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="6b9d7548-33bb-4517-b5e8-60caf47edba7/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Predict2.5-2B", + revision="15a82a2ec231bc318692aa0456a36537c806e7d4", + filename="auto/multiview/6b9d7548-33bb-4517-b5e8-60caf47edba7_ema_bf16.pt", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="f740321e-2cd6-4370-bbfe-545f4eca2065", + name="nvidia/Cosmos-Predict2.5-2B/robot/multiview-agibot", + experiment="multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_agibot_frameinit", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_diffusion_v2/official_runs_vid2vid/multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_agibot_frameinit/checkpoints/iter_000016500", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Predict2.5-2B", + revision="fbe72c18d152053029a19db3b211cf78671ad422", + filename="f740321e-2cd6-4370-bbfe-545f4eca2065/model_ema_bf16.pt", + ), + ), +) + + +_register_checkpoint( + CheckpointConfig( + uuid="0e8177cc-0db5-4cfd-a8a4-b820c772f4fc", + name="nvidia/Cosmos-Transfer2.5-2B/general/multiview-camera", + experiment="multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_multicam_syncam", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_diffusion_v2/official_runs_vid2vid/multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_multicam_syncam/checkpoints/iter_000002000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="0e8177cc-0db5-4cfd-a8a4-b820c772f4fc/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else None, + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="7f6b99b7-7fac-4e74-8dbe-a394cb56ef99", + name="nvidia/Cosmos-Transfer2.5-2B/robot/multiview-agibot-camera", + experiment="multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_agibot", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_diffusion_v2/official_runs_vid2vid/multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_agibot/checkpoints/iter_000003000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="7f6b99b7-7fac-4e74-8dbe-a394cb56ef99/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else None, + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="38c6c645-7d41-4560-8eeb-6f4ddc0e6574", + name="nvidia/Cosmos-Predict2.5-2B/robot/action-cond", + experiment="cosmos_predict2p5_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_256x320", + metadata={ + "resolution": "360p", + "fps": 4, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_256x320/checkpoints/iter_000016000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="main", + filename="38c6c645-7d41-4560-8eeb-6f4ddc0e6574/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Predict2.5-2B", + revision="main", + filename="robot/action-cond/38c6c645-7d41-4560-8eeb-6f4ddc0e6574_ema_bf16.pt", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="24a3b7b8-6a3d-432d-b7d1-5d30b9229465", + name="nvidia/Cosmos-Predict2.5-2B/transfer2.5", + experiment="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only/checkpoints/iter_000037000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="24a3b7b8-6a3d-432d-b7d1-5d30b9229465/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else None, + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="575edf0f-d973-4c74-b52c-69929a08d0a5", + name="nvidia/Cosmos-Predict2.5-2B/base/distilled", + experiment="dmd2_trigflow_distill_cosmos_predict2_2B_bidirectional_TnI2V", + metadata={ + "size": "2B", + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointFileS3( + uri="s3://bucket/cosmos_predict2_distill/predict2_distill/dmd2_trigflow_distill_cosmos_predict2_2B_bidirectional/checkpoints/iter_000007500/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="cb56c64d7e5bb20a50b1e39b4429b967522e91d4", + filename="575edf0f-d973-4c74-b52c-69929a08d0a5/model_ema_bf16.pt", + ), + ), +) + +# ----------------------------------------------------------------------------- +# Cosmos-Predict2.5-14B +# ----------------------------------------------------------------------------- +_register_checkpoint( + CheckpointConfig( + uuid="54937b8c-29de-4f04-862c-e67b04ec41e8", + name="nvidia/Cosmos-Predict2.5-14B/base/pre-trained", + experiment="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma", + metadata={ + "size": "14B", + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointFileS3( + uri="s3://bucket/cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma/checkpoints/iter_000012500/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="54937b8c-29de-4f04-862c-e67b04ec41e8/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Predict2.5-14B", + revision="03eb354f35eae0d6e0c1be3c9f94d8551e125570", + filename="base/pre-trained/54937b8c-29de-4f04-862c-e67b04ec41e8_ema_bf16.pt", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="e21d2a49-4747-44c8-ba44-9f6f9243715f", + name="nvidia/Cosmos-Predict2.5-14B/base/post-trained", + experiment="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma", + metadata={ + "size": "14B", + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointFileS3( + uri="s3://bucket/cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_GRPO-reason_embeddings-Index-26-Size-14B-Res-720-Fps-16-posttrain_data-HQ_V7_RF_MERGE_GENERAL_steps20_every2_lr3e-6_guidance0_scorekeyoverall_reward_databeta0.01_mincon0/checkpoints/iter_000000128/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9e46ea0945ac591f9d4abe810ecd78cde112fb82", + filename="e21d2a49-4747-44c8-ba44-9f6f9243715f/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Predict2.5-14B", + revision="2bc4ca5ba5a20b9858a7ddb856bc82d70b030fbe", + filename="base/post-trained/e21d2a49-4747-44c8-ba44-9f6f9243715f_ema_bf16.pt", + ), + ), +) + +# ----------------------------------------------------------------------------- +# Cosmos-Transfer2.5-2B +# ----------------------------------------------------------------------------- +_register_checkpoint( + CheckpointConfig( + uuid="61f5694b-0ad5-4ecd-8ad7-c8545627d125", + name="nvidia/Cosmos-Transfer2.5-2B/general/edge", + experiment="edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="61c640b2f4092cb0868c9d9941fa505a750ccd4d", + filename="61f5694b-0ad5-4ecd-8ad7-c8545627d125/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Transfer2.5-2B", + revision="b67b64abda3801a9aceddbff2bdb86126c06db74", + filename="general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt", + ), + ) +) + +_register_checkpoint( + CheckpointConfig( + uuid="626e6618-bfcd-4d9a-a077-1409e2ce353f", + name="nvidia/Cosmos-Transfer2.5-2B/general/depth", + experiment="depth_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv4p1_20250823_64N_rectified_flow_refimdrop0pt5", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_transfer2/vid2vid_2B_control/depth_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv4p1_20250823_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000044000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="61c640b2f4092cb0868c9d9941fa505a750ccd4d", + filename="626e6618-bfcd-4d9a-a077-1409e2ce353f/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Transfer2.5-2B", + revision="dea7737ca29dd8d9086413c6dc5724b8250a0bb4", + filename="general/depth/626e6618-bfcd-4d9a-a077-1409e2ce353f_ema_bf16.pt", + ), + ) +) + +_register_checkpoint( + CheckpointConfig( + uuid="ba2f44f2-c726-4fe7-949f-597069d9b91c", + name="nvidia/Cosmos-Transfer2.5-2B/general/blur", + experiment="vis_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5_filterb3g5m2", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_transfer2/vid2vid_2B_control/vis_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5_filterb3g5m2/checkpoints/iter_000036000/", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="8ecf7ad717c10b9f796bab731eba311e480fcf58", + filename="ba2f44f2-c726-4fe7-949f-597069d9b91c/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Transfer2.5-2B", + revision="eb5325b77d358944da58a690157dd2b8071bbf85", + filename="general/blur/ba2f44f2-c726-4fe7-949f-597069d9b91c_ema_bf16.pt", + ), + ) +) + +_register_checkpoint( + CheckpointConfig( + uuid="5136ef49-6d8d-42e8-8abf-7dac722a304a", + name="nvidia/Cosmos-Transfer2.5-2B/general/seg", + experiment="seg_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv4p2_20250823_64N_rectified_flow_refimdrop0pt5", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_transfer2/vid2vid_2B_control/seg_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv4p2_20250823_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000043000/", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="8ecf7ad717c10b9f796bab731eba311e480fcf58", + filename="5136ef49-6d8d-42e8-8abf-7dac722a304a/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Transfer2.5-2B", + revision="23057a4167b89de89a4a397fdbf3887994d115eb", + filename="general/seg/5136ef49-6d8d-42e8-8abf-7dac722a304a_ema_bf16.pt", + ), + ) +) + +_register_checkpoint( + CheckpointConfig( + uuid="ecd0ba00-d598-4f94-aa09-e8627899c431", + name="nvidia/Cosmos-Transfer2.5-2B/general/edge", + experiment="edge_720p_t24_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_nonuniform_hqv3p1_20250714_64N_rectified_flow_mock_data", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_nonuniform_hqv3p1_20250714_64N_rectified_flow/checkpoints/iter_000029000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="ecd0ba00-d598-4f94-aa09-e8627899c431/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Transfer2.5-2B", + revision="bd963eabcfc2d61dc4ea365cacf41d45ac480aa5", + filename="general/edge/ecd0ba00-d598-4f94-aa09-e8627899c431_ema_bf16.pt", + ), + ), +) + + +_register_checkpoint( + CheckpointConfig( + uuid="fcab44fe-6fe7-492e-b9c6-67ef8c1a52ab", + name="nvidia/Cosmos-Transfer2.5-2B/general/seg", + experiment="seg_720p_t24_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_nonuniform_hqv4p2_20250823_64N_rectified_flow", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_transfer2/vid2vid_2B_control/seg_720p_t24_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_nonuniform_hqv4p2_20250823_64N_rectified_flow/checkpoints/iter_000031000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="fcab44fe-6fe7-492e-b9c6-67ef8c1a52ab/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Transfer2.5-2B", + revision="bd963eabcfc2d61dc4ea365cacf41d45ac480aa5", + filename="general/seg/fcab44fe-6fe7-492e-b9c6-67ef8c1a52ab_ema_bf16.pt", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="20d9fd0b-af4c-4cca-ad0b-f9b45f0805f1", + name="nvidia/Cosmos-Transfer2.5-2B/general/blur", + experiment="vis_720p_t24_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_nonuniform_hqv3p1_20250714_64N_rectified_flow", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_transfer2/vid2vid_2B_control/vis_720p_t24_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_nonuniform_hqv3p1_20250714_64N_rectified_flow/checkpoints/iter_000043000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="20d9fd0b-af4c-4cca-ad0b-f9b45f0805f1/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Transfer2.5-2B", + revision="bd963eabcfc2d61dc4ea365cacf41d45ac480aa5", + filename="general/blur/20d9fd0b-af4c-4cca-ad0b-f9b45f0805f1_ema_bf16.pt", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="0f214f66-ae98-43cf-ab25-d65d09a7e68f", + name="nvidia/Cosmos-Transfer2.5-2B/general/depth", + experiment="depth_720p_t24_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_nonuniform_hqv4p1_20250823_64N_rectified_flow", + metadata={ + "resolution": "720p", + "fps": 16, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_transfer2/vid2vid_2B_control/depth_720p_t24_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_nonuniform_hqv4p1_20250823_64N_rectified_flow/checkpoints/iter_000028000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="0f214f66-ae98-43cf-ab25-d65d09a7e68f/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Transfer2.5-2B", + revision="bd963eabcfc2d61dc4ea365cacf41d45ac480aa5", + filename="general/depth/0f214f66-ae98-43cf-ab25-d65d09a7e68f_ema_bf16.pt", + ), + ), +) + +_register_checkpoint( + CheckpointConfig( + uuid="4ecc66e9-df19-4aed-9802-0d11e057287a", + name="nvidia/Cosmos-Transfer2.5-2B/auto/multiview", + experiment="buttercup_transfer2p5_2b_mv_7views_res720p_fps10_t8_fromfinetuned12knofpsuniform_mads720pmulticaps29frames_world_scenario_nofps_uniform", + metadata={ + "resolution": "720p", + "fps": 10, + "views": 7, + "frames": 29, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_transfer2_multiview/cosmos2_mv/buttercup_transfer2p5_2b_mv_7views_res720p_fps10_t8_fromfinetuned12knofpsuniform_mads720pmulticaps29frames_world_scenario_nofps_uniform-0/checkpoints/iter_000006500/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Transfer2.5-2B", + revision="00c591edab119e8a6ca06e6e091351a04ce0ecc9", + filename="auto/multiview/4ecc66e9-df19-4aed-9802-0d11e057287a_ema_bf16.pt", + ), + ) +) + +_register_checkpoint( + CheckpointConfig( + uuid="b5ab002d-a120-4fbf-a7f9-04af8615710b", + name="nvidia/Cosmos-Transfer2.5-2B/auto/multiview", + experiment="buttercup_transfer2p5_2b_mv_7views_res720p_fps10_t8_frombase5knofps_mads720pmulticaps29frames_world_scenario_resumefrom21k", + metadata={ + "resolution": "720p", + "fps": 16, + "views": 7, + "frames": 29, + }, + s3=CheckpointDirS3( + uri="s3://bucket/cosmos_transfer2_multiview/cosmos2_mv/buttercup_transfer2p5_2b_mv_7views_res720p_fps10_t8_frombase5knofps_mads720pmulticaps29frames_world_scenario_resumefrom21k-0/checkpoints/iter_000010000/model", + ), + hf=CheckpointFileHf( + repository="nvidia/Cosmos-Experimental", + revision="9a02ed8daa8c6c7718ac09da06488bfd1d363cb6", + filename="b5ab002d-a120-4fbf-a7f9-04af8615710b/model_ema_bf16.pt", + ) + if EXPERIMENTAL_CHECKPOINTS + else CheckpointFileHf( + repository="nvidia/Cosmos-Transfer2.5-2B", + revision="bd963eabcfc2d61dc4ea365cacf41d45ac480aa5", + filename="auto/multiview/b5ab002d-a120-4fbf-a7f9-04af8615710b_ema_bf16.pt", + ), + ), +) + + +def get_checkpoint_by_uuid(checkpoint_uuid: str) -> CheckpointConfig: + """Return checkpoint config for UUID.""" + if checkpoint_uuid not in _CHECKPOINTS_BY_UUID: + raise ValueError(f"Checkpoint UUID {checkpoint_uuid} not found.") + return _CHECKPOINTS_BY_UUID[checkpoint_uuid] + + +def get_checkpoint_by_s3(checkpoint_s3: str) -> CheckpointConfig: + """Return checkpoint config for S3 URI.""" + checkpoint_s3 = checkpoint_s3.rstrip("/") + if checkpoint_s3 not in _CHECKPOINTS_BY_S3: + raise ValueError(f"Checkpoint S3 {checkpoint_s3} not found.") + return _CHECKPOINTS_BY_S3[checkpoint_s3] + + +@functools.lru_cache +def get_checkpoint_by_hf(checkpoint_hf: str) -> str: + """Download checkpoint from HuggingFace and return local path.""" + # Parse hf://org/repo/path/to/file.pth + assert checkpoint_hf.startswith("hf://"), f"Not a HuggingFace URI: {checkpoint_hf}" + hf_path = checkpoint_hf[5:] # Remove "hf://" prefix + # Split into repo_id (org/repo) and filename (path/to/file.pth) + parts = hf_path.split("/") + if len(parts) < 3: + raise ValueError( + f"Invalid HuggingFace URI format: {checkpoint_hf}. Expected format: hf://org/repo/path/to/file.pth" + ) + repo_id = "/".join(parts[:2]) # org/repo + filename = "/".join(parts[2:]) # path/to/file.pth + log.info(f"Downloading checkpoint from HuggingFace: {repo_id}/{filename}") + path = hf_hub_download( + repo_id=repo_id, + repo_type="model", + filename=filename, + ) + assert os.path.exists(path), path + return path + + +@functools.lru_cache +def get_checkpoint_path(checkpoint_uri: str) -> str: + """Return checkpoint path for S3 URI, HuggingFace URI, or local path. + + Supports: + - S3 URIs: s3://bucket/path/to/checkpoint + - HuggingFace URIs: hf://org/repo/path/to/file.pth + - Local paths: /path/to/checkpoint + """ + if INTERNAL: + return checkpoint_uri + checkpoint_uri = checkpoint_uri.rstrip("/") + if checkpoint_uri.startswith("s3://"): + return get_checkpoint_by_s3(checkpoint_uri).path + if checkpoint_uri.startswith("hf://"): + return get_checkpoint_by_hf(checkpoint_uri) + if not os.path.exists(checkpoint_uri): + raise ValueError(f"Checkpoint path {checkpoint_uri} does not exist.") + return checkpoint_uri diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/checkpoint_db_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/checkpoint_db_test.py new file mode 100644 index 0000000000000000000000000000000000000000..6432d8c963457cf27ad6fc8a08df0bf7636580d9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/checkpoint_db_test.py @@ -0,0 +1,76 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from pathlib import Path + +import pytest + +from cosmos_policy._src.imaginaire.utils.checkpoint_db import ( + _CHECKPOINTS_BY_UUID, + get_checkpoint_by_s3, + get_checkpoint_by_uuid, + get_checkpoint_path, +) + + +@pytest.mark.L0 +def test_get_checkpoint_file(): + uuid = "685afcaa-4de2-42fe-b7b9-69f7a2dee4d8" + s3_uri = "s3://bucket/cosmos_diffusion_v2/pretrain_weights/tokenizer/wan2pt1/Wan2.1_VAE.pth" + config = get_checkpoint_by_uuid(uuid) + assert config.s3 is not None + assert config.hf is not None + assert get_checkpoint_by_s3(s3_uri) is config + assert get_checkpoint_path(s3_uri) == config.path + assert get_checkpoint_path(config.path) == config.path + + +@pytest.mark.L1 +def test_get_checkpoint_hf_file(): + uuid = "685afcaa-4de2-42fe-b7b9-69f7a2dee4d8" + config = get_checkpoint_by_uuid(uuid) + hf_path = Path(config.hf.path) + assert hf_path.is_file() + assert hf_path.suffix == ".pth" + + +@pytest.mark.L0 +def test_get_checkpoint_dir(): + uuid = "7219c6c7-f878-4137-bbdb-76842ea85e70" + s3_uri = "s3://bucket/cosmos_reasoning1/pretrained/Qwen_tokenizer/Qwen/Qwen2.5-VL-7B-Instruct" + config = get_checkpoint_by_uuid(uuid) + assert config.s3 is not None + assert config.hf is not None + assert get_checkpoint_by_s3(s3_uri) is config + assert get_checkpoint_path(s3_uri) == config.path + assert get_checkpoint_path(config.path) == config.path + + +@pytest.mark.L1 +def test_get_checkpoint_hf_dir(): + uuid = "7219c6c7-f878-4137-bbdb-76842ea85e70" + config = get_checkpoint_by_uuid(uuid) + hf_path = Path(config.hf.path) + assert hf_path.is_dir() + assert hf_path.joinpath("tokenizer.json").is_file() + + +@pytest.mark.L1 +def test_all_checkpoints(): + for config in _CHECKPOINTS_BY_UUID.values(): + # Check Hugging Face checkpoint + if config.hf is not None: + hf_path = Path(config.hf.path) + assert hf_path.exists() diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/checkpointer.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/checkpointer.py new file mode 100644 index 0000000000000000000000000000000000000000..63a6ae68cbec88c0dea2036331b691520f15d51a --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/checkpointer.py @@ -0,0 +1,502 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import os +import threading +from typing import TYPE_CHECKING, List, NamedTuple, Tuple + +import torch + +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import callback, distributed, log, misc, object_store + +if TYPE_CHECKING: + from cosmos_policy._src.imaginaire.config import CheckpointConfig, JobConfig + +TORCH_VERSION: Tuple[int, ...] = tuple(int(x) for x in torch.__version__.split(".")[:2]) +if TORCH_VERSION >= (1, 11): + from torch.ao import quantization + from torch.ao.quantization import FakeQuantizeBase, ObserverBase +elif ( + TORCH_VERSION >= (1, 8) + and hasattr(torch.quantization, "FakeQuantizeBase") + and hasattr(torch.quantization, "ObserverBase") +): + from torch import quantization + from torch.quantization import FakeQuantizeBase, ObserverBase + + +class Checkpointer: + """The checkpointer class. Supports checkpoint saving/loading to both local disk or object store.""" + + def __init__(self, config_checkpoint: CheckpointConfig, config_job: JobConfig, callbacks: callback.CallBackGroup): + """Constructor of the checkpointer. + + Args: + config_checkpoint (CheckpointConfig): The config object for the checkpointer. + """ + # Set the callback functions. + self.callbacks = callbacks + + self.checkpoint_dir_local = f"{config_job.path_local}/checkpoints" + self.checkpoint_dir_object_store = f"{config_job.path}/checkpoints" + self.save_to_object_store = config_checkpoint.save_to_object_store.enabled + self.load_from_object_store = config_checkpoint.load_from_object_store.enabled + self.strict_resume = config_checkpoint.strict_resume + self.load_path = config_checkpoint.load_path or None + self.load_training_state = config_checkpoint.load_training_state + self.only_load_scheduler_state = config_checkpoint.only_load_scheduler_state + self.save_thread = None + # Create the object store client interface. + if self.save_to_object_store: + self.object_store_saver = object_store.ObjectStore(config_checkpoint.save_to_object_store) + if self.load_from_object_store: + self.object_store_loader = object_store.ObjectStore(config_checkpoint.load_from_object_store) + + def save( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int, + ) -> None: + """Save network weights, optimizer parameters, scheduler parameters to a checkpoint. + + Args: + model (ImaginaireModel): The PyTorch model. + optimizer (torch.optim.Optimizer): The model optimizer. + scheduler (torch.optim.lr_scheduler.LRScheduler): The optimization scheduler. + grad_scaler (torch.amp.GradScaler): The gradient scaler (for mixed precision training). + iteration (int): Current iteration number. + """ + self.callbacks.on_save_checkpoint_start(model, iteration) + + checkpoint_file = f"iter_{iteration:09}.pt" + + if distributed.get_rank() == 0: + state_dict = dict( + model=model.state_dict(), + optimizer=optimizer.state_dict(), + scheduler=scheduler.state_dict(), + grad_scaler=grad_scaler.state_dict(), + iteration=iteration, + ) + state_dict = misc.to(state_dict, device="cpu") + self.callbacks.on_save_checkpoint(model, state_dict=state_dict) + # Wait for previous saver thread to end. + if self.save_thread: + self.save_thread.join() + # Run the checkpoint saver in a separate thread. + self.save_thread = threading.Thread( + target=self._save_worker_object_store if self.save_to_object_store else self._save_worker_local, + daemon=False, + args=(state_dict, checkpoint_file, distributed.get_rank()), + ) + self.save_thread.start() + + # Note: Checkpoints are saved on a separate thread and this callback is not accurate. + # Please check logs from on_save_checkpoint_success() for better accuracy + self.callbacks.on_save_checkpoint_end(model=None, iteration=iteration) + + @misc.timer("checkpoint saving (local)") + def _save_worker_local(self, state_dict: dict[str, torch.Tensor], checkpoint_file: str, rank: int = 0) -> None: + """Worker to save checkpoint to local disk, spawned with a child thread (runs in parallel with the training). + + Args: + state_dict (dict[str, torch.Tensor]): The state dict of the model/optimizer/scheduler. + checkpoint_file (str): The file name of the model checkpoint. + rank (int): GPU device (default: 0). + """ + checkpoint_path = os.path.join(self.checkpoint_dir_local, checkpoint_file) + os.makedirs(self.checkpoint_dir_local, exist_ok=True) + try: + torch.save(state_dict, checkpoint_path) + if rank == 0: + self._write_latest_checkpoint_file(checkpoint_file) + log.success(f"Saved checkpoint (local): {checkpoint_path}") + iteration = int(checkpoint_file.replace("iter_", "").replace(".pt", "")) + self.callbacks.on_save_checkpoint_success(iteration=iteration) + except Exception as e: # noqa: BLE001 + log.exception(f"Checkpoint failed to save (local): {e}") + + @misc.timer("checkpoint saving (object store)") + def _save_worker_object_store( + self, state_dict: dict[str, torch.Tensor], checkpoint_file: str, rank: int = 0 + ) -> None: + """Worker to upload checkpoint to object store, spawned with a child thread (in parallel with the training). + + Args: + state_dict (dict[str, torch.Tensor]): The state dict of the model/optimizer/scheduler. + checkpoint_file (str): The file name of the model checkpoint. + rank (int): GPU device (default: 0). + """ + checkpoint_path = os.path.join(self.checkpoint_dir_object_store, checkpoint_file) + try: + self.object_store_saver.save_object(state_dict, key=checkpoint_path, type="torch") + if rank == 0: + self._write_latest_checkpoint_file(checkpoint_file) + log.success(f"Saved checkpoint (object store): {checkpoint_path}") + iteration = int(checkpoint_file.replace("iter_", "").replace(".pt", "")) + self.callbacks.on_save_checkpoint_success(iteration=iteration) + except Exception as e: # noqa: BLE001 + log.exception(f"Checkpoint failed to upload (object store): {e}") + + @misc.timer("checkpoint loading") + def load( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer | None = None, + scheduler: torch.optim.lr_scheduler.LRScheduler | None = None, + grad_scaler: torch.amp.GradScaler | None = None, + ) -> int: + """Load network weights and optimizer states from a checkpoint in a single process. + + The priority of the checkpoint loading logic is: + 1. Attempt to resume training if possible by looking for latest_checkpoint.txt under the same name. + 2. If no latest checkpoint were found, it loads the model weights specified by config_checkpoint.path. + - This is typically used for inference mode. + - If config_checkpoint.load_optimizer_state is True, then also load the optimizer and scheduler states. + 3. If none of the above, randomly initialize the model parameters and train from scratch. + + Args: + model (ImaginaireModel): The PyTorch model. + optimizer (torch.optim.Optimizer | None): The model optimizer (default: None). + scheduler (torch.optim.lr_scheduler.LRScheduler | None): The optimization scheduler (default: None). + grad_scaler (torch.amp.GradScaler | None): The gradient scaler (for mixed precision training). + + Returns: + iteration (int): the iteration number to start/resume from. + """ + self.callbacks.on_load_checkpoint_start(model) + + latest_checkpoint_file = self._read_latest_checkpoint_file() + if latest_checkpoint_file is not None: + # 1. Resume training from latest_checkpoint.txt under the same name. + checkpoint_dir = ( + self.checkpoint_dir_object_store if self.load_from_object_store else self.checkpoint_dir_local + ) + checkpoint_path = os.path.join(checkpoint_dir, latest_checkpoint_file) + resume = True + only_resume_scheduler = True + else: + if self.load_path: + # 2. Load the module weights specified by config_checkpoint.path. + checkpoint_path = self.load_path + resume = self.load_training_state + only_resume_scheduler = self.only_load_scheduler_state + else: + # 3. Randomly initialize the model parameters and train from scratch. + checkpoint_path = None + resume = False + only_resume_scheduler = False + # Load checkpoint. + if checkpoint_path is not None: + self._check_checkpoint_exists(checkpoint_path) + if self.load_from_object_store: + log.info(f"Loading checkpoint (object store): {checkpoint_path}") + state_dict = self.object_store_loader.load_object(key=checkpoint_path, type="torch") + log.success(f"Complete loading checkpoint (object store): {checkpoint_path}") + else: + log.info(f"Loading checkpoint (local): {checkpoint_path}") + state_dict = torch.load(checkpoint_path, map_location=lambda storage, loc: storage) + log.success(f"Complete loading checkpoint (local): {checkpoint_path}") + self.callbacks.on_load_checkpoint(model, state_dict=state_dict) + # Load the state dicts. + log.info("- Loading the model...") + model.load_state_dict(state_dict["model"], strict=self.strict_resume) + if resume or only_resume_scheduler: + iteration = state_dict["iteration"] + assert scheduler + log.info("- Loading the scheduler...") + scheduler.load_state_dict(state_dict["scheduler"]) + scheduler.last_epoch = iteration + else: + iteration = 0 + if resume: + assert optimizer + log.info("- Loading the optimizer...") + optimizer.load_state_dict(state_dict["optimizer"]) + log.info("- Loading the gradient scaler...") + grad_scaler.load_state_dict(state_dict["grad_scaler"]) + log.success(f"Done with loading the checkpoint (iteration {iteration}).") + else: + log.success("Done with loading the checkpoint.") + else: + # Checkpoint not found and not specified. We will train everything from scratch. + iteration = 0 + log.info("Training from scratch.") + torch.cuda.empty_cache() + + self.callbacks.on_load_checkpoint_end(model, iteration=iteration, checkpoint_path=checkpoint_path) + + return iteration + + def _read_latest_checkpoint_file(self) -> str | None: + """Get the file name of the latest saved checkpoint. If it doesn't exist, return None. + + Returns: + checkpoint_file (str | None): file name of the latest saved checkpoint. + """ + checkpoint_file = None + if self.load_from_object_store: + latest_path = os.path.join(self.checkpoint_dir_object_store, "latest_checkpoint.txt") + if self.object_store_loader.object_exists(key=latest_path): + checkpoint_file = self.object_store_loader.load_object(key=latest_path, type="text").strip() + else: + latest_path = os.path.join(self.checkpoint_dir_local, "latest_checkpoint.txt") + if os.path.isfile(latest_path): + checkpoint_file = open(latest_path).read().strip() + return checkpoint_file + + def _write_latest_checkpoint_file(self, checkpoint_file: str) -> None: + """Track the file name of the latest saved checkpoint. + + Args: + checkpoint_file (str): file name of the latest saved checkpoint. + """ + content = f"{checkpoint_file}\n" + if self.save_to_object_store: + latest_path = os.path.join(self.checkpoint_dir_object_store, "latest_checkpoint.txt") + self.object_store_saver.save_object(content, key=latest_path, type="text") + else: + latest_path = os.path.join(self.checkpoint_dir_local, "latest_checkpoint.txt") + with open(latest_path, "w") as file: + file.write(content) + + def _check_checkpoint_exists(self, checkpoint_path: str) -> None: + """If the file checkpoint_path does not exist, raise an error. + + Args: + checkpoint_path (str): full path to the checkpoint. + """ + if self.load_from_object_store: + if not self.object_store_loader.object_exists(key=checkpoint_path): + raise FileNotFoundError(f"File not found (object store): {checkpoint_path}") + else: + if not os.path.exists(checkpoint_path): + raise FileNotFoundError(f"File not found (local): {checkpoint_path}") + + def finalize(self) -> None: + """Finalize the checkpointer.""" + if self.save_thread: + self.save_thread.join() + + +class _IncompatibleKeys( + NamedTuple( + "IncompatibleKeys", + [ + ("missing_keys", List[str]), + ("unexpected_keys", List[str]), + ("incorrect_shapes", List[Tuple[str, Tuple[int], Tuple[int]]]), + ], + ) +): + pass + + +class MultiRankCheckpointer(Checkpointer): + def save( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int, + ) -> None: + """Save network weights, optimizer parameters, scheduler parameters to a checkpoint. + + Args: + model (ImaginaireModel): The PyTorch model. + optimizer (torch.optim.Optimizer): The model optimizer. + scheduler (torch.optim.lr_scheduler.LRScheduler): The optimization scheduler. + grad_scaler (torch.amp.GradScaler): The gradient scaler (for mixed precision training). + iteration (int): Current iteration number. + """ + # checkpoint_file = f"iter_{iteration:09}.pt" + postfix, _, total_ema_num = model.get_ckpt_postfix() + checkpoint_file = f"iter_{iteration:09}{postfix}.pt" + save_ranks = list(range(total_ema_num)) + for _rank in save_ranks: + if distributed.get_rank() == _rank: + state_dict = dict( + model=model.state_dict(), + optimizer=optimizer.state_dict(), + scheduler=scheduler.state_dict(), + grad_scaler=grad_scaler.state_dict(), + iteration=iteration, + ) + state_dict = misc.to(state_dict, device="cpu") + self.callbacks.on_save_checkpoint(model, state_dict=state_dict) + # Wait for previous saver thread to end. + if self.save_thread: + self.save_thread.join() + # Run the checkpoint saver in a separate thread. + self.save_thread = threading.Thread( + target=self._save_worker_object_store if self.save_to_object_store else self._save_worker_local, + daemon=False, + args=(state_dict, checkpoint_file, distributed.get_rank()), + ) + self.save_thread.start() + + @misc.timer("checkpoint loading") + def load( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer | None = None, + scheduler: torch.optim.lr_scheduler.LRScheduler | None = None, + grad_scaler: torch.amp.GradScaler | None = None, + ) -> int: + """Load network weights and optimizer states from a checkpoint in a single process. + + The priority of the checkpoint loading logic is: + 1. Attempt to resume training if possible by looking for latest_checkpoint.txt under the same name. + 2. If no latest checkpoint were found, it loads the model weights specified by config_checkpoint.path. + - This is typically used for inference mode. + - If config_checkpoint.load_optimizer_state is True, then also load the optimizer and scheduler states. + 3. If none of the above, randomly initialize the model parameters and train from scratch. + + Args: + model (ImaginaireModel): The PyTorch model. + optimizer (torch.optim.Optimizer | None): The model optimizer (default: None). + scheduler (torch.optim.lr_scheduler.LRScheduler | None): The optimization scheduler (default: None). + grad_scaler (torch.amp.GradScaler | None): The gradient scaler (for mixed precision training). + + Returns: + iteration (int): the iteration number to start/resume from. + """ + latest_checkpoint_file = self._read_latest_checkpoint_file() + if latest_checkpoint_file is not None: + # different from base checkpointer, this support multi-EMA + postfix, _, total_ema_num = model.get_ckpt_postfix() + latest_checkpoint_file = latest_checkpoint_file.replace(".pt", f"{postfix}.pt") + # 1. Resume training from latest_checkpoint.txt under the same name. + checkpoint_dir = ( + self.checkpoint_dir_object_store if self.load_from_object_store else self.checkpoint_dir_local + ) + checkpoint_path = os.path.join(checkpoint_dir, latest_checkpoint_file) + resume = True + else: + if self.load_path: + # 2. Load the module weights specified by config_checkpoint.path. + checkpoint_path = self.load_path + # different from base checkpointer, this support multi-EMA + postfix, _, total_ema_num = model.get_ckpt_postfix() + checkpoint_path = checkpoint_path.replace(".pt", f"{postfix}.pt") + resume = self.load_training_state + else: + # 3. Randomly initialize the model parameters and train from scratch. + checkpoint_path = None + resume = False + # Load checkpoint. + if checkpoint_path is not None: + self._check_checkpoint_exists(checkpoint_path) + if self.load_from_object_store: + log.info(f"Loading checkpoint (object store): {checkpoint_path}") + state_dict = self.object_store_loader.load_object(key=checkpoint_path, type="torch") + log.success(f"Complete loading checkpoint (object store): {checkpoint_path}") + else: + log.info(f"Loading checkpoint (local): {checkpoint_path}") + state_dict = torch.load(checkpoint_path, map_location=lambda storage, loc: storage) + log.success(f"Complete loading checkpoint (local): {checkpoint_path}") + self.callbacks.on_load_checkpoint(model, state_dict=state_dict) + # Load the state dicts. + log.info("- Loading the model...") + log.critical(model.load_state_dict(state_dict["model"], strict=self.strict_resume)) + if resume: + iteration = state_dict["iteration"] + assert optimizer and scheduler + log.info("- Loading the optimizer...") + optimizer.load_state_dict(state_dict["optimizer"]) + log.info("- Loading the scheduler...") + scheduler.load_state_dict(state_dict["scheduler"]) + scheduler.last_epoch = iteration + log.info("- Loading the gradient scaler...") + grad_scaler.load_state_dict(state_dict["grad_scaler"]) + log.success(f"Done with loading the checkpoint (iteration {iteration}).") + else: + iteration = 0 + log.success("Done with loading the checkpoint.") + else: + # Checkpoint not found and not specified. We will train everything from scratch. + iteration = 0 + log.info("Training from scratch.") + torch.cuda.empty_cache() + return iteration + + +# https://github.com/facebookresearch/fvcore/blob/9d683aae73fb899dd35d6cf6720e5ef567761c57/fvcore/common/checkpoint.py +def non_strict_load_model(model: torch.nn.Module, checkpoint_state_dict: dict) -> _IncompatibleKeys: + # workaround https://github.com/pytorch/pytorch/issues/24139 + model_state_dict = model.state_dict() + incorrect_shapes = [] + for k in list(checkpoint_state_dict.keys()): + if k in model_state_dict: + if "_extra_state" in k: # Key introduced by TransformerEngine for FP8 + log.warning(f"Skipping key {k} introduced by TransformerEngine for FP8 in the checkpoint.") + continue + model_param = model_state_dict[k] + # Allow mismatch for uninitialized parameters + if TORCH_VERSION >= (1, 8) and isinstance(model_param, torch.nn.parameter.UninitializedParameter): + continue + if not isinstance(model_param, torch.Tensor): + raise ValueError( + f"Find non-tensor parameter {k} in the model. type: {type(model_param)} {type(checkpoint_state_dict[k])}, please check if this key is safe to skip or not." + ) + + shape_model = tuple(model_param.shape) + shape_checkpoint = tuple(checkpoint_state_dict[k].shape) + if shape_model != shape_checkpoint: + has_observer_base_classes = ( + TORCH_VERSION >= (1, 8) + and hasattr(quantization, "ObserverBase") + and hasattr(quantization, "FakeQuantizeBase") + ) + if has_observer_base_classes: + # Handle the special case of quantization per channel observers, + # where buffer shape mismatches are expected. + def _get_module_for_key(model: torch.nn.Module, key: str) -> torch.nn.Module: + # foo.bar.param_or_buffer_name -> [foo, bar] + key_parts = key.split(".")[:-1] + cur_module = model + for key_part in key_parts: + cur_module = getattr(cur_module, key_part) + return cur_module + + cls_to_skip = ( + ObserverBase, + FakeQuantizeBase, + ) + target_module = _get_module_for_key(model, k) + if isinstance(target_module, cls_to_skip): + # Do not remove modules with expected shape mismatches + # them from the state_dict loading. They have special logic + # in _load_from_state_dict to handle the mismatches. + continue + + incorrect_shapes.append((k, shape_checkpoint, shape_model)) + checkpoint_state_dict.pop(k) + incompatible = model.load_state_dict(checkpoint_state_dict, strict=False) + # Remove keys with "_extra_state" suffix, which are non-parameter items introduced by TransformerEngine for FP8 handling + missing_keys = [k for k in incompatible.missing_keys if "_extra_state" not in k] + unexpected_keys = [k for k in incompatible.unexpected_keys if "_extra_state" not in k] + return _IncompatibleKeys( + missing_keys=missing_keys, + unexpected_keys=unexpected_keys, + incorrect_shapes=incorrect_shapes, + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/cluster_env.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/cluster_env.py new file mode 100644 index 0000000000000000000000000000000000000000..55787cf59d993cdbe8c4e2ddc9e6f60adf6b2d93 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/cluster_env.py @@ -0,0 +1,166 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +from enum import Enum +from functools import lru_cache +from typing import Dict + + +class ClusterType(Enum): + LOCAL = "local" + NGC = "ngc" + SLURM = "slurm" + + +class ClusterEnvInfo(Enum): + BASIC = "basic" + DETAILED = "detailed" + ALL = "all" + + +NGC_ENV_BASIC_VARS = [ + "NGC_JOB_ID", + "NGC_ARRAY_SIZE", + "NGC_GPUS_PER_NODE", +] + +SLURM_ENV_BASIC_VARS = [ + "SLURM_JOB_USER", + "SLURM_JOB_PARTITION", + "SLURM_LOG_DIR", + "SLURM_JOBID", + "SLURM_NNODES", + "SLURM_JOB_NAME", + "SLURM_JOB_NODELIST", + "SLURMD_NODENAME", +] + + +@lru_cache() +def is_local() -> bool: + """ + Check if the code is running on a local machine. + """ + return not is_ngc() and not is_slurm() + + +@lru_cache() +def is_ngc() -> bool: + """ + Check if the code is running on NGC. + """ + return "NGC_ARRAY_SIZE" in os.environ + + +@lru_cache() +def is_slurm() -> bool: + """ + Check if the code is running on SLURM. + """ + return "SLURM_JOB_ID" in os.environ + + +def get_ngc_env(level: ClusterEnvInfo = ClusterEnvInfo.BASIC) -> Dict[str, str]: + """ + Retrieves NVIDIA GPU Cloud (NGC) environment variables based on the specified detail level. + The function filters environment variables to include only those relevant to NGC, + differentiated by the detail level specified. + + Parameters: + level (ClusterInfoLevel): The level of detail for the information returned. + Defaults to ClusterInfoLevel.BASIC. + + Returns: + dict: A dictionary containing the environment variables. If the level is BASIC, + it includes only predefined key variables that are considered basic. + If the level is DETAILED, it includes all environment variables that start + with "NGC_". + + Raises: + ValueError: If an unknown level is specified, an exception is raised indicating that the + level is not recognized. + """ + if level == ClusterEnvInfo.BASIC: + return {k: os.environ[k] for k in NGC_ENV_BASIC_VARS if k in os.environ} + elif level == ClusterEnvInfo.DETAILED: + return {k: os.environ[k] for k in os.environ if k.startswith("NGC_")} + elif level == ClusterEnvInfo.ALL: + return {k: v for k, v in os.environ} + else: + raise ValueError(f"Unknown level {level}") + + +def get_slurm_env(level: ClusterEnvInfo = ClusterEnvInfo.BASIC) -> Dict[str, str]: + """ + Retrieves SLURM environment variables based on the specified detail level. + This function filters the environment variables related to the SLURM job scheduler + environment based on the provided detail level of the cluster information. + + Parameters: + level (ClusterEnvInfo): The detail level of the environment variables to retrieve. + This can be BASIC, DETAILED, or ALL. Defaults to BASIC. + + Returns: + Dict[str, str]: A dictionary containing the SLURM environment variables. The contents of + the dictionary vary based on the level: + - BASIC: Returns predefined key variables important for basic SLURM variables. + - DETAILED: Includes all variables that start with "SLURM_". + - ALL: Returns all environment variables available in the current session. + + Raises: + ValueError: If an unknown level is specified, it raises an exception indicating + that the level is not recognized. + """ + if level == ClusterEnvInfo.BASIC: + return {k: os.environ[k] for k in SLURM_ENV_BASIC_VARS if k in os.environ} + elif level == ClusterEnvInfo.DETAILED: + return {k: os.environ[k] for k in os.environ if k.startswith("SLURM_")} + elif level == ClusterEnvInfo.ALL: + return {k: v for k, v in os.environ.items()} + else: + raise ValueError(f"Unknown level {level}") + + +def get_cluster_env(level: ClusterEnvInfo = ClusterEnvInfo.BASIC) -> Dict[str, str]: + """ + Retrieves a combination of environment variables from the cluster, merging information from + both NVIDIA GPU Cloud (NGC) and SLURM environments based on the specified detail level. + This function provides a unified dictionary of environment settings that are crucial for + applications running in clustered computing environments. + + Parameters: + level (ClusterEnvInfo): The level of detail for the environment variables to be retrieved. + The level can be BASIC, DETAILED, or ALL. Defaults to BASIC. + - BASIC: Gathers basic environment variables from both NGC and SLURM. + - DETAILED: Includes more detailed information from both NGC and SLURM. + - ALL: Combines all available environment variables from the system + with NGC and SLURM specific ones. + + Returns: + Dict[str, str]: A dictionary containing key-value pairs of environment variables. + Initially includes the current working directory under the key 'PWD'. + """ + env_info = { + "PWD": os.getcwd(), # Always include the present working directory. + } + if level == ClusterEnvInfo.ALL: + env_info.update(os.environ) # Adds all system environment variables. + return env_info + + # For BASIC and DETAILED levels, merge environment variables from NGC and SLURM: + env_info.update(get_ngc_env(level)) + env_info.update(get_slurm_env(level)) + return env_info diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/config_helper.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/config_helper.py new file mode 100644 index 0000000000000000000000000000000000000000..ce9358b536afe553fa3264f953209eb780abeac6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/config_helper.py @@ -0,0 +1,219 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import importlib +import importlib.util +import os +import pkgutil +import sys +from dataclasses import fields as dataclass_fields +from dataclasses import is_dataclass +from typing import Any, Dict, Optional + +import attr +import attrs +from hydra import compose, initialize +from hydra.core.config_store import ConfigStore +from hydra.core.global_hydra import GlobalHydra +from omegaconf import DictConfig, OmegaConf + +from cosmos_policy._src.imaginaire.config import Config +from cosmos_policy._src.imaginaire.utils import log + + +def is_attrs_or_dataclass(obj) -> bool: + """ + Check if the object is an instance of an attrs class or a dataclass. + + Args: + obj: The object to check. + + Returns: + bool: True if the object is an instance of an attrs class or a dataclass, False otherwise. + """ + return is_dataclass(obj) or attr.has(type(obj)) + + +def get_fields(obj): + """ + Get the fields of an attrs class or a dataclass. + + Args: + obj: The object to get fields from. Must be an instance of an attrs class or a dataclass. + + Returns: + list: A list of field names. + + Raises: + ValueError: If the object is neither an attrs class nor a dataclass. + """ + if is_dataclass(obj): + return [field.name for field in dataclass_fields(obj)] + elif attr.has(type(obj)): + return [field.name for field in attr.fields(type(obj))] + else: + raise ValueError("The object is neither an attrs class nor a dataclass.") + + +def override(config: Config, overrides: Optional[list[str]] = None, remove_defaults: bool = False) -> Config: + """ + :param config: the instance of class `Config` (usually from `make_config`) + :param overrides: list of overrides for config + :return: the composed instance of class `Config` + """ + # Store the class of the config for reconstruction after overriding. + # config_class = type(config) + + def remove_defaults_filter(f, _): + return f.name != "defaults" + + # Convert Config object to a DictConfig object + config_dict = attrs.asdict(config, filter=remove_defaults_filter if remove_defaults else None) + config_omegaconf = DictConfig(content=config_dict, flags={"allow_objects": True}) + # Enforce "--" separator between the script arguments and overriding configs. + if overrides: + if overrides[0] != "--": + raise ValueError( + f'Hydra config overrides must be separated with a "--" token. but got overrides={overrides}, and overrides[0]={overrides[0]}' + ) + overrides = overrides[1:] + # Use Hydra to handle overrides + cs = ConfigStore.instance() + cs.store(name="config", node=config_omegaconf) + if not GlobalHydra().is_initialized(): + with initialize(version_base=None): + config_omegaconf = compose(config_name="config", overrides=overrides) + OmegaConf.resolve(config_omegaconf) + else: + config_omegaconf = compose(config_name="config", overrides=overrides) + OmegaConf.resolve(config_omegaconf) + + def config_from_dict(ref_instance: Any, kwargs: Any) -> Any: + """ + Construct an instance of the same type as ref_instance using the provided dictionary or data or unstructured data + + Args: + ref_instance: The reference instance to determine the type and fields when needed + kwargs: A dictionary of keyword arguments to use for constructing the new instance or primitive data or unstructured data + + Returns: + Any: A new instance of the same type as ref_instance constructed using the provided kwargs or the primitive data or unstructured data + + Raises: + AssertionError: If the fields do not match or if extra keys are found. + Exception: If there is an error constructing the new instance. + """ + is_type = is_attrs_or_dataclass(ref_instance) + if not is_type: + return kwargs + else: + ref_fields = set(get_fields(ref_instance)) + assert isinstance(kwargs, dict) or isinstance(kwargs, DictConfig), ( + "kwargs must be a dictionary or a DictConfig" + ) + keys = set(kwargs.keys()) + + # ref_fields must equal to or include all keys + extra_keys = keys - ref_fields + assert ref_fields == keys or keys.issubset(ref_fields), ( + f"Fields mismatch: {ref_fields} != {keys}. Extra keys found: {extra_keys} \n \t when constructing {type(ref_instance)} with {keys}" + ) + + resolved_kwargs: Dict[str, Any] = {} + for f in keys: + resolved_kwargs[f] = config_from_dict(getattr(ref_instance, f), kwargs[f]) + try: + new_instance = type(ref_instance)(**resolved_kwargs) + except Exception as e: + log.error(f"Error when constructing {type(ref_instance)} with {resolved_kwargs}") + log.error(e) + raise e + return new_instance + + config = config_from_dict(config, config_omegaconf) + + return config + + +def get_config_module(config_file: str) -> str: + if not config_file.endswith(".py"): + log.error("Config file cannot be specified as module.") + log.error("Please provide the path to the Python config file (relative to the Imaginaire4 root).") + # Convert to importable module format. + config_module = config_file.replace("/", ".").replace(".py", "") + if importlib.util.find_spec(config_module) is None: + raise ValueError(f"Imaginaire4 config module ({config_module}) not found.") + return config_module + + +def import_module(full_module_name: str, reload: bool = False): + """ + Import a module by name. + + Args: + full_module_name: The fully qualified name of the module to import. + reload: If True, reload the module if it's already imported. + """ + if full_module_name in sys.modules and reload: + importlib.reload(sys.modules[full_module_name]) + else: + importlib.import_module(full_module_name) + + +def import_all_modules_from_package(package_path: str, reload: bool = False, skip_underscore: bool = True) -> None: + """ + Import all modules from the specified package path recursively. + + This function is typically used in conjunction with Hydra to ensure that all modules + within a specified package are imported, which is necessary for registering configurations. + + Example usage: + ```python + import_all_modules_from_package("projects.cosmos.diffusion.v1.config.experiment", reload=True, skip_underscore=False) + ``` + + Args: + package_path (str): The dotted path to the package from which to import all modules. + reload (bool): Flag to determine whether to reload modules if they're already imported. + skip_underscore (bool): If True, skips importing modules that start with an underscore. + """ + log.critical(f"{'Reloading' if reload else 'Importing'} all modules from package {package_path}") + package = importlib.import_module(package_path) + package_directory = package.__path__ + + def import_modules_recursively(directory: str, prefix: str) -> None: + """ + Recursively imports or reloads all modules in the given directory. + + Args: + directory (str): The file system path to the current package directory. + prefix (str): The module prefix (e.g., 'projects.cosmos.diffusion.v1.config'). + """ + for _, module_name, is_pkg in pkgutil.iter_modules([directory]): + if skip_underscore and module_name.startswith("_"): + log.debug(f"Skipping module {module_name} as it starts with an underscore") + continue + + full_module_name = f"{prefix}.{module_name}" + log.debug(f"{'Reloading' if reload else 'Importing'} module {full_module_name}") + + import_module(full_module_name, reload=reload) + + if is_pkg: + sub_package_directory = os.path.join(directory, module_name) + import_modules_recursively(sub_package_directory, full_module_name) + + for directory in package_directory: + import_modules_recursively(directory, package_path) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/context_managers.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/context_managers.py new file mode 100644 index 0000000000000000000000000000000000000000..9d6f7e2963958c4fcb010ed7f753ba05d0107d28 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/context_managers.py @@ -0,0 +1,55 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from contextlib import ExitStack, contextmanager +from typing import Generator + +from cosmos_policy._src.imaginaire.utils.misc import timer + + +@contextmanager +def data_loader_init() -> Generator[None, None, None]: + """ + Wrap the data loader initialization with multiple context managers used for telemetry and one logger. + """ + contexts = [ + timer("init_data_loader"), + ] + with ExitStack() as stack: + yield [stack.enter_context(cm) for cm in contexts] + + +@contextmanager +def model_init(set_barrier: bool = False) -> Generator[None, None, None]: + """ + Wrap the instantiation of the model with multiple context managers used for telemetry and one logger. + """ + contexts = [ + timer("init_model"), + ] + with ExitStack() as stack: + yield [stack.enter_context(cm) for cm in contexts] + + +@contextmanager +def distributed_init() -> Generator[None, None, None]: + """ + Wrap the distributed initialization, used for telemetry and timers + """ + contexts = [ + timer("init_distributed"), + ] + with ExitStack() as stack: + yield [stack.enter_context(cm) for cm in contexts] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/context_parallel.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/context_parallel.py new file mode 100644 index 0000000000000000000000000000000000000000..8ee3995f03f4e7454d90b562d6fcdfef45936bc4 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/context_parallel.py @@ -0,0 +1,255 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math +from typing import Optional + +try: + import megatron.core.parallel_state as parallel_state + + USE_MEGATRON = True +except ImportError: + USE_MEGATRON = False + +import torch +from torch import Tensor +from torch.distributed import ProcessGroup, all_gather, broadcast_object_list, get_process_group_ranks, get_world_size +from torch.distributed.utils import _verify_param_shape_across_processes + +from cosmos_policy._src.imaginaire.utils import distributed + + +def split_inputs_cp(x: Tensor, seq_dim: int, cp_group: ProcessGroup) -> Tensor: + """ + Split input tensor along the sequence dimension for checkpoint parallelism. + + This function divides the input tensor into equal parts along the specified + sequence dimension, based on the number of ranks in the checkpoint parallelism group. + It then selects the part corresponding to the current rank. + + Args: + x: Input tensor to be split. + seq_dim: The dimension along which to split the input (sequence dimension). + cp_group: The process group for checkpoint parallelism. + + Returns: + A slice of the input tensor corresponding to the current rank. + + Raises: + AssertionError: If the sequence dimension is not divisible by the number of ranks. + """ + cp_ranks = get_process_group_ranks(cp_group) + cp_size = len(cp_ranks) + + assert x.shape[seq_dim] % cp_size == 0, f"{x.shape[seq_dim]} cannot divide cp_size {cp_size}" + x = x.view(*x.shape[:seq_dim], cp_size, x.shape[seq_dim] // cp_size, *x.shape[(seq_dim + 1) :]) + seq_idx = torch.tensor([cp_group.rank()], device=x.device) + x = x.index_select(seq_dim, seq_idx) + # Note that the new sequence length is the original sequence length / cp_size + x = x.view(*x.shape[:seq_dim], -1, *x.shape[(seq_dim + 2) :]) + return x + + +@torch.compiler.disable +def cat_outputs_cp(x: Tensor, seq_dim: int, cp_group: ProcessGroup) -> Tensor: + """ + Concatenate outputs from different ranks in the checkpoint parallelism group. + + This function gathers tensors from all ranks in the checkpoint parallelism group + and concatenates them along the specified sequence dimension. + + The function is decorated with @torch.compiler.disable because it contains distributed + operations and dynamic tensor creation based on runtime rank information that seem to be + incompatible with torch.compile's static graph compilation. + + Args: + x: Input tensor to be concatenated. + seq_dim: The dimension along which to concatenate the tensors (sequence dimension). + cp_group: The process group for checkpoint parallelism. + + Returns: + A tensor that is the concatenation of tensors from all ranks in the cp_group. + + Raises: + RuntimeError: If the gather operation fails. + """ + # Get the world size (number of processes in the group) + world_size = get_world_size(cp_group) + + # Create a list to store tensors from all ranks + gathered_tensors = [torch.zeros_like(x) for _ in range(world_size)] + + # Gather tensors from all ranks + try: + all_gather(gathered_tensors, x, group=cp_group) + except RuntimeError as e: + raise RuntimeError(f"Failed to gather tensors: {e}") + + # Concatenate the gathered tensors along the specified dimension + return torch.cat(gathered_tensors, dim=seq_dim) + + +def cat_outputs_cp_with_grad(x: Tensor, seq_dim: int, cp_group: ProcessGroup) -> Tensor: + """ + Concatenate outputs from different ranks in the context parallelism group. + + This function gathers tensors from all ranks in the checkpoint parallelism group + and concatenates them along the specified sequence dimension. + + It retains computational graph locally for each rank by replacing the portion of the tensor with original output. + + Args: + x: Input tensor to be concatenated. + seq_dim: The dimension along which to concatenate the tensors (sequence dimension). + cp_group: The process group for checkpoint parallelism. + + Returns: + A tensor that is the concatenation of tensors from all ranks in the cp_group. + + Raises: + RuntimeError: If the gather operation fails. + """ + # Get the world size (number of processes in the group) + cp_size = cp_group.size() + assert cp_size > 0, "cp_size should be greater than 0" + + # Create a list to store tensors from all ranks + gathered_tensors = [torch.zeros_like(x) for _ in range(cp_size)] + + # Gather tensors from all ranks + try: + all_gather(gathered_tensors, x, group=cp_group) + except RuntimeError as e: + raise RuntimeError(f"Failed to gather tensors: {e}") + + rank = cp_group.rank() + gathered_tensors[rank] = x + # Concatenate the gathered tensors along the specified dimension + return torch.cat(gathered_tensors, dim=seq_dim) + + +@torch.compiler.disable +def robust_broadcast(tensor: torch.Tensor, src: int, pg: ProcessGroup, is_check_shape: bool = False) -> torch.Tensor: + """ + Perform a robust broadcast operation that works regardless of tensor shapes on different ranks. + + The function is decorated with @torch.compiler.disable because it contains distributed + operations and dynamic tensor creation based on runtime rank information that seem to be + incompatible with torch.compile's static graph compilation. + + Args: + tensor (torch.Tensor): The tensor to broadcast (on src rank) or receive (on other ranks). + src (int): The source rank for the broadcast. Defaults to 0. + + Returns: + torch.Tensor: The broadcasted tensor on all ranks. + """ + # First, broadcast the shape of the tensor + if distributed.get_rank() == src: + shape = torch.tensor(tensor.shape, dtype=torch.long).cuda() + else: + shape = torch.empty(tensor.dim(), dtype=torch.long).cuda() + if is_check_shape: + _verify_param_shape_across_processes(pg, [shape]) + torch.distributed.broadcast(shape, src, group=pg) + + # Resize the tensor on non-src ranks if necessary + if distributed.get_rank() != src: + tensor = tensor.new_empty(shape.tolist()).type_as(tensor) + + # Now broadcast the tensor data + torch.distributed.broadcast(tensor, src, group=pg) + + return tensor + + +def broadcast( + item: torch.Tensor | str | None, process_group: Optional[ProcessGroup] = None +) -> torch.Tensor | str | None: + """ + Broadcast the item from the minimum rank in the specified group(s). + """ + if process_group is None: + return item + + min_rank = min(get_process_group_ranks(process_group)) + if isinstance(item, torch.Tensor): # assume the device is cuda + item = robust_broadcast(item, min_rank, process_group) + elif item is not None: + broadcastable_list = [item] + broadcast_object_list(broadcastable_list, min_rank, group=process_group) + item = broadcastable_list[0] + return item + + +def broadcast_split_tensor( + tensor: torch.Tensor, + seq_dim: int, + process_group: Optional[ProcessGroup] = None, +) -> torch.Tensor: + """ + Broadcast the tensor from the minimum rank in the specified group(s). + """ + if tensor is None: + return tensor + min_rank = min(get_process_group_ranks(process_group)) + tensor = robust_broadcast(tensor, min_rank, process_group) + return split_inputs_cp(tensor, seq_dim, process_group) + + +def find_split( + shape_tensor: torch.Size, cp_size: int, patch_values: tuple[int, int, int] = (1, 2, 2), view_factor: int = 1 +) -> torch.Size: + """ + Find the shape of input tensor for post-CP split, taking into account both temporal and spatial split, as well as patching values. + The split by width is not possible currently, due to memory stride issues, which break quality. This is checked + by an assert. + + The spatial split is achieved by flattening the input video into a single dimension before CP split is performed, + and rearranging it back into [T, H, W] format after the CP split, since the input passed to the model must still be in [T, H, W] format. + + Args: + shape_tensor (torch.Size): The shape of the Tensor that we want to split. Needs to be in [B, C, T, H, W] format. + cp_size (int): The Context Parallelism size that we want to use. + patch_values (tuple[int, int, int], optional): The patch values that are applied inside the Diffusion model. + First element of the tuple is temporal patch size. Two next elements are the spatial patch sizes. + The default value is (1, 2, 2) + view_factor (int, optional): The number of views that are present in the temporal dimension. Default value is 1. + + Returns: + The torch.Size of how the post-split tensor should look like in [T, H, W] dimensions. + + """ + if not USE_MEGATRON: + raise ImportError("No megatron.core package found, which is required for Context Parallelism usage.") + B, C, T, H, W = shape_tensor + ret = [] + assert T % view_factor == 0 + T = T // view_factor + cp_size_t = 1 + for i, size in enumerate([T, H, W]): + if i == 2 and cp_size > 1: + raise ValueError( + f"Split by width dimension is not currently supported due to quality issues. Width dimension would be split by a factor of {cp_size}. Lower the CP size to avoid splitting by width." + ) + patch_size = patch_values[i] + gcd = math.gcd(size // patch_size, cp_size) + cp_size = cp_size // gcd + if i == 0: + cp_size_t = gcd + ret.append(size // gcd) + # Saving the CP size in the temporal dimension for VideoPositionEmb embeddings calculation + parallel_state.cp_size_t = cp_size_t + return torch.Size(ret) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/count_params.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/count_params.py new file mode 100644 index 0000000000000000000000000000000000000000..79c03199e8cb19a141659e394042a5dccb678413 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/count_params.py @@ -0,0 +1,23 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from torch import nn + + +def count_params(model: nn.Module, verbose=False) -> int: + total_params = sum(p.numel() for p in model.parameters() if p.requires_grad) + if verbose: + print(f"{model.__class__.__name__} has {total_params * 1.0e-6:.2f} M params.") + return total_params diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/dataloader.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/dataloader.py new file mode 100644 index 0000000000000000000000000000000000000000..b2bac67619b8e9399de578d2e16cd018becfb249 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/dataloader.py @@ -0,0 +1,105 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Iterable, Iterator + +import torch +import torch.distributed as dist +import torch.utils.data + + +class MultiEpochsDataLoader(torch.utils.data.DataLoader): + """A dataloader that relentlessly samples from the dataset. + + This eliminates the overhead of prefetching data before each epoch. + Ref: https://github.com/rwightman/pytorch-image-models/blob/master/timm/data/loader.py + """ + + def __init__(self, *args, **kwargs) -> None: + super().__init__(*args, **kwargs) + self._DataLoader__initialized = False + if self.batch_sampler is None: + self.sampler = _RepeatSampler(self.sampler) # type: ignore + else: + self.batch_sampler = _RepeatSampler(self.batch_sampler) # type: ignore + self._DataLoader__initialized = True + self.iterator = super().__iter__() + + def __len__(self) -> int: + return len(self.sampler) if self.batch_sampler is None else len(self.batch_sampler.sampler) # type: ignore + + def __iter__(self) -> Iterable: + for _ in range(len(self)): + yield next(self.iterator) + + +class _RepeatSampler: + """A sampler wrapper that repeats data sampling forever. + + Args: + sampler (Sampler): Data sampler object. + """ + + def __init__(self, sampler: torch.utils.data.Sampler): + self.sampler = sampler + + def __iter__(self) -> Iterator: + while True: + yield from iter(self.sampler) + + +class DistributedEvalSampler(torch.utils.data.Sampler): + """Distributed data sampler for evaluation. + + Ref: https://github.com/SeungjunNah/DeepDeblur-PyTorch/blob/master/src/data/sampler.py (by snah) + DistributedEvalSampler is different from DistributedSampler in that it does not pad extra samples to make it + evenly divisible. It should not be used for training, or the distributed processes could hang forever. + DistributedEvalSampler is for evaluation purpose where synchronization does not happen every epoch. + Synchronization should be done outside the dataloader loop. + """ + + def __init__(self, dataset: torch.utils.data.Dataset, shuffle: bool = False, seed: int = 0): + """Constructor of DistributedEvalSampler, + + Args: + dataset (torch.utils.data.Dataset): Dataset used for sampling. + shuffle (bool): Whether to shuffle the indices (default: False). + seed (int): Random seed for shuffling if enabled (default: 0). + """ + self.dataset = dataset + self.num_replicas = dist.get_world_size() + self.rank = dist.get_rank() + self.dataset_size = len(self.dataset) # type: ignore + indices = list(range(self.dataset_size)) + indices = indices[self.rank : self.dataset_size : self.num_replicas] + self.num_samples = len(indices) + self.shuffle = shuffle + self.seed = seed + + def __iter__(self) -> Iterator: + if self.shuffle: + # Deterministically shuffle based on epoch and seed. + gen = torch.Generator() + gen.manual_seed(self.seed) + indices = torch.randperm(self.dataset_size, generator=gen).tolist() + else: + indices = list(range(self.dataset_size)) + # Subsample. + indices = indices[self.rank : self.dataset_size : self.num_replicas] + assert len(indices) == self.num_samples + return iter(indices) + + def __len__(self) -> int: + return self.num_samples diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/dataset_utils.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/dataset_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..04e313ec67738b9c1384c161a37c0b699a08c02e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/dataset_utils.py @@ -0,0 +1,345 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Adapted from: +https://github.com/bytedance/IRASim/blob/main/dataset/dataset_util.py +""" + +import base64 +import math +import os +from io import BytesIO + +import numpy as np +import torch +import torch.distributed as dist +import torchvision.transforms.functional as F +from PIL import Image + + +def is_dist_avail_and_initialized(): + if not dist.is_available(): + return False + if not dist.is_initialized(): + return False + return True + + +def get_rank(): + if not is_dist_avail_and_initialized(): + return 0 + return dist.get_rank() + + +def get_1d_sincos_pos_embed_from_grid(embed_dim, pos): + """ + embed_dim: output dimension for each position + pos: a list of positions to be encoded: size (M,) + out: (M, D) + """ + assert embed_dim % 2 == 0 + omega = np.arange(embed_dim // 2, dtype=np.float32) + omega /= embed_dim / 2.0 + omega = 1.0 / 10000**omega # (D/2,) + + pos = pos.reshape(-1) # (M,) + out = np.einsum("m,d->md", pos, omega) # (M, D/2), outer product + + emb_sin = np.sin(out) # (M, D/2) + emb_cos = np.cos(out) # (M, D/2) + + emb = np.concatenate([emb_sin, emb_cos], axis=1) # (M, D) + return emb + + +def get_2d_sincos_pos_embed_from_grid(embed_dim, grid): + assert embed_dim % 2 == 0 + + # use half of dimensions to encode grid_h + emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0]) # (H*W, D/2) + emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1]) # (H*W, D/2) + + emb = np.concatenate([emb_h, emb_w], axis=1) # (H*W, D) + return emb + + +def get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False): + """ + grid_size: int of the grid height and width + return: + pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token) + """ + grid_h = np.arange(grid_size, dtype=np.float32) + grid_w = np.arange(grid_size, dtype=np.float32) + grid = np.meshgrid(grid_w, grid_h) # here w goes first + grid = np.stack(grid, axis=0) + + grid = grid.reshape([2, 1, grid_size, grid_size]) + pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid) + if cls_token: + pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0) + return pos_embed + + +def b64_2_img(data: str): + image_b64 = base64.b64decode(data) + img = Image.open(BytesIO(image_b64)).convert("RGB") + return img + + +def get_continuous_action(d_acts, c_act_max, c_act_min, n_bins): + c_act_max = c_act_max.to(d_acts.device) + c_act_min = c_act_min.to(d_acts.device) + c_acts = d_acts / (n_bins - 1) * (c_act_max - c_act_min) + c_act_min + return c_acts + + +def alpha2rotm(a): + """Alpha euler angle to rotation matrix.""" + rotm = np.array([[1, 0, 0], [0, np.cos(a), -np.sin(a)], [0, np.sin(a), np.cos(a)]]) + return rotm + + +def beta2rotm(b): + """Beta euler angle to rotation matrix.""" + rotm = np.array([[np.cos(b), 0, np.sin(b)], [0, 1, 0], [-np.sin(b), 0, np.cos(b)]]) + return rotm + + +def gamma2rotm(c): + """Gamma euler angle to rotation matrix.""" + rotm = np.array([[np.cos(c), -np.sin(c), 0], [np.sin(c), np.cos(c), 0], [0, 0, 1]]) + return rotm + + +def euler2rotm(euler_angles): + """Euler angle (ZYX) to rotation matrix.""" + alpha = euler_angles[0] + beta = euler_angles[1] + gamma = euler_angles[2] + + rotm_a = alpha2rotm(alpha) + rotm_b = beta2rotm(beta) + rotm_c = gamma2rotm(gamma) + + rotm = rotm_c @ rotm_b @ rotm_a + + return rotm + + +def isRotm(R): + # Checks if a matrix is a valid rotation matrix. + # Forked from Andy Zeng + Rt = np.transpose(R) + shouldBeIdentity = np.dot(Rt, R) + I = np.identity(3, dtype=R.dtype) + n = np.linalg.norm(I - shouldBeIdentity) + return n < 1e-6 + + +def rotm2euler(R): + # Forked from: https://learnopencv.com/rotation-matrix-to-euler-angles/ + # R = Rz * Ry * Rx + assert isRotm(R) + sy = math.sqrt(R[0, 0] * R[0, 0] + R[1, 0] * R[1, 0]) + singular = sy < 1e-6 + + if not singular: + x = math.atan2(R[2, 1], R[2, 2]) + y = math.atan2(-R[2, 0], sy) + z = math.atan2(R[1, 0], R[0, 0]) + else: + x = math.atan2(-R[1, 2], R[1, 1]) + y = math.atan2(-R[2, 0], sy) + z = 0 + + # (-pi , pi] + while x > np.pi: + x -= 2 * np.pi + while x <= -np.pi: + x += 2 * np.pi + while y > np.pi: + y -= 2 * np.pi + while y <= -np.pi: + y += 2 * np.pi + while z > np.pi: + z -= 2 * np.pi + while z <= -np.pi: + z += 2 * np.pi + return np.array([x, y, z]) + + +def quat2rotm(quat): + """Quaternion to rotation matrix. + + Args: + quat (4, numpy array): quaternion x, y, z, w + Returns: + rotm (3x3 numpy array): rotation matrix + """ + w = quat[3] + x = quat[0] + y = quat[1] + z = quat[2] + + s = w * w + x * x + y * y + z * z + + rotm = np.array( + [ + [1 - 2 * (y * y + z * z) / s, 2 * (x * y - z * w) / s, 2 * (x * z + y * w) / s], + [2 * (x * y + z * w) / s, 1 - 2 * (x * x + z * z) / s, 2 * (y * z - x * w) / s], + [2 * (x * z - y * w) / s, 2 * (y * z + x * w) / s, 1 - 2 * (x * x + y * y) / s], + ] + ) + + return rotm + + +def rotm2quat(R): + """Convert 3x3 rotation matrix to quaternion (w, x, y, z).""" + R = np.array(R, dtype=float) + trace = np.trace(R) + + if trace > 0: + s = 0.5 / np.sqrt(trace + 1.0) + w = 0.25 / s + x = (R[2, 1] - R[1, 2]) * s + y = (R[0, 2] - R[2, 0]) * s + z = (R[1, 0] - R[0, 1]) * s + else: + if R[0, 0] > R[1, 1] and R[0, 0] > R[2, 2]: + s = 2.0 * np.sqrt(1.0 + R[0, 0] - R[1, 1] - R[2, 2]) + w = (R[2, 1] - R[1, 2]) / s + x = 0.25 * s + y = (R[0, 1] + R[1, 0]) / s + z = (R[0, 2] + R[2, 0]) / s + elif R[1, 1] > R[2, 2]: + s = 2.0 * np.sqrt(1.0 + R[1, 1] - R[0, 0] - R[2, 2]) + w = (R[0, 2] - R[2, 0]) / s + x = (R[0, 1] + R[1, 0]) / s + y = 0.25 * s + z = (R[1, 2] + R[2, 1]) / s + else: + s = 2.0 * np.sqrt(1.0 + R[2, 2] - R[0, 0] - R[1, 1]) + w = (R[1, 0] - R[0, 1]) / s + x = (R[0, 2] + R[2, 0]) / s + y = (R[1, 2] + R[2, 1]) / s + z = 0.25 * s + + return np.array([w, x, y, z]) + + +def get_converted_fp32_paths(deepspeed_ckpt_path): + deepspeed_ckpt_path = deepspeed_ckpt_path.rstrip("/") + ckpt_dir = os.path.dirname(deepspeed_ckpt_path) + ckpt_name = os.path.basename(deepspeed_ckpt_path) + fp32_ckpt_name = f"{ckpt_name}.fp32.pt" + converted_path = os.path.join(ckpt_dir, fp32_ckpt_name) + return converted_path + + +class Resize_Preprocess: + def __init__(self, size): + """ + Initialize the preprocessing class with the target size. + Args: + size (tuple): The target height and width as a tuple (height, width). + """ + self.size = size + + def __call__(self, video_frames): + """ + Apply the transformation to each frame in the video. + Args: + video_frames (torch.Tensor): A tensor representing a batch of video frames. + Returns: + torch.Tensor: The transformed video frames. + """ + # Resize each frame in the video + resized_frames = torch.stack([F.resize(frame, self.size, antialias=True) for frame in video_frames]) + return resized_frames + + +class Preprocess: + def __init__(self, size): + self.size = size + + def __call__(self, clip): + clip = Preprocess.resize_scale(clip, self.size[0], self.size[1], interpolation_mode="bilinear") + return clip + + def __repr__(self) -> str: + return f"{self.__class__.__name__}(size={self.size})" + + @staticmethod + def resize_scale(clip, target_height, target_width, interpolation_mode): + target_ratio = target_height / target_width + H = clip.size(-2) + W = clip.size(-1) + clip_ratio = H / W + if clip_ratio > target_ratio: + scale_ = target_width / W + else: + scale_ = target_height / H + return torch.nn.functional.interpolate(clip, scale_factor=scale_, mode=interpolation_mode, align_corners=False) + + +class ToTensorVideo: + """ + Convert tensor data type from uint8 to float, divide value by 255.0 and + permute the dimensions of clip tensor + """ + + def __init__(self): + pass + + def __call__(self, clip): + """ + Args: + clip (torch.tensor, dtype=torch.uint8): Size is (T, C, H, W) + Return: + clip (torch.tensor, dtype=torch.float): Size is (T, C, H, W) + """ + return to_tensor(clip) + + def __repr__(self) -> str: + return self.__class__.__name__ + + +def to_tensor(clip): + """ + Convert tensor data type from uint8 to float, divide value by 255.0 and + permute the dimensions of clip tensor + Args: + clip (torch.tensor, dtype=torch.uint8): Size is (T, C, H, W) + Return: + clip (torch.tensor, dtype=torch.float): Size is (T, C, H, W) + """ + _is_tensor_video_clip(clip) + if not clip.dtype == torch.uint8: + raise TypeError("clip tensor should have data type uint8. Got %s" % str(clip.dtype)) + # return clip.float().permute(3, 0, 1, 2) / 255.0 + return clip.float() / 255.0 + + +def _is_tensor_video_clip(clip): + if not torch.is_tensor(clip): + raise TypeError("clip should be Tensor. Got %s" % type(clip)) + + if not clip.ndimension() == 4: + raise ValueError("clip should be 4D. Got %dD" % clip.dim()) + + return True diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/denoise_prediction.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/denoise_prediction.py new file mode 100644 index 0000000000000000000000000000000000000000..a209db0eba28a8d8bcb527bfbaca6f5e361ace14 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/denoise_prediction.py @@ -0,0 +1,28 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +from dataclasses import dataclass +from typing import Optional + +import torch + + +@dataclass +class DenoisePrediction: + x0: torch.Tensor # clean data prediction + eps: Optional[torch.Tensor] = None # noise prediction + logvar: Optional[torch.Tensor] = None # log variance of noise prediction, can be used a confidence / uncertainty diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/device.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/device.py new file mode 100644 index 0000000000000000000000000000000000000000..515ea78b1e1b05a16fa75774eb4aa8bcd6250b15 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/device.py @@ -0,0 +1,114 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import gc +import math +import os + +import pynvml +from loguru import logger as logging + + +def get_gpu_architecture(): + """ + Retrieves the GPU architecture of the available GPUs. + + Returns: + str: The GPU architecture, which can be "H100", "A100", or "Other". + """ + try: + pynvml.nvmlInit() + device_count = pynvml.nvmlDeviceGetCount() + for i in range(device_count): + handle = pynvml.nvmlDeviceGetHandleByIndex(i) + model_name = pynvml.nvmlDeviceGetName(handle) + if isinstance(model_name, bytes): + model_name = model_name.decode("utf-8") + print(f"GPU {i}: Model: {model_name}") + + # Check for specific models like H100 or A100 + if "H100" in model_name or "H200" in model_name: + return "H100" + elif "A100" in model_name: + return "A100" + elif "L40S" in model_name: + return "L40S" + elif "B200" in model_name: + return "B200" + except pynvml.NVMLError as error: + print(f"Failed to get GPU info: {error}") + finally: + pynvml.nvmlShutdown() + + # return "Other" incase of non hopper/ampere or error + return "Other" + + +class GPUArchitectureNotSupported(Exception): + """ + Custom exception raised when the expected GPU architecture is not supported. + """ + + pass + + +def print_gpu_mem(str=None): + try: + pynvml.nvmlInit() + meminfo = pynvml.nvmlDeviceGetMemoryInfo(pynvml.nvmlDeviceGetHandleByIndex(0)) + logging.info( + f"{str}: {meminfo.used / 1024 / 1024}/{meminfo.total / 1024 / 1024}MiB used ({meminfo.free / 1024 / 1024}MiB free)" + ) + except pynvml.NVMLError as error: + print(f"Failed to get GPU memory info: {error}") + + +def force_gc(): + print_gpu_mem() + print("gc()") + gc.collect() + print_gpu_mem() + print("empty cuda cache") + # print(torch.cuda.memory_summary()) + print_gpu_mem() + + +def gpu0_has_80gb_or_less(): + try: + pynvml.nvmlInit() + meminfo = pynvml.nvmlDeviceGetMemoryInfo(pynvml.nvmlDeviceGetHandleByIndex(0)) + return meminfo.total / 1024 / 1024 / 1024 <= 80 + except pynvml.NVMLError as error: + print(f"Failed to get GPU memory info: {error}") + + +class Device: + _nvml_affinity_elements = math.ceil(os.cpu_count() / 64) # type: ignore + + def __init__(self, device_idx: int): + super().__init__() + self.handle = pynvml.nvmlDeviceGetHandleByIndex(device_idx) + + def get_name(self) -> str: + return pynvml.nvmlDeviceGetName(self.handle) + + def get_cpu_affinity(self) -> list[int]: + affinity_string = "" + for j in pynvml.nvmlDeviceGetCpuAffinity(self.handle, Device._nvml_affinity_elements): + # assume nvml returns list of 64 bit ints + affinity_string = "{:064b}".format(j) + affinity_string + affinity_list = [int(x) for x in affinity_string] + affinity_list.reverse() # so core 0 is in 0th element of list + return [i for i, e in enumerate(affinity_list) if e != 0] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/disabled_train.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/disabled_train.py new file mode 100644 index 0000000000000000000000000000000000000000..af961d31dc0932f0b86b546d6a3143a2db80b363 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/disabled_train.py @@ -0,0 +1,22 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any + + +def disabled_train(self: Any, mode: bool = True) -> Any: + """Overwrite model.train with this function to make sure train/eval mode + does not change anymore.""" + return self diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/distributed.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/distributed.py new file mode 100644 index 0000000000000000000000000000000000000000..e52f096f5dac114fdbf2ff469965e89a36858d78 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/distributed.py @@ -0,0 +1,443 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import collections +import collections.abc +import ctypes +import functools +import os +from contextlib import contextmanager +from datetime import timedelta +from typing import TYPE_CHECKING, Any, Callable, Container, Optional + +import pynvml +import torch +import torch.distributed as dist +from torch.distributed import get_process_group_ranks + +from cosmos_policy._src.imaginaire.utils.device import Device + +if dist.is_available(): + from torch.distributed.distributed_c10d import _get_default_group + from torch.distributed.utils import _sync_module_states, _verify_param_shape_across_processes + +from cosmos_policy._src.imaginaire.utils import log + +if TYPE_CHECKING: + from cosmos_policy._src.imaginaire.config import DDPConfig + +try: + from megatron.core import parallel_state +except ImportError: + print("Megatron-core is not installed.") + + +def init() -> int | None: + """Initialize distributed training.""" + if dist.is_initialized(): + return torch.cuda.current_device() + + # Set GPU affinity. + pynvml.nvmlInit() + local_rank = int(os.getenv("LOCAL_RANK", 0)) + try: + device = Device(local_rank) + os.sched_setaffinity(0, device.get_cpu_affinity()) + except pynvml.NVMLError as e: + log.warning(f"Failed to set device affinity: {e}") + # Set up NCCL communication. + os.environ["TORCH_NCCL_BLOCKING_WAIT"] = "0" + os.environ["TORCH_NCCL_ASYNC_ERROR_HANDLING"] = "1" + if dist.is_available(): + torch.cuda.set_device(local_rank) + # Get the timeout value from environment variable + timeout_seconds = os.getenv("TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC", 1800) + # Convert the timeout to an integer (if it isn't already) and then to a timedelta + timeout_timedelta = timedelta(seconds=int(timeout_seconds)) + dist.init_process_group(backend="nccl", init_method="env://", timeout=timeout_timedelta) + log.critical( + f"Initialized distributed training with local rank {local_rank} with timeout {timeout_seconds}", + rank0_only=False, + ) + # Increase the L2 fetch granularity for faster speed. + _libcudart = ctypes.CDLL("libcudart.so") + # Set device limit on the current device. + p_value = ctypes.cast((ctypes.c_int * 1)(), ctypes.POINTER(ctypes.c_int)) + _libcudart.cudaDeviceSetLimit(ctypes.c_int(0x05), ctypes.c_int(128)) + _libcudart.cudaDeviceGetLimit(p_value, ctypes.c_int(0x05)) + log.info(f"Training with {get_world_size()} GPUs.") + + +def get_rank(group: Optional[dist.ProcessGroup] = None) -> int: + """Get the rank (GPU device) of the worker. + + Returns: + rank (int): The rank of the worker. + """ + rank = 0 + if dist.is_available() and dist.is_initialized(): + rank = dist.get_rank(group) + return rank + + +def get_world_size(group: Optional[dist.ProcessGroup] = None) -> int: + """Get world size. How many GPUs are available in this job. + + Returns: + world_size (int): The total number of GPUs available in this job. + """ + world_size = 1 + if dist.is_available() and dist.is_initialized(): + world_size = dist.get_world_size(group) + return world_size + + +def is_rank0() -> bool: + """Check if current process is the master GPU. + + Returns: + (bool): True if this function is called from the master GPU, else False. + """ + return get_rank() == 0 + + +def is_local_rank0() -> bool: + """Check if current process is the local master GPU in the current node. + + Returns: + (bool): True if this function is called from the local master GPU, else False. + """ + return torch.cuda.current_device() == 0 + + +def rank0_only(func: Callable) -> Callable: + """Apply this function only to the master GPU. + + Example usage: + @rank0_only + def func(x): + return x + 3 + + Args: + func (Callable): a function. + + Returns: + (Callable): A function wrapper executing the function only on the master GPU. + """ + + @functools.wraps(func) + def wrapper(*args, **kwargs): # noqa: ANN202 + if is_rank0(): + return func(*args, **kwargs) + else: + return None + + return wrapper + + +def barrier() -> None: + """Barrier for all GPUs.""" + if dist.is_available() and dist.is_initialized(): + dist.barrier() + + +def rank0_first(func: Callable) -> Callable: + """run the function on rank 0 first, then on other ranks.""" + + @functools.wraps(func) + def wrapper(*args, **kwargs): # noqa: ANN202 + if is_rank0(): + result = func(*args, **kwargs) + barrier() + if not is_rank0(): + result = func(*args, **kwargs) + return result + + return wrapper + + +def parallel_model_wrapper(config_ddp: DDPConfig, model: torch.nn.Module) -> torch.nn.Module | DistributedDataParallel: + """Wraps the model to enable data parallalism for training across multiple GPU devices. + + Args: + config_ddp (DDPConfig): The data parallel config. + model (torch.nn.Module): The PyTorch module. + + Returns: + model (torch.nn.Module | DistributedDataParallel): The data parallel model wrapper + if distributed environment is available, otherwise return the original model. + """ + if dist.is_available() and dist.is_initialized(): + local_rank = int(os.getenv("LOCAL_RANK", 0)) + try: + ddp_group = parallel_state.get_data_parallel_group(with_context_parallel=True) + except Exception as e: + log.info(e) + log.info("parallel_state not initialized, treating all GPUs equally for DDP") + ddp_group = None + + model = DistributedDataParallel( + model, + device_ids=[local_rank], + output_device=local_rank, + find_unused_parameters=config_ddp.find_unused_parameters, + static_graph=config_ddp.static_graph, + broadcast_buffers=config_ddp.broadcast_buffers, + process_group=ddp_group, + ) + return model + + +class DistributedDataParallel(torch.nn.parallel.DistributedDataParallel): + """This extends torch.nn.parallel.DistributedDataParallel with .training_step(). + + This borrows the concept of `forward-redirection` from Pytorch lightning. It wraps an ImaginaireModel such that + model.training_step() would be executed when calling self.training_step(), while preserving the behavior of calling + model() for Pytorch modules. Internally, this is a double rerouting mechanism (training_step -> forward -> + training_step), allowing us to preserve the function names and signatures. + """ + + def __init__(self, model: torch.nn.Module, *args, **kwargs): + super().__init__(model, *args, **kwargs) + self.show_sync_grad_static_graph_warning = True + + def training_step(self, *args, **kwargs) -> Any: + # Cache the original model.forward() method. + original_forward = self.module.forward + + def wrapped_training_step(*_args, **_kwargs): # noqa: ANN202 + # Unpatch immediately before calling training_step() because itself may want to call the real forward. + self.module.forward = original_forward + # The actual .training_step(). + return self.module.training_step(*_args, **_kwargs) + + # Patch the original_module's forward so we can redirect the arguments back to the real method. + self.module.forward = wrapped_training_step + # Call self, which implicitly calls self.forward() --> model.forward(), which is now model.training_step(). + # Without calling self.forward() or model.forward() explciitly, implicit hooks are also executed. + return self(*args, **kwargs) + + +@contextmanager +def ddp_sync_grad(model, enabled): + r""" + Context manager to enable/disable gradient synchronizations across DDP processes for DDP model. + Modified from: + https://pytorch.org/docs/stable/_modules/torch/nn/parallel/distributed.html#DistributedDataParallel.no_sync + Note that this is incompatible with static_graph=True and will be an no-op if static_graph=True. + + Within this context, gradients will be accumulated on module + variables, which will later be synchronized in the first + forward-backward pass exiting the context. + + .. warning:: + The forward pass should be included inside the context manager, or + else gradients will still be synchronized. + """ + assert isinstance(model, torch.nn.Module) + if isinstance(model, DistributedDataParallel): + old_require_backward_grad_sync = model.require_backward_grad_sync + if model.static_graph and model.require_backward_grad_sync != enabled: + if model.show_sync_grad_static_graph_warning: + log.warning("DDP static_graph=True is incompatible with sync_grad(). Performance will be reduced.") + model.show_sync_grad_static_graph_warning = False + else: + model.require_backward_grad_sync = enabled + try: + yield + finally: + if isinstance(model, DistributedDataParallel): + model.require_backward_grad_sync = old_require_backward_grad_sync + + +def collate_batches(data_batches: list[dict[str, torch.Tensor]]) -> torch.Tensor | dict[str, torch.Tensor]: + """Aggregate the list of data batches from all devices and process the results. + + This is used for gathering validation data batches with cosmos_policy._src.imaginaire.utils.dataloader.DistributedEvalSampler. + It will return the data/output of the entire validation set in its original index order. The sizes of data_batches + in different ranks may differ by 1 (if dataset size is not evenly divisible), in which case a dummy sample will be + created before calling dis.all_gather(). + + Args: + data_batches (list[dict[str, torch.Tensor]]): List of tensors or (hierarchical) dictionary where + leaf entries are tensors. + + Returns: + data_gather (torch.Tensor | dict[str, torch.Tensor]): tensors or (hierarchical) dictionary where + leaf entries are concatenated tensors. + """ + if isinstance(data_batches[0], torch.Tensor): + # Concatenate the local data batches. + data_concat = torch.cat(data_batches, dim=0) # type: ignore + # Get the largest number of local samples from all ranks to determine whether to dummy-pad on this rank. + max_num_local_samples = torch.tensor(len(data_concat), device="cuda") + dist.all_reduce(max_num_local_samples, op=dist.ReduceOp.MAX) + if len(data_concat) < max_num_local_samples: + assert len(data_concat) + 1 == max_num_local_samples + dummy = torch.empty_like(data_concat[:1]) + data_concat = torch.cat([data_concat, dummy], dim=0) + dummy_count = torch.tensor(1, device="cuda") + else: + dummy_count = torch.tensor(0, device="cuda") + # Get all concatenated batches from all ranks and concatenate again. + dist.all_reduce(dummy_count, op=dist.ReduceOp.SUM) + data_concat = all_gather_tensor(data_concat.contiguous()) + data_collate = torch.stack(data_concat, dim=1).flatten(start_dim=0, end_dim=1) + # Remove the dummy samples. + if dummy_count > 0: + data_collate = data_collate[:-dummy_count] + elif isinstance(data_batches[0], collections.abc.Mapping): + data_collate = dict() + for key in data_batches[0].keys(): + data_collate[key] = collate_batches([data[key] for data in data_batches]) # type: ignore + else: + raise TypeError + return data_collate + + +@torch.no_grad() +def all_gather_tensor(tensor: torch.Tensor) -> list[torch.Tensor]: + """Gather the corresponding tensor from all GPU devices to a list. + + Args: + tensor (torch.Tensor): Pytorch tensor. + + Returns: + tensor_list (list[torch.Tensor]): A list of Pytorch tensors gathered from all GPU devices. + """ + tensor_list = [torch.zeros_like(tensor) for _ in range(get_world_size())] + dist.all_gather(tensor_list, tensor) + return tensor_list + + +def broadcast(tensor, src, group=None, async_op=False): + world_size = get_world_size() + if world_size < 2: + return tensor + dist.broadcast(tensor, src=src, group=group, async_op=async_op) + + +def dist_reduce_tensor(tensor, rank=0, reduce="mean"): + r"""Reduce to rank 0""" + world_size = get_world_size() + if world_size < 2: + return tensor + with torch.no_grad(): + dist.reduce(tensor, dst=rank) + if get_rank() == rank: + if reduce == "mean": + tensor /= world_size + elif reduce == "sum": + pass + else: + raise NotImplementedError + return tensor + + +def sync_model_states( + model: torch.nn.Module, + process_group: Optional[dist.ProcessGroup] = None, + src: int = 0, + params_and_buffers_to_ignore: Optional[Container[str]] = None, + broadcast_buffers: bool = True, +): + """ + Modify based on DDP source code + Synchronizes the parameters and buffers of a model across different processes in a distributed setting. + + This function ensures that all processes in the specified process group have the same initial parameters and + buffers from the source rank, typically rank 0. It is useful when different processes start with different model + states and a synchronization is required to ensure consistency across all ranks. + + Args: + model (nn.Module): The model whose parameters and buffers are to be synchronized. + process_group (dist.ProcessGroup, optional): The process group for communication. If None, + the default group is used. Defaults to None. + src (int, optional): The source rank from which parameters and buffers will be broadcasted. + Defaults to 0. + params_and_buffers_to_ignore (Optional[Container[str]], optional): A container of parameter and buffer + names to exclude from synchronization. Defaults to None, which means all parameters and buffers are + included. + broadcast_buffers (bool, optional): Whether to broadcast buffers or not. Defaults to True. + + Side Effects: + This function modifies the state of the model in-place to synchronize it with the source rank's model state. + + Raises: + RuntimeError: If the shapes of parameters across processes do not match, a runtime error will be raised. + + Examples: + >>> # downloading duplicated model weights from s3 in each rank and save network bandwidth + >>> # useful and save our time when model weights are huge + >>> if dist.get_rank == 0: + >>> model.load_state_dict(network_bound_weights_download_fn(s3_weights_path)) + >>> dist.barrir() + >>> sync_model_states(model) # sync rank0 weights to other ranks + """ + if not dist.is_available() or not dist.is_initialized(): + return + if process_group is None: + process_group = _get_default_group() + if not params_and_buffers_to_ignore: + params_and_buffers_to_ignore = set() + + log.info( + f"Synchronizing model states from rank {src} to all ranks in process group {get_process_group_ranks(process_group)}." + ) + + # Build tuple of (module, parameter) for all parameters that require grads. + modules_and_parameters = [ + (module, parameter) + for module_name, module in model.named_modules() + for parameter in [ + param + # Note that we access module.named_parameters instead of + # parameters(module). parameters(module) is only needed in the + # single-process multi device case, where it accesses replicated + # parameters through _former_parameters. + for param_name, param in module.named_parameters(recurse=False) + if f"{module_name}.{param_name}" not in params_and_buffers_to_ignore + # if param.requires_grad + # and f"{module_name}.{param_name}" not in params_and_buffers_to_ignore + ] + ] + + # Deduplicate any parameters that might be shared across child modules. + memo = set() + modules_and_parameters = [ + # "p not in memo" is the deduplication check. + # "not memo.add(p)" is always True, and it's only there to cause "add(p)" if needed. + (m, p) + for m, p in modules_and_parameters + if p not in memo and not memo.add(p) # type: ignore[func-returns-value] + ] + + # Build list of parameters. + parameters = [parameter for _, parameter in modules_and_parameters] + if len(parameters) == 0: + return + + _verify_param_shape_across_processes(process_group, parameters) + + _sync_module_states( + module=model, + process_group=process_group, + broadcast_bucket_size=int(250 * 1024 * 1024), + src=src, + params_and_buffers_to_ignore=params_and_buffers_to_ignore, + broadcast_buffers=broadcast_buffers, + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/ema.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/ema.py new file mode 100644 index 0000000000000000000000000000000000000000..9bb9b9c89e2b01cfd19df45286c681ba9c2c4245 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/ema.py @@ -0,0 +1,366 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +from contextlib import contextmanager, nullcontext +from typing import TYPE_CHECKING, Any, Generator, List, Optional, Union + +import numpy as np +import torch + +try: + from megatron.core import parallel_state + + USE_MEGATRON = True +except ImportError: + USE_MEGATRON = False + +from cosmos_policy._src.imaginaire.utils import distributed, log + +if TYPE_CHECKING: + from cosmos_policy._src.imaginaire.model import ImaginaireModel + + +class FastEmaModelUpdater: + """ + This class is used to update target model~(EMA) given source model~(regular model) and beta. + The method interaface mimic :class:`EMAModelTracker` and :class:`PowerEMATracker`. + Different from two classes, this class does not maintain the EMA model weights as buffers. It expects the user to have two module with same architecture and weights shape. + The class is proposed to work with FSDP model where above two classes are not working as expected. Besides, it is strange to claim model weights as buffers and do unnecessary name changing in :class:`EMAModelTracker` and :class:`PowerEMATracker`. Moeving forward, we should use this class instead of above two classes. + """ + + def __init__(self): + # Flag to indicate whether the cache is taken or not. Useful to avoid cache overwrite + self.is_cached = False + + def update_average(self, src_model: torch.nn.Module, tgt_model: torch.nn.Module, beta: float = 0.9999) -> None: + target_list = [] + source_list = [] + for tgt_params, src_params in zip(tgt_model.parameters(), src_model.parameters()): + assert tgt_params.dtype == torch.float32, ( + f"EMA model only works in FP32 dtype, got {tgt_params.dtype} instead." + ) + target_list.append(tgt_params) + source_list.append(src_params.data) + torch._foreach_mul_(target_list, beta) + torch._foreach_add_(target_list, source_list, alpha=1.0 - beta) + + def copy_to(self, src_model: torch.nn.Module, tgt_model: torch.nn.Module) -> None: + for tgt_params, src_params in zip(tgt_model.parameters(), src_model.parameters()): + tgt_params.data.copy_(src_params.data) + + def cache(self, parameters: Any, is_cpu: bool = False) -> None: + """Save the current parameters for restoring later. + + Args: + parameters (iterable): Iterable of torch.nn.Parameter to be temporarily stored. + """ + assert self.is_cached is False, "EMA cache is already taken. Did you forget to restore it?" + device = "cpu" if is_cpu else "cuda" + self.collected_params = [param.clone().to(device) for param in parameters] + self.is_cached = True + + def restore(self, parameters: Any) -> None: + """Restore the parameters in self.collected_params. + + Useful to validate the model with EMA parameters without affecting the + original optimization process. Store the parameters before copy_to(). + After validation (or model saving), use this to restore the former parameters. + + Args: + parameters (iterable): Iterable of torch.nn.Parameter to be updated with the stored parameters. + """ + assert self.is_cached, "EMA cache is not taken yet." + for c_param, param in zip(self.collected_params, parameters, strict=False): + param.data.copy_(c_param.data.type_as(param.data)) + self.collected_params = [] + # Release the cache after we call restore + self.is_cached = False + + +def get_buffer_name(param_name: str, torch_compile_buffer_renaming: bool = False) -> str: + """ + This function creates buffer name used by EMA from parameter's name + + Args: + param_name (str): Model's parameter name + Returns: + buffer_name (str): buffer name to be used for given parameter name + """ + + buffer_name = param_name.replace(".", "-") + + if torch_compile_buffer_renaming: + # torch.compile() adds _orig_mod to state dict names, this way we get original name + buffer_name = buffer_name.replace("_orig_mod-", "") + + return buffer_name + + +class EMAModelTracker(torch.nn.Module): + """This is a class to track the EMA model weights. + + The EMA weights are registered as buffers, which are extractable as state dicts. The names follow those of the + regular weights, except all "." are replaced with "-" (limitation of register_buffer()). This is similar to SDXL's + implementation of EMA. There are no optimizable parameters. + TODO(snah): multi-EMA weights. + + Attributes: + collected_params (list): temporarily stores the regular weights while in EMA mode. + beta (float): EMA decay rate. (default: 0.9999). + torch_compile_buffer_renaming (bool): whether to remove '_orig_mod-' from buffer names when torch.compile is used + """ + + def __init__(self, model: ImaginaireModel, beta: float = 0.9999, torch_compile_buffer_renaming: bool = False): + """Constructor of the EMA model weight tracker. + + Args: + model (ImaginaireModel): The PyTorch model. + beta (float): EMA decay rate. (default: 0.9999). + """ + super().__init__() + self.torch_compile_buffer_renaming: bool = torch_compile_buffer_renaming + if not 0.0 <= beta <= 1.0: + raise ValueError("Decay must be between 0 and 1") + self.beta = beta + for name, param in model.named_parameters(): + if param.requires_grad: + buffer_name = get_buffer_name(name, self.torch_compile_buffer_renaming) + self.register_buffer(buffer_name, param.clone().detach().data) + self.collected_params = [] + # Flag to indicate whether the cache is taken or not. Useful to avoid cache overwrite + self.is_cached = False + + @torch.no_grad() + def update_average(self, model: ImaginaireModel, iteration: Optional[int] = None) -> None: + del iteration + target_list = [] + source_list = [] + ema_buffers = self.state_dict() + for name, param in model.named_parameters(): + if param.requires_grad: + buffer_name = get_buffer_name(name, self.torch_compile_buffer_renaming) + buffer = ema_buffers[buffer_name] + assert buffer.dtype == torch.float32, f"EMA model only works in FP32 dtype, got {buffer.dtype} instead." + target_list.append(buffer) + source_list.append(param.data) + torch._foreach_mul_(target_list, self.beta) + torch._foreach_add_(target_list, source_list, alpha=1.0 - self.beta) + + def copy_to(self, model: ImaginaireModel) -> None: + ema_buffers = self.state_dict() + for name, param in model.named_parameters(): + if param.requires_grad: + buffer_name = get_buffer_name(name, self.torch_compile_buffer_renaming) + buffer = ema_buffers[buffer_name] + param.data.copy_(buffer.data) + + def cache(self, parameters: Any, is_cpu: bool = False) -> None: + """Save the current parameters for restoring later. + + Args: + parameters (iterable): Iterable of torch.nn.Parameter to be temporarily stored. + """ + assert self.is_cached is False, "EMA cache is already taken. Did you forget to restore it?" + device = "cpu" if is_cpu else "cuda" + self.collected_params = [param.clone().to(device) for param in parameters] + self.is_cached = True + + def restore(self, parameters: Any) -> None: + """Restore the parameters in self.collected_params. + + Useful to validate the model with EMA parameters without affecting the + original optimization process. Store the parameters before copy_to(). + After validation (or model saving), use this to restore the former parameters. + + Args: + parameters (iterable): Iterable of torch.nn.Parameter to be updated with the stored parameters. + """ + assert self.is_cached, "EMA cache is not taken yet." + for c_param, param in zip(self.collected_params, parameters, strict=False): + param.data.copy_(c_param.data.type_as(param.data)) + self.collected_params = [] + # Release the cache after we call restore + self.is_cached = False + + @classmethod + def initialize_multi_rank_ema( + cls, model: torch.nn.Module, rate: Union[float, List[float]], num: int = 1, enabled: bool = True + ) -> Optional[EMAModelTracker]: + """ + Class method to initialize per rank EMA Model Tracker with different rate. + Each rank will have a different rate based on the given configuration, resulting in different EMA weights. + + Args: + model (torch.nn.Module): The neural network model to be tracked. + rate (Union[float, List[float]]): The decay rate(s) for the EMA. If a list is provided, + it corresponds to rates for different ranks. + num (int, optional): The number of leading ranks to consider for different rates. + Defaults to 1. + enabled (bool, optional): Flag to enable or disable the creation of the tracker. + If False, returns None. Defaults to True. + + Returns: + Optional[EMAModelTracker]: An instance of EMAModelTracker if enabled, otherwise None. + + Example: + >>> model = torch.nn.Linear(10, 2) + >>> tracker = EMAModelTracker.initialize_ema_from_settings(model, rate=[0.1, 0.2], num=2) + >>> print(tracker) + + Notes: + If `rate` is a list and the current rank is less than `num`, the rate for the current rank + is used. If the current rank exceeds `num`, the first rate in the list is used by default. + """ + if not enabled: + return None + if USE_MEGATRON and parallel_state.is_initialized(): + cur_dp_rank = parallel_state.get_data_parallel_rank(with_context_parallel=True) + log.critical(f"using MCore parallel_state for EMA initialization. DP RANK: {cur_dp_rank}", rank0_only=False) + log.warning("It should not used together with FSDP!") + else: + cur_dp_rank = distributed.get_rank() + log.critical(f"using torch.distributed for EMA initialization. DP RANK: {cur_dp_rank}", rank0_only=False) + rate = rate if isinstance(rate, list) else [rate] + num = min(num, len(rate)) + rate = rate[cur_dp_rank] if cur_dp_rank < num else rate[0] + if cur_dp_rank < num: + print(f"EMAModelTracker: rank {cur_dp_rank}, rate {rate}") + return cls(model, rate) + + +class PowerEMATracker(EMAModelTracker): + def __init__(self, model: ImaginaireModel, s: float = 0.1, torch_compile_buffer_renaming: bool = False): + """Constructor of the EMA model weight tracker. + + Args: + model (ImaginaireModel): The PyTorch model. + s (float): EMA decay rate. See EDM2 paper + torch_compile_buffer_renaming (bool): whether to remove '_orig_mod-' from buffer names when torch.compile is used + """ + super().__init__(model=model, beta=0.0, torch_compile_buffer_renaming=torch_compile_buffer_renaming) + self.exp = np.roots([1, 7, 16 - s**-2, 12 - s**-2]).real.max() + + @torch.no_grad() + def update_average(self, model: ImaginaireModel, iteration: Optional[int] = None) -> None: + if iteration == 0: + beta = 0.0 + else: + i = iteration + 1 + beta = (1 - 1 / i) ** (self.exp + 1) + self.beta = beta + + super().update_average(model, iteration) + + @classmethod + def initialize_multi_rank_ema( + cls, model: torch.nn.Module, rate: float, num: int, enabled: bool = True + ) -> Optional[PowerEMATracker]: + """ + Class method to initialize per rank EMA Model Tracker with different rate. + Each rank will have a different rate based on the given configuration, resulting in different EMA weights. + + Args: + model (torch.nn.Module): The neural network model for which the EMA tracker is being set up. + num (int): The number of ranks for which the rate adjustment is applied. Beyond this, the rate remains unchanged. + rate (float): The base decay rate for the EMA calculation. + enabled (bool, optional): Flag to enable or disable the initialization of the tracker. If False, returns None. + Defaults to True. + + Returns: + Optional[PowerEMATracker]: An instance of PowerEMATracker with adjusted rate if enabled, otherwise None. + + Raises: + None + + Example: + >>> model = torch.nn.Linear(10, 2) + >>> tracker = PowerEMATracker.initialize_multi_rank_ema(model, num=3, rate=0.99) + >>> print(tracker) + + Notes: + The decay rate is modified by dividing it by 2 raised to the power of the rank for each rank less than `num`. + If the rank is greater than or equal to `num`, the base rate is used without modification. This approach + allows higher ranked processes to have a less aggressive decay, potentially reflecting their delayed synchronization + in a distributed training scenario. + """ + if not enabled: + return None + if USE_MEGATRON and parallel_state.is_initialized(): + cur_dp_rank = parallel_state.get_data_parallel_rank(with_context_parallel=True) + log.critical(f"using MCore parallel_state for EMA initialization. DP RANK: {cur_dp_rank}", rank0_only=False) + log.warning("It should not used together with FSDP!") + else: + cur_dp_rank = distributed.get_rank() + log.critical(f"using torch.distributed for EMA initialization. DP RANK: {cur_dp_rank}", rank0_only=False) + + divider = 2**cur_dp_rank if cur_dp_rank < num else 1 + if cur_dp_rank < num: + print(f"PowerEMATracker: rank {cur_dp_rank}, rate {rate / divider}") + return cls(model, rate / divider) + + +@contextmanager +def ema_scope(model: ImaginaireModel, enabled: bool = False, context: str | None = None) -> Generator[None, None, None]: + """Context manager for switching between regular and EMA model weights. + + This function is a dispatcher that handles two main cases: + 1. If the model has its own `ema_scope` method, it will be used. + This allows models to define custom EMA logic (e.g., for FSDP). + 2. If not, it falls back to a generic mechanism that expects the model + to have a `.ema` attribute containing an EMA tracker object. + + Args: + model (ImaginaireModel): The PyTorch model. + enabled (bool): Whether switching to EMA weights is enabled (default: False). + context (str | None): A logging context string, passed to the model's ema_scope if used. + """ + + def scope_function(): + if enabled: + has_custom_scope = hasattr(model, "ema_scope") and callable(model.ema_scope) + has_generic_ema = hasattr(model, "ema") and isinstance( + model.ema, (FastEmaModelUpdater, EMAModelTracker, PowerEMATracker) + ) + assert has_custom_scope or has_generic_ema + + if has_custom_scope: + return model.ema_scope(context=context) + else: + return ema_scope_generic(model) + else: + return nullcontext() + + with scope_function(): + yield + + +@contextmanager +def ema_scope_generic(model: ImaginaireModel) -> Generator[None, None, None]: + """Generic context manager for switching between regular and EMA model weights. + + Args: + model (ImaginaireModel): The PyTorch model, which must have a `.ema` attribute. + """ + model.ema.cache(model.parameters()) + model.ema.copy_to(model) + + log.info("EMA: switched to EMA weights.") + try: + yield + finally: + model.ema.restore(model.parameters()) + log.info("EMA: restored regular weights.") diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/ema_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/ema_test.py new file mode 100644 index 0000000000000000000000000000000000000000..1d7a2ddbe58018d8f92f2837512e7990b2692270 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/ema_test.py @@ -0,0 +1,160 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import pytest +import torch + +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils.ema import EMAModelTracker, PowerEMATracker, ema_scope + + +@pytest.fixture +def linear_model_instance(): + beta = 0.9 + # Fix the seed for model initialization + torch.manual_seed(0) + model = ImaginaireModel() + model.net = torch.nn.Linear(10, 2) + model.ema = EMAModelTracker(model, beta) + return model + + +@pytest.fixture +def linear_model_no_bias_instance(): + beta = 0.9 + # Fix the seed for model initialization + torch.manual_seed(0) + model = ImaginaireModel() + model.net = torch.nn.Linear(10, 2, bias=False) + model.ema = EMAModelTracker(model, beta) + return model + + +@pytest.mark.L0 +def test_val_error(linear_model_instance): + model = linear_model_instance + torch.manual_seed(0) + x_train = torch.rand((100, 10)) + y_train = torch.rand(100).round().long() + x_val = torch.rand((100, 10)) + y_val = torch.rand(100).round().long() + optimizer = torch.optim.Adam(model.parameters(), lr=1e-2) + + model.train() + for _ in range(2): + logits = model.net(x_train) + loss = torch.nn.functional.cross_entropy(logits, y_train) + optimizer.zero_grad() + loss.backward() + optimizer.step() + model.ema.update_average(model) + + model.eval() + logits = model.net(x_val) + loss_orig = torch.nn.functional.cross_entropy(logits, y_val) + print(f"Original loss: {loss_orig}") + + with ema_scope(model, True): + logits = model.net(x_val) + loss_ema = torch.nn.functional.cross_entropy(logits, y_val) + print(f"EMA loss: {loss_ema}") + assert loss_ema < loss_orig, "EMA loss was not lower" + + logits = model.net(x_val) + loss_orig2 = torch.nn.functional.cross_entropy(logits, y_val) + assert torch.allclose(loss_orig, loss_orig2), "Restored model was not the same as stored model" + + +@pytest.mark.L0 +@pytest.mark.parametrize("ema_tracker", [EMAModelTracker, PowerEMATracker]) +def test_ema_update(linear_model_no_bias_instance, ema_tracker): + model = linear_model_no_bias_instance + with torch.no_grad(): + model.net.weight.fill_(0.0) + + if ema_tracker == EMAModelTracker: + model.ema = ema_tracker(model, beta=0.9) + elif ema_tracker == PowerEMATracker: + model.ema = ema_tracker(model, s=0.1) + else: + raise ValueError(f"Unknown EMA tracker: {ema_tracker}. Please use EMAModelTracker or PowerEMATracker.") + + # Check that the ema weights were initialized correctly. + ema_weight = model.ema.state_dict()["net-weight"] + assert torch.all(ema_weight == 0.0), "EMA weights were not initialized correctly" + + with torch.no_grad(): + model.net.weight.fill_(1.0) + + # Iteration is used to compute beta in power ema, but is not used in regular ema. + model.ema.update_average(model, iteration=1) + + # Check that the ema weights were updated correctly. + ema_weight = model.ema.state_dict()["net-weight"] + assert torch.allclose(ema_weight, torch.full(size=(1,), fill_value=(1.0 - model.ema.beta))), ( + "EMA update was incorrect" + ) + + # Check that the regular model weights were not changed from the ema update. + assert torch.all(model.net.weight == 1.0), "EMA update shouldn't have changed the regular model weights" + + # Check that the ema weights were copied back to the model correctly. + model.ema.copy_to(model) + assert torch.allclose(model.net.weight, torch.full(size=(1,), fill_value=(1.0 - model.ema.beta))), ( + "EMA weights were copied to the model incorrectly" + ) + + +@pytest.mark.L1 +@pytest.mark.parametrize("ema_tracker", [EMAModelTracker, PowerEMATracker]) +def test_ema_update_torch_compile(linear_model_no_bias_instance, ema_tracker): + model = linear_model_no_bias_instance + with torch.no_grad(): + model.net.weight.fill_(0.0) + + if ema_tracker == EMAModelTracker: + model.ema = ema_tracker(model, beta=0.9, torch_compile_buffer_renaming=True) + elif ema_tracker == PowerEMATracker: + model.ema = ema_tracker(model, s=0.1, torch_compile_buffer_renaming=True) + else: + raise ValueError(f"Unknown EMA tracker: {ema_tracker}. Please use EMAModelTracker or PowerEMATracker.") + + # Compilation should take place after EMA was created + model.net = torch.compile(model.net) + + # Check that the ema weights were initialized correctly. + ema_weight = model.ema.state_dict()["net-weight"] + assert torch.all(ema_weight == 0.0), "EMA weights were not initialized correctly" + + with torch.no_grad(): + model.net.weight.fill_(1.0) + + # Iteration is used to compute beta in power ema, but is not used in regular ema. + model.ema.update_average(model, iteration=1) + + # Check that the ema weights were updated correctly. + ema_weight = model.ema.state_dict()["net-weight"] + assert torch.allclose(ema_weight, torch.full(size=(1,), fill_value=(1.0 - model.ema.beta))), ( + "EMA update was incorrect" + ) + + # Check that the regular model weights were not changed from the ema update. + assert torch.all(model.net.weight == 1.0), "EMA update shouldn't have changed the regular model weights" + + # Check that the ema weights were copied back to the model correctly. + model.ema.copy_to(model) + assert torch.allclose(model.net.weight, torch.full(size=(1,), fill_value=(1.0 - model.ema.beta))), ( + "EMA weights were copied to the model incorrectly" + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/embedding_concat_strategy.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/embedding_concat_strategy.py new file mode 100644 index 0000000000000000000000000000000000000000..612f45a9e96823a891e648143b6f4e6a648f3c70 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/embedding_concat_strategy.py @@ -0,0 +1,25 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from enum import Enum + + +class EmbeddingConcatStrategy(str, Enum): + FULL_CONCAT = "full_concat" # Concatenate embeddings all layers + MEAN_POOLING = "mean_pooling" # Average pool embeddings all layers + POOL_EVERY_N_LAYERS_AND_CONCAT = "pool_every_n_layers_and_concat" # Pool every n layers and concatenatenate + + def __str__(self) -> str: + return self.value diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/fsdp_helper.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/fsdp_helper.py new file mode 100644 index 0000000000000000000000000000000000000000..761b819320590155c1a97c496b1a73e21af49289 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/fsdp_helper.py @@ -0,0 +1,159 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +from contextlib import contextmanager +from functools import partial + +import torch +from torch.distributed.algorithms._checkpoint.checkpoint_wrapper import ( + CheckpointImpl, + apply_activation_checkpointing, + checkpoint_wrapper, +) +from torch.distributed.device_mesh import init_device_mesh +from torch.distributed.fsdp import FullyShardedDataParallel as FSDP +from torch.distributed.fsdp._runtime_utils import ( + _post_forward, + _post_forward_reshard, + _pre_forward, + _pre_forward_unshard, + _root_pre_forward, +) +from torch.distributed.utils import _p_assert + +from cosmos_policy._src.imaginaire.utils import distributed, log + + +def apply_fsdp_checkpointing(model, list_block_cls): + """apply activation checkpointing to model + returns None as model is updated directly + """ + log.critical("--> applying fdsp activation checkpointing...") + non_reentrant_wrapper = partial( + checkpoint_wrapper, + # offload_to_cpu=False, + checkpoint_impl=CheckpointImpl.NO_REENTRANT, + ) + + def check_fn(submodule): + result = False + for block_cls in list_block_cls: + if isinstance(submodule, block_cls): + result = True + break + return result + + apply_activation_checkpointing(model, checkpoint_wrapper_fn=non_reentrant_wrapper, check_fn=check_fn) + + +@contextmanager +def possible_fsdp_scope( + model: torch.nn.Module, +): + enabled = isinstance(model, FSDP) + if enabled: + assert not torch.is_grad_enabled(), "FSDP context should be entered with grad disabled" + handle = model._handle + args, kwargs = [0], dict(dummy=0) + with torch.autograd.profiler.record_function("FullyShardedDataParallel.possible_fsdp_scope"): + args, kwargs = _root_pre_forward(model, model, args, kwargs) + unused = None + args, kwargs = _pre_forward( + model, + handle, + _pre_forward_unshard, + model._fsdp_wrapped_module, + args, + kwargs, + ) + if handle: + _p_assert( + handle.flat_param.device == model.compute_device, + "Expected `FlatParameter` to be on the compute device " + f"{model.compute_device} but got {handle.flat_param.device}", + ) + try: + yield None + finally: + if enabled: + output = {"output": 1} + _post_forward(model, handle, _post_forward_reshard, model, unused, output) + + +def hsdp_device_mesh(replica_group_size=None, sharding_group_size=None, device=None): + """ + Initializes a device mesh for use with Hybrid Sharding strategy in FSDP (HSDP) training. + + This function requires explicit sizes for replica and sharding groups to accommodate models + whose GPU fit is unknown, providing flexibility in distributed training setups. + + Args: + replica_group_size (int): The size of each replica group. Must be provided to ensure + the model fits within the available resources. + sharding_group_size (int): The size of each sharding group that the model can fit. Must be provided to + ensure the correct distribution of model parameters. + device (str, optional): The device to use (e.g., "cuda:0"). If None, defaults to "cuda" + with the local rank as the device index. + + Returns: + A device mesh object compatible with FSDP. + + Raises: + ValueError: If replica_group_size or sharding_group_size are not provided, or if the + world size is not evenly divisible by the sharding group size. + RuntimeError: If a valid device mesh cannot be created. + + Usage: + If your model fits on 4 GPUS, and you have 3 nodes of 8 GPUs, then: + Sharding_Group_Size = 4 + Replica_Groups_Size = (24 total gpus, 4 per sharding group) = 6 Replica Groups + >>> device_mesh = initialize_device_mesh(replica_group_size, sharding_group_size) + >>> sharded_model = FSDP(model, device_mesh=device_mesh, ...) + """ + + # world_size = int(os.getenv("WORLD_SIZE", "1")) + world_size = distributed.get_world_size() + if sharding_group_size is None: + sharding_group_size = min(world_size, 8) + sharding_group_size = min(sharding_group_size, world_size) + if replica_group_size is None: + replica_group_size = world_size // sharding_group_size + + device = device or "cuda" + + if world_size % sharding_group_size != 0: + raise ValueError( + f"World size {world_size} is not evenly divisible by sharding group size {sharding_group_size}." + ) + + if (world_size // sharding_group_size) % replica_group_size != 0: + raise ValueError( + f"The calculated number of replica groups is not evenly divisible by " + f"replica_group_size {replica_group_size}." + ) + + device_mesh = init_device_mesh( + device, (replica_group_size, sharding_group_size), mesh_dim_names=("replicate", "shard") + ) + if device_mesh is None: + raise RuntimeError("Failed to create a valid device mesh.") + + log.critical( + f"Device mesh initialized with replica group size {replica_group_size} and sharding group size {sharding_group_size}" + ) + + return device_mesh diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/fused_adam.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/fused_adam.py new file mode 100644 index 0000000000000000000000000000000000000000..f98b33ccd44fd7e1e9cfd378e6e7f4dc5df3ddfd --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/fused_adam.py @@ -0,0 +1,383 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import torch +import transformer_engine as te +import transformer_engine_torch as tex + +from cosmos_policy._src.imaginaire.utils import distributed, log + + +class FusedAdam(torch.optim.Optimizer): + """Implements Adam algorithm. + + Currently GPU-only. Requires Apex to be installed via + ``pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./``. + + This version of fused Adam implements 2 fusions. + + * Fusion of the Adam update's elementwise operations + * A multi-tensor apply launch that batches the elementwise updates applied to all the model's parameters + into one or a few kernel launches. + + :class:`FusedAdam` may be used as a drop-in replacement for ``torch.optim.AdamW``, + or ``torch.optim.Adam`` with ``adam_w_mode=False``:: + + opt = FusedAdam(model.parameters(), lr = ....) + ... + opt.step() + + .. warning:: + A previous version of :class:`FusedAdam` allowed a number of additional arguments to ``step``. + These additional arguments are now deprecated and unnecessary. + + Adam was been proposed in `Adam: A Method for Stochastic Optimization`_. + + Arguments: + params (iterable): iterable of parameters to optimize or dicts defining + parameter groups. + lr (float, optional): learning rate. (default: 1e-3) + betas (Tuple[float, float], optional): coefficients used for computing + running averages of gradient and its square. (default: (0.9, 0.999)) + eps (float, optional): term added to the denominator to improve + numerical stability. (default: 1e-8) + weight_decay (float, optional): weight decay (L2 penalty) (default: 0) + amsgrad (boolean, optional): whether to use the AMSGrad variant of this + algorithm from the paper `On the Convergence of Adam and Beyond`_ + (default: False) NOT SUPPORTED in FusedAdam! + adam_w_mode (boolean, optional): Apply L2 regularization or weight decay + True for decoupled weight decay(also known as AdamW) (default: True) + capturable (bool, optional): whether to use the version of the optimizer + that can be used with CUDA Graphs. (default: False) + master_weights (bool, optional): whether to maintain FP32 master weights + in the optimizer with FP16 mixed precision training, currently can + only be used with capturable set to True. (default: False) + + .. _Adam - A Method for Stochastic Optimization: + https://arxiv.org/abs/1412.6980 + .. _On the Convergence of Adam and Beyond: + https://openreview.net/forum?id=ryQu7f-RZ + """ + + def __init__( + self, + params, + lr=1e-3, + bias_correction=True, + betas=(0.9, 0.999), + eps=1e-8, + adam_w_mode=True, + weight_decay=0.0, + amsgrad=False, + capturable=False, + master_weights=False, + ): + if amsgrad: + raise RuntimeError("FusedAdam does not support the AMSGrad variant.") + if master_weights and not capturable: + raise RuntimeError("Master weights is currently only supported with the capturable version.") + # If the optimizer is capturable then LR should be a tensor (on GPU) + log.warning(f"FusedAdam master_weights: {master_weights} capturable: {capturable}") + lr = torch.tensor(lr, dtype=torch.float32) if capturable else lr + defaults = dict(lr=lr, bias_correction=bias_correction, betas=betas, eps=eps, weight_decay=weight_decay) + super(FusedAdam, self).__init__(params, defaults) + self.adam_w_mode = 1 if adam_w_mode else 0 + + self.capturable = capturable + self.master_weights = master_weights + + self.param_groups_master = None + + if capturable: + for idx, group in enumerate(self.param_groups): + if len(group["params"]) == 0: + continue + device = group["params"][0].device + for item in ["lr"]: + if isinstance(group[item], float): + group[item] = torch.tensor(group[item], dtype=torch.float32) + self.param_groups[idx][item] = group[item].to(device=device) + + self._step_supports_amp_scaling = True + + # Skip buffer + self._dummy_overflow_buf = torch.tensor([0], dtype=torch.int, device="cuda") + self.multi_tensor_adam = tex.multi_tensor_adam + self.multi_tensor_adam_capturable = tex.multi_tensor_adam_capturable + self.multi_tensor_adam_capturable_master = tex.multi_tensor_adam_capturable_master + + def step(self, closure=None, grads=None, output_params=None, scale=None, grad_norms=None, grad_scaler=None): + """Performs a single optimization step. + + Arguments: + closure (callable, optional): A closure that reevaluates the model + and returns the loss. + + The remaining arguments are deprecated, and are only retained (for the moment) for error-checking purposes. + """ + if any(p is not None for p in [grads, output_params, scale, grad_norms]): + raise RuntimeError( + "FusedAdam has been updated. " + "Simply initialize it identically to torch.optim.Adam, and call step() with no arguments." + ) + loss = None + if closure is not None: + loss = closure() + + if self.param_groups_master is None: + # Create full precision master weights + self.param_groups_master = [] + for i, pg in enumerate(self.param_groups): + param_list = pg["params"] + self.param_groups_master.append( + { + "params": [p.clone().detach().float() if self.master_weights else None for p in param_list], + } + ) + + for group, group_master in zip(self.param_groups, self.param_groups_master): + if len(group["params"]) == 0: + continue + device = group["params"][0].device + bias_correction = 1 if "bias_correction" in group and group["bias_correction"] else 0 + beta1, beta2 = group["betas"] + + # assume same step across group now to simplify things + # per parameter step can be easily support by making it tensor, or pass list into kernel + if "step" in group: + if self.capturable: + group["step"] = ( + group["step"].to(device=device) + if isinstance(group["step"], torch.Tensor) + else torch.tensor(group["step"], dtype=torch.int32, device=device) + ) + group["step"] += (self._dummy_overflow_buf != 1).to(torch.int) + else: + group["step"] += 1 + else: + group["step"] = 1 if not self.capturable else torch.tensor([1], dtype=torch.int, device=device) + + if self.capturable: + group["lr"] = ( + group["lr"].to(device=device) + if isinstance(group["lr"], torch.Tensor) + else torch.tensor(group["lr"], dtype=torch.float32, device=device) + ) + + # create lists for multi-tensor apply + g_16, p_16, m_16, v_16 = [], [], [], [] + g_bf, p_bf, m_bf, v_bf = [], [], [], [] + g_32, p_32, m_32, v_32 = [], [], [], [] + p_16_master = [] + p_32_master = [] + bf16_master = [] + + for p, p_master in zip(group["params"], group_master["params"]): + if p.grad is None: + continue + if p.grad.data.is_sparse: + raise RuntimeError( + "FusedAdam does not support sparse gradients, please consider SparseAdam instead" + ) + + state = self.state[p] + # State initialization + if len(state) == 0: + # Exponential moving average of gradient values + state["exp_avg"] = torch.zeros_like(p.data).float() + # Exponential moving average of squared gradient values + state["exp_avg_sq"] = torch.zeros_like(p.data).float() + + if p.dtype == torch.float16: + if self.master_weights: + p_16_master.append(p_master.data) + g_16.append(p.grad.data) + p_16.append(p.data) + m_16.append(state["exp_avg"]) + v_16.append(state["exp_avg_sq"]) + elif p.dtype == torch.bfloat16: + if self.master_weights: + bf16_master.append(p_master.data) + g_bf.append(p.grad) + p_bf.append(p) + m_bf.append(state["exp_avg"]) + v_bf.append(state["exp_avg_sq"]) + elif p.dtype == torch.float32: + if self.master_weights: + p_32_master.append(p_master.data) + g_32.append(p.grad.data) + p_32.append(p.data) + m_32.append(state["exp_avg"]) + v_32.append(state["exp_avg_sq"]) + else: + raise RuntimeError("FusedAdam only support fp16 and fp32.") + + # If the optimizer is capturable, then if there's a grad scaler it works + # on the GPU + a different multi_tensor_applier should be called + if self.capturable: + # overflow check of gradients + found_inf = ( + grad_scaler._check_inf_per_device(self)[device] + if grad_scaler is not None + else torch.zeros((1,), device=device) + ) + self._dummy_overflow_buf.copy_(found_inf) + + # get unscale scale factor + scale, inv_scale = None, None + if grad_scaler: + scale = grad_scaler._get_scale_async() + inv_scale = scale.double().reciprocal().float() + else: + scale = torch.ones((1,), device=device, dtype=torch.float32) + inv_scale = torch.ones((1,), device=device, dtype=torch.float32) + + if len(g_16) > 0: + te.pytorch.optimizers.multi_tensor_applier( + ( + self.multi_tensor_adam_capturable_master + if self.master_weights + else self.multi_tensor_adam_capturable + ), + self._dummy_overflow_buf, + [g_16, p_16, m_16, v_16, p_16_master] if self.master_weights else [g_16, p_16, m_16, v_16], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + inv_scale, + ) + + if len(g_bf) > 0: + te.pytorch.optimizers.multi_tensor_applier( + ( + self.multi_tensor_adam_capturable_master + if self.master_weights + else self.multi_tensor_adam_capturable + ), + self._dummy_overflow_buf, + [g_bf, p_bf, m_bf, v_bf, bf16_master] if self.master_weights else [g_bf, p_bf, m_bf, v_bf], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + inv_scale, + ) + + if len(g_32) > 0: + te.pytorch.optimizers.multi_tensor_applier( + ( + self.multi_tensor_adam_capturable_master + if self.master_weights + else self.multi_tensor_adam_capturable + ), + self._dummy_overflow_buf, + [g_32, p_32, m_32, v_32, p_32_master] if self.master_weights else [g_32, p_32, m_32, v_32], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + inv_scale, + ) + else: + if len(g_16) > 0: + te.pytorch.optimizers.multi_tensor_applier( + self.multi_tensor_adam, + self._dummy_overflow_buf, + [g_16, p_16, m_16, v_16], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + ) + + if len(g_bf) > 0: + te.pytorch.optimizers.multi_tensor_applier( + self.multi_tensor_adam, + self._dummy_overflow_buf, + [g_bf, p_bf, m_bf, v_bf], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + ) + + if len(g_32) > 0: + te.pytorch.optimizers.multi_tensor_applier( + self.multi_tensor_adam, + self._dummy_overflow_buf, + [g_32, p_32, m_32, v_32], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + ) + + return loss + + def load_state_dict(self, state_dict): + super().load_state_dict(state_dict) + for group in self.param_groups: + if self.capturable: + group["lr"] = ( + group["lr"].cuda() + if isinstance(group["lr"], torch.Tensor) + else torch.tensor(group["lr"], dtype=torch.float32).cuda() + ) + + if "step" in group: + if self.capturable: + if distributed.get_rank() == 0: + step = ( + group["step"].cuda() + if isinstance(group["step"], torch.Tensor) + else torch.tensor([group["step"]], dtype=torch.int32).cuda() + ) + else: + step = torch.zeros(1, dtype=torch.int32).cuda() + # make it compatible with FSDP optimizer + distributed.broadcast(step, 0) + group["step"] = step + elif isinstance(group["step"], torch.Tensor): + group["step"] = group["step"].item() + for p in group["params"]: + state = self.state[p] + if "exp_avg" in state: + state["exp_avg"] = state["exp_avg"].float() + state["exp_avg_sq"] = state["exp_avg_sq"].float() diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/fused_nan_to_num.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/fused_nan_to_num.py new file mode 100644 index 0000000000000000000000000000000000000000..520a83e1da1b6eefec7f72b0c27ba2359db45bfb --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/fused_nan_to_num.py @@ -0,0 +1,24 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import List + +import torch + + +@torch.jit.script +def fused_nan_to_num(params: List[torch.Tensor]): + for param in params: + torch.nan_to_num(param, nan=0.0, posinf=0.0, neginf=0.0, out=param) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/graph.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/graph.py new file mode 100644 index 0000000000000000000000000000000000000000..8e7b4e27fc65d1a8a03ff0c83eb12af9ebe53326 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/graph.py @@ -0,0 +1,444 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""A rework of make_graphed_callabled function from TransformerEngine so that it works with inference-only.""" + +from typing import Any, Callable, Dict, Optional, Tuple, TypeVar, Union + +import torch +from torch._C import _graph_pool_handle +from torch.utils._pytree import tree_flatten as _tree_flatten +from torch.utils._pytree import tree_unflatten as _tree_unflatten +from transformer_engine.pytorch.distributed import get_all_rng_states, graph_safe_rng_available +from transformer_engine.pytorch.module.base import TransformerEngineBaseModule + +from cosmos_policy._src.imaginaire.utils import log + +__all__ = ["create_cuda_graph"] + + +_IS_GRAPH_CAPTURING = False + +_T = TypeVar("_T") +SingleOrTuple = Union[_T, Tuple[_T, ...]] + + +def set_capture_start() -> None: + """Record beginning of `make_graphed_callables`.""" + global _IS_GRAPH_CAPTURING + _IS_GRAPH_CAPTURING = True + + +def set_capture_end() -> None: + """Record end of `make_graphed_callables`.""" + global _IS_GRAPH_CAPTURING + _IS_GRAPH_CAPTURING = False + + +def is_graph_capturing() -> None: + """Return whether within `make_graphed_callables`.""" + return _IS_GRAPH_CAPTURING + + +def graph_pool_handle(): + """ + Returns an opaque token representing the id of a graph memory pool. + """ + return _graph_pool_handle() + + +def _make_graphed_callables( + callables: SingleOrTuple[Callable], + sample_args: SingleOrTuple[Tuple[torch.Tensor, ...]], + num_warmup_iters: int = 3, + sample_kwargs: Optional[SingleOrTuple[Dict[str, Any]]] = None, + pool: Optional[Tuple[int, ...]] = None, +) -> SingleOrTuple[Callable]: + """ + Helper method for `make_graphed_callables` + """ + + if torch.is_autocast_enabled() and torch.is_autocast_cache_enabled(): + raise RuntimeError( + "make_graphed_callables does not support the autocast caching. Please set `cache_enabled=False`." + ) + + # Default is to pass no kwargs to callables + if sample_kwargs is None: + if isinstance(callables, tuple): + sample_kwargs = tuple({} for _ in range(len(sample_args))) + else: + sample_kwargs = {} + + # Canonicalize args as tuples + just_one_callable = False + if not isinstance(callables, tuple): + just_one_callable = True + callables = (callables,) + sample_args = (sample_args,) + sample_kwargs = (sample_kwargs,) + + # Check sizes of args + assert len(sample_args) == len(callables) + assert len(sample_kwargs) == len(callables) + + # Check callables + for c in callables: + if isinstance(c, torch.nn.Module): + assert len(c._backward_hooks) == 0 and len(c._forward_hooks) == 0 and len(c._forward_pre_hooks) == 0, ( + "Modules must not have hooks registered at the time they are passed. " + + "However, registering hooks on modules after passing them " + + "through make_graphed_callables is allowed." + ) + assert all(b.requires_grad is False for b in c.buffers()), ( + "In any :class:`~torch.nn.Module` passed to " + + ":func:`~make_graphed_callables`, only parameters may be trainable. " + + "All buffers must have ``requires_grad=False``." + ) + + # Flatten callable arguments + per_callable_kwargs_keys = [list(kwargs.keys()) for kwargs in sample_kwargs] + flatten_sample_args = [] + for args, kwargs, kwargs_keys in zip(sample_args, sample_kwargs, per_callable_kwargs_keys): + flatten_arg, _ = _tree_flatten(args) + flatten_kwarg, _ = _tree_flatten([kwargs[key] for key in kwargs_keys]) + flatten_sample_args.append(tuple(flatten_arg + flatten_kwarg)) + assert all(isinstance(arg, torch.Tensor) for arg in flatten_arg), ( + "In the beta API, sample_args " + + "for each callable must contain only Tensors. Other types are not allowed." + ) + + # If a callable is an nn.Module, its graph's full input surface is the args the user explicitly + # passes to forward (ie, its sample_args) AND the module's parameter attributes. + per_callable_len_user_args = [len(args) for args in flatten_sample_args] + per_callable_module_params = [tuple(c.parameters()) if isinstance(c, torch.nn.Module) else () for c in callables] + per_callable_static_input_surfaces = [ + flatten_sample_args[i] + per_callable_module_params[i] for i in range(len(callables)) + ] + + fwd_graphs = [torch.cuda.CUDAGraph() for _ in range(len(flatten_sample_args))] + graph_callables = [None for _ in range(len(flatten_sample_args))] + + # For cases with multiple active RNG states, e.g. TP. + if graph_safe_rng_available(): + for _, state in get_all_rng_states().items(): + for fwd_graph in fwd_graphs: + fwd_graph.register_generator_state(state) + + mempool = graph_pool_handle() if pool is None else pool + + # Warmup + # Hopefully prevents cudnn benchmarking and other lazy-initialization cuda work + # from ending up in any captures. + torch.cuda.synchronize() + + # Get warmup func and func_idx. + warmup_func_idx = [] + warmup_func = [] + for func_idx, func in enumerate(callables): + warmup_func_idx.append(func_idx) + warmup_func.append(func) + assert len(warmup_func) == len(sample_args), f"Warmup runs {len(warmup_func)} don't match args {len(sample_args)}." + assert len(warmup_func_idx) == len(set(warmup_func_idx)), ( + f"Warmup runs {len(warmup_func)} but only {len(set(warmup_func_idx))} are unique." + ) + + # Filter the TE modules that cudagraph can access. + visited_te_modules = set() + + def hook_fn(module, inputs, outputs): # pylint: disable=unused-argument + if isinstance(module, TransformerEngineBaseModule): + visited_te_modules.add(module) + + # Run warmup and do the above filtering. + with torch.cuda.stream(torch.cuda.Stream()): + for func_idx, func in zip(warmup_func_idx, warmup_func): + args = sample_args[func_idx] + kwargs = sample_kwargs[func_idx] + for _ in range(num_warmup_iters): + hooks = [] + for module in func.modules(): + hook = module.register_forward_hook(hook_fn) + hooks.append(hook) + outputs, _ = _tree_flatten(func(*args, **kwargs)) + for hook in hooks: + hook.remove() + del outputs + # The following code is added specifically for MCore's special requirements, + # aimed at preventing warmup from altering the control flow. + for module in func.modules(): + if hasattr(module, "is_first_microbatch"): + module.is_first_microbatch = True + torch.cuda.synchronize() + + # All captures here share a mempool. To avoid replays corrupting each other's memory, + # the safest approach is to capture all passes in the same order they'll run: + # Capture forward graphs + per_callable_static_outputs = [] + per_callable_output_unflatten_spec = [] + graph_id = 0 + for func, args, kwargs, fwd_graph in zip(callables, sample_args, sample_kwargs, fwd_graphs): + with torch.cuda.graph(fwd_graph, pool=mempool): + outputs = func(*args, **kwargs) + graph_callables[graph_id] = func + graph_id += 1 + + flatten_outputs, spec = _tree_flatten(outputs) + per_callable_static_outputs.append(tuple(flatten_outputs)) + per_callable_output_unflatten_spec.append(spec) + + def make_graphed_autograd_function( + fwd_graph, + module_params, + kwargs_keys, + len_user_args, + output_unflatten_spec, + static_input_surface, + static_outputs, + ): + class Graphed(torch.autograd.Function): + """Autograd function for graph replay.""" + + @staticmethod + def forward(ctx, *inputs): + # pylint: disable=missing-function-docstring + + # Copy values from new tensors into static tensors + for i in range(len_user_args): + if static_input_surface[i].data_ptr() != inputs[i].data_ptr(): + static_input_surface[i].copy_(inputs[i]) + + # Replay forward graph + fwd_graph.replay() + assert isinstance(static_outputs, tuple) + return tuple(o.detach() for o in static_outputs) + + def functionalized(*user_args, **user_kwargs): + # Check that required kwargs are provided + for key in kwargs_keys: + if key not in user_kwargs: + raise TypeError( + f"Graphed callable was initialized with kwarg {key} ,but it was not provided in graph replay" + ) + + # Runs the autograd function with inputs == all inputs to + # the graph that might require grad (explicit user args + + # module parameters) + # Assumes module params didn't change since capture. + flatten_user_args, _ = _tree_flatten(user_args) + flatten_user_kwargs, _ = _tree_flatten([user_kwargs[key] for key in kwargs_keys]) + func_args = tuple(flatten_user_args) + tuple(flatten_user_kwargs) + module_params + out = Graphed.apply(*func_args) + return _tree_unflatten(out, output_unflatten_spec) + + return functionalized + + # Put together the final graphed callables + ret = [] + for i in range(len(sample_args)): + graphed = make_graphed_autograd_function( + fwd_graphs[i], + per_callable_module_params[i], + per_callable_kwargs_keys[i], + per_callable_len_user_args[i], + per_callable_output_unflatten_spec[i], + per_callable_static_input_surfaces[i], + per_callable_static_outputs[i], + ) + + func = graph_callables[i] + if isinstance(func, torch.nn.Module): + + def make_graphed_forward(func, graph_training_state, graphed, orig_fwd): + def new_fwd(*user_args, **user_kwargs): + # If the module's training-or-eval state matches what we graphed, + # run the graph, otherwise run the original forward method + if func.training == graph_training_state: + return graphed(*user_args, **user_kwargs) + return orig_fwd(*user_args, **user_kwargs) + + return new_fwd + + forward = make_graphed_forward(func, func.training, graphed, func.forward) + ret.append(forward) + else: + ret.append(graphed) + + if just_one_callable: + return ret[0] + + return tuple(ret) + + +def make_graphed_callables_forward( + modules: SingleOrTuple[Callable], + sample_args: SingleOrTuple[Tuple[torch.Tensor, ...]], + num_warmup_iters: int = 3, + sample_kwargs: Optional[SingleOrTuple[Dict[str, Any]]] = None, + pool: Optional[Tuple[int, ...]] = None, +) -> Union[Callable, Tuple[Callable, ...]]: + """ + Make CUDA graph version of Transformer Engine modules + A variation of PyTorch's `make_graphed_callables` utility function. + `original PyTorch implementation `_ + for more documentation. + Graphing parameters + ------------------- + modules: (tuple of) callable + Callable or callables to graph. + sample_args: (tuple of) tuple of torch.Tensor + Positional arguments to callable(s). + num_warmup_iters: int, default = 3 + Number of warmup iterations. + sample_kwargs: (tuple of) dict, optional + Keyword arguments to callable(s) + pool: (tuple of) int, default = `None`, optional + An instance returned from function `torch.cuda.graph_pool_handle` that hints + this graph may share memory with the indicated pool. + """ + set_capture_start() + + # Handle single module. + just_one_callable = False + if not isinstance(modules, tuple): + just_one_callable = True + modules = (modules,) + + forward_funcs = [] + for module in modules: + assert isinstance(module, torch.nn.Module), f"Graphing for {type(module)} is not supported." + forward_funcs.append(module) + + if just_one_callable: + forward_funcs = forward_funcs[0] + else: + forward_funcs = tuple(forward_funcs) + + # Save RNG state. + if graph_safe_rng_available(): + generators = [ + torch.cuda.default_generators[torch.cuda.current_device()], + *get_all_rng_states().values(), + ] + original_rng_states = [state.get_state() for state in generators] + else: + original_rng_states = torch.cuda.get_rng_state() + + graphed_callables = _make_graphed_callables( + forward_funcs, + sample_args, + num_warmup_iters=num_warmup_iters, + sample_kwargs=sample_kwargs, + pool=pool, + ) + + # Ensures warmup does not affect numerics for ops such as dropout. + if graph_safe_rng_available(): + for gen, state in zip(generators, original_rng_states): + gen.set_state(state) + else: + torch.cuda.set_rng_state(original_rng_states) + set_capture_end() + return graphed_callables + + +def create_cuda_graph( + cuda_graphs_storage: dict, + blocks: torch.nn.ModuleList, + tensor_args: list[Any], + tensor_kwargs: dict[str, Any], + extra_key: Optional[str] = None, +) -> str: + def _make_dummy_tensor_like(t: torch.Tensor) -> torch.Tensor: + if t.dtype.is_floating_point: + return torch.randn(t.shape, device=t.device, dtype=t.dtype) + if t.dtype == torch.bool: + return torch.zeros(t.shape, device=t.device, dtype=t.dtype) + if t.dtype in (torch.uint8, torch.int8, torch.int16, torch.int32, torch.int64): + if t.numel() > 0: + low = int(t.min().item()) + high = int(t.max().item()) + if high == low: + high = low + 1 + else: + high = high + 1 + else: + low, high = 0, 1 + return torch.randint(low, high, t.shape, device=t.device, dtype=t.dtype) + # Fallback: use zeros for uncommon dtypes (e.g., complex) to avoid dtype/range pitfalls. + return torch.zeros(t.shape, device=t.device, dtype=t.dtype) + + def _make_dummy_tree(x: Any) -> Any: + flat, spec = _tree_flatten(x) + dummy_flat: list[torch.Tensor] = [] + for leaf in flat: + if not isinstance(leaf, torch.Tensor): + raise TypeError( + f"create_cuda_graph only supports pytrees of torch.Tensor leaves; got leaf type {type(leaf)}" + ) + dummy = _make_dummy_tensor_like(leaf) + dummy.requires_grad = leaf.requires_grad + dummy_flat.append(dummy) + return _tree_unflatten(dummy_flat, spec) + + real_args = [arg for arg in tensor_args if arg is not None] + real_kwargs = {k: v for k, v in tensor_kwargs.items() if v is not None} + + # Shapes key must reflect all tensor leaves (supports tuple/list/dict structures). + flat_tensors: list[torch.Tensor] = [] + for arg in real_args: + flat, _ = _tree_flatten(arg) + for leaf in flat: + if not isinstance(leaf, torch.Tensor): + raise TypeError( + f"create_cuda_graph only supports pytrees of torch.Tensor leaves; got leaf type {type(leaf)}" + ) + flat_tensors.append(leaf) + for _, kwarg in real_kwargs.items(): + flat, _ = _tree_flatten(kwarg) + for leaf in flat: + if not isinstance(leaf, torch.Tensor): + raise TypeError( + f"create_cuda_graph only supports pytrees of torch.Tensor leaves; got leaf type {type(leaf)}" + ) + flat_tensors.append(leaf) + + shapes_key = "_".join(str(shape_component) for t in flat_tensors for shape_component in t.shape) + if extra_key: + shapes_key = f"{shapes_key}_{extra_key}" + if shapes_key not in cuda_graphs_storage: + callables = [] + sample_args = [] + sample_kwargs = [] + for block in blocks: + callables.append(block) + args = [] + kwargs = {} + for arg in real_args: + args.append(_make_dummy_tree(arg)) + for name, kwarg in real_kwargs.items(): + kwargs[name] = _make_dummy_tree(kwarg) + sample_args.append(tuple(args)) + sample_kwargs.append(kwargs) + + log.critical(f"Creating graph for shape {shapes_key}") + cuda_graphs_storage[shapes_key] = make_graphed_callables_forward( + tuple(callables), + tuple(sample_args), + sample_kwargs=tuple(sample_kwargs), + num_warmup_iters=11, + ) + log.critical(f"Created graph for shape {shapes_key}") + return shapes_key diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/helper_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/helper_test.py new file mode 100644 index 0000000000000000000000000000000000000000..ea96266504f873b89b05b76d14fb70b15be28997 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/helper_test.py @@ -0,0 +1,168 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""Adapted from: + +https://github.com/PyTorchLightning/pytorch-lightning/blob/master/tests/helpers/runif.py +""" + +import importlib.metadata +import subprocess +import sys +from pathlib import Path +from typing import Any, Dict, List, Optional, Union + +import pytest +import torch +from loguru import logger +from packaging.version import Version +from pytest import MarkDecorator + +from cosmos_policy._src.imaginaire.utils.device import get_gpu_architecture + + +class RunIf: + """RunIf wrapper for conditional skipping of tests. + + Fully compatible with `@pytest.mark`. + + Example: + + ```python + @RunIf(min_torch="1.8") + @pytest.mark.parametrize("arg1", [1.0, 2.0]) + def test_wrapper(arg1): + assert arg1 > 0 + ``` + """ + + def __new__( + cls, + min_gpus: int = 0, + min_torch: Optional[str] = None, + max_torch: Optional[str] = None, + min_python: Optional[str] = None, + supported_arch: Optional[List[str]] = None, + requires_file: Optional[Union[str, List[str]]] = None, + requires_package: Optional[Union[str, List[str]]] = None, + **kwargs: Dict[Any, Any], + ) -> MarkDecorator: + """Creates a new `@RunIf` `MarkDecorator` decorator. + + :param min_gpus: Min number of GPUs required to run test. + :param min_torch: Minimum pytorch version to run test. + :param max_torch: Maximum pytorch version to run test. + :param min_python: Minimum python version required to run test. + :param requires_file: File or list of files required to run test. + :param requires_package: Package name or list of package names required to be installed to run test. + :param kwargs: Native `pytest.mark.skipif` keyword arguments. + """ + conditions = [] + reasons = [] + + if min_gpus: + conditions.append(torch.cuda.device_count() < min_gpus) + reasons.append(f"GPUs>={min_gpus}") + + if min_torch: + torch_version = importlib.metadata.version("torch") + conditions.append(Version(torch_version) < Version(min_torch)) + reasons.append(f"torch>={min_torch}") + + if max_torch: + torch_version = importlib.metadata.version("torch") + conditions.append(Version(torch_version) >= Version(max_torch)) + reasons.append(f"torch<{max_torch}") + + if min_python: + py_version = f"{sys.version_info.major}.{sys.version_info.minor}.{sys.version_info.micro}" + conditions.append(Version(py_version) < Version(min_python)) + reasons.append(f"python>={min_python}") + + if supported_arch: + if isinstance(supported_arch, str): + supported_arch = [supported_arch] + gpu_arch = get_gpu_architecture() + conditions.extend([gpu_arch not in supported_arch]) + reasons.append(f"supported_arch arch={','.join(supported_arch)}") + + if requires_file: + if isinstance(requires_file, str): + requires_file = [requires_file] + conditions.extend([not Path(file).exists() for file in requires_file]) + reasons.append(f"requires file={','.join(requires_file)}") + + if requires_package: + if isinstance(requires_package, str): + requires_package = [requires_package] + for package in requires_package: + try: + __import__(package) + except ImportError: + conditions.extend([True]) + reasons.append(f"Package {package} is not installed.") + + reasons = [rs for cond, rs in zip(conditions, reasons) if cond] + return pytest.mark.skipif( + condition=any(conditions), + reason=f"Requires: [{' + '.join(reasons)}]", + **kwargs, + ) + + +def run_command( + cmd: str, max_retry_counter: int = 3, is_raise: bool = True, capture_output: bool = True +) -> subprocess.CompletedProcess: + """Runs a shell command with the ability to retry upon failure. + + Parameters: + - cmd (str): The shell command to run. + - max_retry_counter (int): Maximum number of retries if the command fails. + - is_raise (bool): Whether to raise an exception and exit the program if the command fails after all retries. + - capture_output (bool): Whether to capture the output of the command. + + Returns: + - subprocess.CompletedProcess: The result of the command execution. + """ + + retry_counter = 0 + while retry_counter < max_retry_counter: + try: + result = subprocess.run(cmd, shell=True, capture_output=capture_output, text=True, check=False) + + # Check if the command was successful (returncode = 0) + if result.returncode == 0: + return result + + retry_counter += 1 + logger.debug( + f"Retry {retry_counter}/{max_retry_counter}: Command '{cmd}' failed with error" + f" code {result.returncode}. Error message: {result.stderr.strip()}" + ) + + except Exception as e: # pylint: disable=broad-except + retry_counter += 1 + logger.debug(f"Retry {retry_counter}/{max_retry_counter}: Command '{cmd}' raised an exception: {e}") + + # If reached here, all retries have failed + error_message = ( + f"Command '{cmd}' failed after {max_retry_counter} retries. Error code: {result.returncode}. " + f"Error message: {result.stderr.strip()}" + ) + if is_raise: + raise RuntimeError(error_message) + + logger.critical(error_message) + return result diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/high_sigma_strategy.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/high_sigma_strategy.py new file mode 100644 index 0000000000000000000000000000000000000000..2e3f5ccd3fc388c92586c96c13436988925b4d4f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/high_sigma_strategy.py @@ -0,0 +1,28 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from enum import Enum + + +class HighSigmaStrategy(str, Enum): + NONE = "none" + UNIFORM80_2000 = "uniform80_2000" + LOGUNIFORM200_100000 = "LOGUNIFORM200_100000" + SHIFT24 = "shift24" + BALANCED_TWO_HEADS_V1 = "balanced_two_heads_v1" + HARDCODED_20steps = "hardcoded_20steps" + + def __str__(self) -> str: + return self.value diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/launch.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/launch.py new file mode 100644 index 0000000000000000000000000000000000000000..5fd4dc83da2dc9f3397e9903d210b78b9443d23b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/launch.py @@ -0,0 +1,176 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import argparse +import os +import sys +import time + +import torch +import wandb +from omegaconf import OmegaConf + +from cosmos_policy._src.imaginaire.config import Config +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.cluster_env import get_cluster_env +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.imaginaire.utils.env_parsers.cred_env_parser import CRED_ENVS + +# Global variable to track S3 readiness +S3_READY = False + + +def log_reproducible_setup(config: Config, args: argparse.Namespace) -> None: + """ + Configures the environment for reproducibility of experiments by setting up + S3 backends for storage, logging important job details, and saving configuration and + environment details both locally and on S3. + This function is crucial for ensuring that all aspects of the computational environment are captured and can be + replicated for future runs or analysis. + + Parameters: + config (Config): A configuration object containing all the settings necessary + for the job, including paths and credentials. + args (argparse.Namespace): An argparse namespace containing the command line + arguments passed to the script. This includes configurations + and any overrides specified at runtime. + + Actions: + - Sets up S3 backend for storing user data and other outputs. + - Logs job paths and critical information regarding job execution. + - Saves the job configuration locally only for the main node in a distributed setting. + - Captures and logs command-line execution details. + - Optionally reads git commit and branch information if available and logs them. + - Saves both job environment information and launch details locally and syncs these to S3. + - Supports conditional integration with Weights & Biases (wandb) for experiment tracking. + + Notes: + - The function is designed to run within a distributed environment where certain actions + (like saving configurations) are restricted to the main node (rank 0). + - It uses the 'easy_io' module for interacting with S3, ensuring files are written and + read correctly from the object store. + - It leverages OmegaConf for saving YAML configurations + - git information is read from 'git_commit.txt' and 'git_branch.txt' files if they exist. + - snapshot codebase is saved as 'codebase.zip' if it exists in the current directory. + + Raises: + FileNotFoundError: If specific files like 'git_commit.txt' or 'codebase.zip' are expected + but not found. + IOError: If there are issues in file handling operations, particularly with file + reading/writing. + """ + + run_timestamp = f"{time.strftime('%Y-%m-%d_%H-%M-%S')}" + time_tensor = torch.ByteTensor(bytearray(run_timestamp, "utf-8")).cuda() + distributed.broadcast(time_tensor, 0) + run_timestamp = time_tensor.cpu().numpy().tobytes().decode("utf-8") + + global S3_READY + if os.path.exists(config.checkpoint.save_to_object_store.credentials) or CRED_ENVS.APP_ENV in [ + "prod", + "dev", + "stg", + ]: + easy_io.set_s3_backend( + backend_args={ + "backend": "s3", + "path_mapping": { + "s3://timestamps_rundir/": f"s3://{config.checkpoint.save_to_object_store.bucket}/{config.job.path}/job_runs/{run_timestamp}/", + "s3://rundir/": f"s3://{config.checkpoint.save_to_object_store.bucket}/{config.job.path}/", + }, + "s3_credential_path": config.checkpoint.save_to_object_store.credentials, + } + ) + S3_READY = True + else: + log.warning("S3 credentials not found. Skipping easy_io S3 setup.") + + log.warning(f"Job path: {config.job.path}") + job_info = get_cluster_env() + # save cfg to local + if distributed.get_rank() == 0: + job_local_path = config.job.path_local + log.critical(f"Job local path: {job_local_path}") + os.makedirs(config.job.path_local, exist_ok=True) + launch_info = { + "cmd": " ".join(sys.argv), + "args_cfg_path": args.config, + "args_override": args.opts, + } + + job_info["job_local_path"] = str(job_local_path) + job_info["s3"] = f"s3://{config.checkpoint.save_to_object_store.bucket}/{config.job.path}/" + # optional read git_commit.txt and save git commit id + if os.path.exists("git_commit.txt"): + with open("git_commit.txt", "r") as f: + job_info["commit_id"] = f.read().strip() + log.critical(f"Commit id: {job_info['commit_id']}") + if os.path.exists("git_branch.txt"): + with open("git_branch.txt", "r") as f: + job_info["git_branch"] = f.read().strip() + log.critical(f"git branch: {job_info['git_branch']}") + + with open(f"{job_local_path}/job_env.yaml", "w") as f: + OmegaConf.save(job_info, f) + with open(f"{job_local_path}/launch_info.yaml", "w") as f: + OmegaConf.save(launch_info, f) + if wandb.run: + wandb.run.config.update({f"JOB_INFO/{k}": v for k, v in job_info.items()}, allow_val_change=True) + + # by default, we upload run in ngc and slurm + if config.upload_reproducible_setup: + # sync to s3 + if S3_READY: + log.critical( + f"Uploading reproducible setup to s3://{config.checkpoint.save_to_object_store.bucket}/{config.job.path}/job_runs/{run_timestamp}/" + ) + + config_pkl_save_fp = f"{config.job.path_local}/config.pkl" + easy_io.copyfile_from_local( + config_pkl_save_fp, f"s3://timestamps_rundir/{config_pkl_save_fp.split('/')[-1]}" + ) + config_yaml_save_fp = config_pkl_save_fp.replace(".pkl", ".yaml") + easy_io.copyfile_from_local( + config_yaml_save_fp, f"s3://timestamps_rundir/{config_yaml_save_fp.split('/')[-1]}" + ) + easy_io.copyfile_from_local(f"{job_local_path}/job_env.yaml", "s3://timestamps_rundir/job_env.yaml") + easy_io.copyfile_from_local( + f"{job_local_path}/launch_info.yaml", + "s3://timestamps_rundir/launch_info.yaml", + ) + if os.path.exists("codebase.zip"): + easy_io.copyfile_from_local("codebase.zip", "s3://timestamps_rundir/codebase.zip") + if os.path.exists("code.tar.gz"): + easy_io.copyfile_from_local("code.tar.gz", "s3://timestamps_rundir/code.tar.gz") + if os.path.exists("git_diff.txt"): + easy_io.copyfile_from_local("git_diff.txt", "s3://timestamps_rundir/git_diff.txt") + if easy_io.exists("s3://rundir/job_history.yaml"): + job_history = easy_io.load("s3://rundir/job_history.yaml") + else: + job_history = {} + job_history[len(job_history)] = { + "timestamp": run_timestamp, + "reproduce_dir": f"s3://{config.checkpoint.save_to_object_store.bucket}/{config.job.path}/job_runs/{run_timestamp}/", + **launch_info, + } + print(job_history) + easy_io.dump(job_history, "s3://rundir/job_history.yaml") + else: + log.warning("S3 credentials not found. Skipping upload of reproducible setup.") + + # save per rank cluster information to s3 + if config.upload_reproducible_setup: + if S3_READY: + easy_io.dump(job_info, f"s3://timestamps_rundir/cluster_env/RANK_{distributed.get_rank():06d}.yaml") diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/log.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/log.py new file mode 100644 index 0000000000000000000000000000000000000000..561057321969b823d12f23ec53264dc5b5203109 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/log.py @@ -0,0 +1,162 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import atexit +import os +import sys +from typing import Any, Optional + +import torch.distributed as dist +from loguru._logger import Core, Logger + +RANK0_ONLY = True +LEVEL = os.environ.get("LOGURU_LEVEL", "INFO") + + +def make_new_logger(depth: int = 1) -> Logger: + return Logger( + core=Core(), + exception=None, + depth=depth, + record=False, + lazy=False, + colors=False, + raw=False, + capture=True, + patchers=[], + extra={}, + ) + + +logger = make_new_logger(depth=1) +atexit.register(logger.remove) + + +def _add_relative_path(record: dict[str, Any]) -> None: + start = os.getcwd() + record["extra"]["relative_path"] = os.path.relpath(record["file"].path, start) + + +*options, _, extra = logger._options # type: ignore +logger._options = tuple([*options, [_add_relative_path], extra]) # type: ignore + + +def init_loguru_stdout() -> None: + logger.remove() + datetime_format = get_datetime_format() + machine_format = get_machine_format() + message_format = get_message_format() + logger.add( + sys.stdout, + level=LEVEL, + format=f"{datetime_format}{machine_format}{message_format}", + filter=_rank0_only_filter, + ) + + +def init_loguru_file(path: str) -> None: + datetime_format = get_datetime_format() + machine_format = get_machine_format() + message_format = get_message_format() + logger.add( + path, + encoding="utf8", + level=LEVEL, + format=f"{datetime_format}{machine_format}{message_format}", + rotation="100 MB", + filter=lambda result: _rank0_only_filter(result) or not RANK0_ONLY, + enqueue=True, + ) + + +def get_datetime_format() -> str: + return "[{time:MM-DD HH:mm:ss}|" + + +def get_machine_format() -> str: + node_id = os.environ.get("NGC_ARRAY_INDEX", "0") + num_nodes = int(os.environ.get("NGC_ARRAY_SIZE", "1")) + machine_format = "" + rank = 0 + if dist.is_available(): + if not RANK0_ONLY and dist.is_initialized(): + rank = dist.get_rank() + world_size = dist.get_world_size() + machine_format = ( + f"[Node{node_id:<3}/{num_nodes:<3}][RANK{rank:<5}/{world_size:<5}]" + "[{process.name:<8}]| " + ) + return machine_format + + +def get_message_format() -> str: + message_format = "{level}|{extra[relative_path]}:{line}:{function}] {message}" + return message_format + + +def _rank0_only_filter(record: Any) -> bool: + is_rank0 = record["extra"].get("rank0_only", True) + if _get_rank() == 0 and is_rank0: + return True + if not is_rank0: + record["message"] = f"[RANK {_get_rank()}] " + record["message"] + return not is_rank0 + + +def trace(message: str, rank0_only: bool = True) -> None: + logger.opt(depth=1).bind(rank0_only=rank0_only).trace(message) + + +def debug(message: str, rank0_only: bool = True) -> None: + logger.opt(depth=1).bind(rank0_only=rank0_only).debug(message) + + +def info(message: str, rank0_only: bool = True) -> None: + logger.opt(depth=1).bind(rank0_only=rank0_only).info(message) + + +def success(message: str, rank0_only: bool = True) -> None: + logger.opt(depth=1).bind(rank0_only=rank0_only).success(message) + + +def warning(message: str, rank0_only: bool = True) -> None: + logger.opt(depth=1).bind(rank0_only=rank0_only).warning(message) + + +def error(message: str, rank0_only: bool = True) -> None: + logger.opt(depth=1).bind(rank0_only=rank0_only).error(message) + + +def critical(message: str, rank0_only: bool = True) -> None: + logger.opt(depth=1).bind(rank0_only=rank0_only).critical(message) + + +def exception(message: str, rank0_only: bool = True) -> None: + logger.opt(depth=1).bind(rank0_only=rank0_only).exception(message) + + +def _get_rank(group: Optional[dist.ProcessGroup] = None) -> int: + """Get the rank (GPU device) of the worker. + + Returns: + rank (int): The rank of the worker. + """ + rank = 0 + if dist.is_available() and dist.is_initialized(): + rank = dist.get_rank(group) + return rank + + +# Execute at import time. +init_loguru_stdout() diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/misc.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/misc.py new file mode 100644 index 0000000000000000000000000000000000000000..404086fae9f71442c94e5e8bac13c23b3144a442 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/misc.py @@ -0,0 +1,651 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import collections +import collections.abc +import functools +import json +import os +import random +from contextlib import ContextDecorator, nullcontext +from dataclasses import fields +from typing import Any, Callable, List, Tuple, TypeVar, Union + +import numpy as np +from loguru import logger as logging + +try: + # pyrefly: ignore # import-error + import straggler +except ImportError: + straggler = None +import termcolor +import torch +import wandb +from torch.distributed._functional_collectives import AsyncCollectiveTensor +from torch.distributed._tensor.api import DTensor + +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.distributed import all_gather_tensor +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.imaginaire.utils.timer import Timer + + +def requires_grad(model: torch.nn.Module, value: bool = True) -> None: + """Set a model to require gradients or not. + + Args: + model (torch.nn.Module): Neural network model. + value (bool): Whether the network requires gradients or not. + """ + for p in model.parameters(): + p.requires_grad = value + + +def to( + data: Any, + device: str | torch.device | None = None, + dtype: torch.dtype | None = None, + memory_format: torch.memory_format = torch.preserve_format, +) -> Any: + """Recursively cast data into the specified device, dtype, and/or memory_format. + + The input data can be a tensor, a list of tensors, a dict of tensors. + See the documentation for torch.Tensor.to() for details. + + Args: + data (Any): Input data. + device (str | torch.device): GPU device (default: None). + dtype (torch.dtype): data type (default: None). + memory_format (torch.memory_format): memory organization format (default: torch.preserve_format). + + Returns: + data (Any): Data cast to the specified device, dtype, and/or memory_format. + """ + assert device is not None or dtype is not None or memory_format is not None, ( + "at least one of device, dtype, memory_format should be specified" + ) + + if isinstance(data, torch.Tensor): + if ( + memory_format == torch.channels_last + and data.dim() != 4 + or memory_format == torch.channels_last_3d + and data.dim() != 5 + ): + memory_format = torch.preserve_format # do not change the memory format + is_cpu = (isinstance(device, str) and device == "cpu") or ( + isinstance(device, torch.device) and device.type == "cpu" + ) + data = data.to( + device=device, + dtype=dtype, + memory_format=memory_format, + non_blocking=(not is_cpu), + ) + return data + elif isinstance(data, collections.abc.Mapping): + return type(data)({key: to(data[key], device=device, dtype=dtype, memory_format=memory_format) for key in data}) + elif isinstance(data, collections.abc.Sequence) and not isinstance(data, (str, bytes)): + return type(data)([to(elem, device=device, dtype=dtype, memory_format=memory_format) for elem in data]) + else: + return data + + +def serialize(data: Any) -> Any: + """Serialize data by hierarchically traversing through iterables. + + Args: + data (Any): Input data. + + Returns: + data (Any): Serialized data. + """ + if isinstance(data, collections.abc.Mapping): + return type(data)({key: serialize(data[key]) for key in data}) + elif isinstance(data, collections.abc.Sequence) and not isinstance(data, (str, bytes)): + return type(data)([serialize(elem) for elem in data]) + else: + try: + json.dumps(data) + except TypeError: + data = str(data) + return data + + +def print_environ_variables(env_vars: list[str]) -> None: + """Print a specific list of environment variables. + + Args: + env_vars (list[str]): List of specified environment variables. + """ + for env_var in env_vars: + if env_var in os.environ: + log.info(f"Environment variable {Color.green(env_var)}: {Color.yellow(os.environ[env_var])}") + else: + log.warning(f"Environment variable {Color.green(env_var)} not set!") + + +def set_random_seed(seed: int, by_rank: bool = False) -> None: + """Set random seed. This includes random, numpy, Pytorch. + + Args: + seed (int): Random seed. + by_rank (bool): if true, each GPU will use a different random seed. + """ + if by_rank: + seed += distributed.get_rank() + log.info(f"Using random seed {seed}.") + random.seed(seed) + np.random.seed(seed) + torch.manual_seed(seed) # sets seed on the current CPU & all GPUs + + +def arch_invariant_rand( + shape: List[int] | Tuple[int], dtype: torch.dtype, device: str | torch.device, seed: int | None = None +): + """Produce a GPU-architecture-invariant randomized Torch tensor. + + Args: + shape (list or tuple of ints): Output tensor shape. + dtype (torch.dtype): Output tensor type. + device (torch.device): Device holding the output. + seed (int): Optional randomization seed. + + Returns: + tensor (torch.tensor): Randomly-generated tensor. + """ + # Create a random number generator, optionally seeded + rng = np.random.RandomState(seed) + + # Generate random numbers using the generator + random_array = rng.standard_normal(shape).astype(np.float32) # Use standard_normal for normal distribution + + # Convert to torch tensor and return + return torch.from_numpy(random_array).to(dtype=dtype, device=device) + + +def get_data_batch_size(data: dict[str, torch.Tensor] | torch.Tensor) -> int: + """Get the batch size from a data batch, a (possibly hierarchical) dictionary of tensors. + + Args: + data (dict[str, torch.Tensor]): Data batch (dictionary of tensors). + + Returns: + batch_size (int): Data batch size. + """ + + def _get_batch_size(input_data: Any) -> Union[int, None]: + """ + Helper function that recursively finds a tensor in the input data + (could be a nested dictionary) and returns its batch size. + """ + if isinstance(input_data, torch.Tensor): + return len(input_data) + elif isinstance(input_data, collections.abc.Mapping): + for key, value in input_data.items(): + batch_size = _get_batch_size(value) + if batch_size is not None: + return batch_size + return None + + batch_size = _get_batch_size(data) + if not isinstance(batch_size, int): + raise ValueError(f"Batch size ({batch_size}) obtained from invalid data: {data}") + return batch_size + + +def parameters_to_buffer(module: torch.nn.Module, persistent: bool = True): + """Convert parameters in a module to buffers. + Buffers do not have its own gradients and thus not updated by backpropagation. + + Args: + module (torch.nn.Module): a module to convert parameters + persistent (bool): If True, buffers are included in state_dict. + """ + named_params = dict() + + for name, param in module.named_parameters(): + named_params[name] = param + + for name, param in named_params.items(): + module_hierarchy = name.split(".") + submodule_name = ".".join(module_hierarchy[:-1]) + submodule = module.get_submodule(submodule_name) + subname = module_hierarchy[-1] + delattr(submodule, subname) + submodule.register_buffer(subname, param, persistent=persistent) + + return + + +T = TypeVar("T", bound=Callable[..., Any]) + + +class timer(Timer): + """Simple CPU timer for timing the execution of code. + + It can be used as either a context manager or a function decorator. The timing result will be logged upon exit. + + Example: + def func_a(): + time.sleep(1) + with timer("func_a"): + func_a() + + @timer("func_b) + def func_b(): + time.sleep(1) + func_b() + """ + + def __init__(self, context: str, debug: bool = False): + super().__init__( + tag=context, + measure_cpu=True, + measure_cuda=False, + unit="s", + debug=debug, + ) + + +class memory_checker(ContextDecorator): # noqa: N801 + """Simple memory checker for a given block of code. + + It can be used as either a context manager or a function decorator. The memory usage will be logged upon exit. + Example: + def func_a(): + torch.rand([int(1024**2)]).float().cuda() + with memory_checker("func_a"): + func_a() + >>> 0.004GB memory used + + @memory_checker("func_b") + def func_b(): + random_var = torch.rand([int(1024**2)]).cuda() + func_b() + """ + + def __init__(self, context: str, debug: bool = False): + self.context = context + self.debug = debug + + def __enter__(self) -> None: + torch.cuda.synchronize() + torch.cuda.reset_peak_memory_stats() + self.initial_memory = torch.cuda.max_memory_allocated() + + def __exit__(self, exc_type, exc_value, traceback) -> None: # noqa: ANN001 + torch.cuda.synchronize() + final_memory = torch.cuda.max_memory_allocated() + message = f"Memory used within {self.context}: {(final_memory - self.initial_memory) / 1024**3:.4f} GB" + if self.debug: + log.debug(message) + else: + log.info(message) + + def __call__(self, func: T) -> T: + @functools.wraps(func) + def wrapper(*args, **kwargs): # noqa: ANN202 + torch.cuda.synchronize() + torch.cuda.reset_peak_memory_stats() + initial_memory = torch.cuda.max_memory_allocated() + result = func(*args, **kwargs) + torch.cuda.synchronize() + final_memory = torch.cuda.max_memory_allocated() + message = f"Memory used within {self.context}: {(final_memory - initial_memory) / 1024**3:.4f} GB" + if self.debug: + log.debug(message) + else: + log.info(message) + return result + + return wrapper # type: ignore + + +class TrainingTimer: + """Timer for timing the execution of code, aggregating over multiple training iterations. + + It is used as a context manager to measure the execution time of code and store the timing results + for each function. The context managers can be nested. + + Attributes: + results (dict): A dictionary to store timing results for various code. + + Example: + timer = Timer() + for i in range(100): + with timer("func_a"): + func_a() + avg_time = sum(timer.results["func_a"]) / len(timer.results["func_a"]) + print(f"func_a() took {avg_time} seconds.") + """ + + def __init__(self) -> None: + self.results = dict() + self.average_results = dict() + self.timers = [] + self.func_stack = [] + self.reset() + + def reset(self) -> None: + self.results = {key: [] for key in self.results} + + def __enter__(self) -> TrainingTimer: + timer = Timer(measure_cpu=True, measure_cuda=False, debug=True, unit="s") + self.timers.append(timer) + timer.start() + return self + + def __exit__(self, exc_type, exc_value, traceback) -> None: # noqa: ANN001 + timer = self.timers.pop() + timer.end() + result = timer.get_cpu_time() + key = self.func_stack.pop() + self.results.setdefault(key, []) + self.results[key].append(result) + + def __call__(self, func_name: str) -> TrainingTimer: + self.func_stack.append(func_name) + return self + + def __getattr__(self, func_name: str) -> TrainingTimer: + return self.__call__(func_name) + + def nested(self, func_name: str) -> TrainingTimer: + return self.__call__(func_name) + + def compute_average_results(self) -> dict[str, float]: + results = dict() + for key, value_list in self.results.items(): + results[key] = sum(value_list) / len(value_list) + return results + + +def timeout_handler(timeout_period: float, signum: int, frame: int) -> None: + # What to do when the process gets stuck. For now, we simply end the process. + error_message = f"Timeout error: more than {timeout_period} seconds passed since the last iteration." + if distributed.is_rank0(): + wandb.alert(title="Timeout error!", text=error_message, level=wandb.AlertLevel.ERROR) + raise TimeoutError(error_message) + + +class Color: + """A convenience class to colorize strings in the console. + + Example: + import + print("This is {Color.red('important')}.") + """ + + @staticmethod + def red(x: str) -> str: + return termcolor.colored(str(x), color="red") + + @staticmethod + def green(x: str) -> str: + return termcolor.colored(str(x), color="green") + + @staticmethod + def blue(x: str) -> str: + return termcolor.colored(str(x), color="blue") + + @staticmethod + def cyan(x: str) -> str: + return termcolor.colored(str(x), color="cyan") + + @staticmethod + def yellow(x: str) -> str: + return termcolor.colored(str(x), color="yellow") + + @staticmethod + def magenta(x: str) -> str: + return termcolor.colored(str(x), color="magenta") + + @staticmethod + def grey(x: str) -> str: + return termcolor.colored(str(x), color="grey") + + +class BufferCnt: + """ + Buffer counter which keeps track of the condition when called and returns True when the condition in met "thres" + amount of times, otherwise returns False. + + Example usage: + buf = BufferCnt(thres=3) + for _ in range(5): + if buf(random.random() > 0.5): + print("We got lucky 3 times out of 5.") + + Args: + thres (int): The amount of times the expression needs to be True before returning True. + reset_over_thres (bool): Whether to reset the buffer after returning True. + """ + + def __init__(self, thres=10, reset_over_thres=False): + self._cnt = 0 + self.thres = thres + self.reset_over_thres = reset_over_thres + + def __call__(self, expre, thres=None): + if expre is True: + self._cnt += 1 + else: + self._cnt = 0 + + if thres is None: + thres = self.thres + + if self._cnt >= thres: + if self.reset_over_thres: + self.reset() + return True + + return False + + @property + def cnt(self): + return self._cnt + + def reset(self): + self._cnt = 0 + + +def dataclass_instance_to_dict(dataclass: Any) -> dict: + """Convert a dataclass to a dictionary. + + Args: + dataclass (Any): Dataclass object. + + Returns: + dict: Dictionary representation of the dataclass. + """ + return {f.name: getattr(dataclass, f.name) for f in fields(dataclass)} + + +def get_local_tensor_if_DTensor(tensor: torch.Tensor | DTensor) -> torch.tensor: + if isinstance(tensor, DTensor): + local = tensor.to_local() + # As per PyTorch documentation, if the communication is not finished yet, we need to wait for it to finish + # https://pytorch.org/docs/stable/distributed.tensor.html#torch.distributed.tensor.DTensor.to_local + if isinstance(local, AsyncCollectiveTensor): + return local.wait() + else: + return local + return tensor + + +class NVTXRangeContext: + """ + Context manager which inserts NVTX range around the current context and optionally calls torch.cuda.synchronize + at the start and the end of the context. + + Args: + name (str): Name of the NVTX range. + enabled (bool): Whether the context manager is enabled. When disabled, it does nothing. Default: True. + synchronize (bool): Whether to call torch.cuda.synchronize() at the start and the end of the context. Default: True. + """ + + def __init__(self, name: str, enabled: bool = True, synchronize: bool = True): + self.name = name + self.enabled = enabled + self.synchronize = synchronize + + def __enter__(self): + if not self.enabled: + return + if self.synchronize: + torch.cuda.synchronize() + torch.cuda.nvtx.range_push(self.name) + + def __exit__(self, exc_type, exc_val, exc_tb): + if not self.enabled: + return + if self.synchronize: + torch.cuda.synchronize() + torch.cuda.nvtx.range_pop() + + +class StragglerDetectorV2: + """StragglerDetectorV2 is a class that allows you to easily integrate "straggler" tool: + https://gitlab-master.nvidia.com/dl/gwe/fault_tolerance_related/straggler/-/tree/cupti?ref_type=heads. + + This tool detects stragglers using low-level CUPTI tool, which can gather kernel execution time with very low overhead. + The execution times are compared across different ranks, as well as to the execution time of the exact same kernels in the past. + This tool can be easily integrated, as it's resilient to any synchronizations, since it captures kernels execution time. + It means that we can wrap the entire forward or backward passes and the stragglers will be identified regardless + of synchronizations happening during the iteration. + + Args: + enabled (bool): Whether the straggler detection is enabled. When disabled, it does nothing. Default: True. + report_freq (int): Generate a report each report_freq iterations that analyzes the GPUs performance. Defaults to 100. + profile_freq (int): Enable the CUPTI profiling each profile_freq iterations. Since the overhead is very low, + the default value is 1. + max_diff (float): Defines the maximum relative difference between the fastest and the slowest rank to determine the slowdown. Defaults to 2.0 + raise_error (bool): Whether to raise error when stragglers are detected enough times. Defaults to True.""" + + def __init__( + self, + enabled: bool = True, + report_freq: int = 100, + profile_freq: int = 1, + max_diff: float = 2.0, + raise_error: bool = True, + ): + self.enabled = enabled + self.report_freq = report_freq + self.profile_freq = profile_freq + self.name = self.__class__.__name__ + self.slowdown_count = BufferCnt(thres=10, reset_over_thres=True) + self.max_diff = max_diff + self.raise_error = raise_error + + def initialize(self): + if self.enabled: + if not straggler: + raise RuntimeError( + "Please install straggler package before using StragglerDetectionV2." + "Package can be installed from here: https://gitlab-master.nvidia.com/dl/osiris/straggler" + ) + + straggler.Detector.initialize( + scores_to_compute=["relative_perf_scores", "individual_perf_scores"], + gather_on_rank0=False, # all ranks results will be available on rank 0 + profiling_interval=self.profile_freq, + ) + + def profile_section(self, name: str, section_enabled: bool, profile_cuda: bool = True): + if section_enabled and self.enabled: + return straggler.Detector.detection_section(name, profile_cuda=profile_cuda) + else: + return nullcontext() + + def _aggregate_section_results(self, local_section_summaries): + data = [] + for key in local_section_summaries: + # straggler reports time in ms + data.append(local_section_summaries[key][straggler.Statistic.MAX] / 1000) + return distributed.all_gather_tensor(torch.tensor(data).cuda()) + + def generate_report(self, iteration): + if self.enabled and iteration % self.report_freq == 0: + report = straggler.Detector.generate_report() + gpu_relative_perf_score = report.gpu_relative_perf_scores[distributed.get_rank()] + gpu_relative_perf_score_gather_list = distributed.all_gather_tensor( + torch.tensor([gpu_relative_perf_score]).cuda() + ) + local_section_data = self._aggregate_section_results(report.local_section_summaries) + if distributed.get_rank() == 0: + stragglers = report.identify_stragglers(gpu_rel_threshold=1 / self.max_diff) + wandb_info = { + f"{self.name}/relative_gpu_perf_{rank}": perf[0].item() + for rank, perf in enumerate(gpu_relative_perf_score_gather_list) + } + for key_id, key in enumerate(report.local_section_summaries): + wandb_info.update( + {f"{self.name}/{key}_{rank:03d}": v[key_id].item() for rank, v in enumerate(local_section_data)} + ) + + data_tensor = torch.tensor(gpu_relative_perf_score_gather_list) + slowest_rank_id = torch.argmin(data_tensor) + wandb_info.update( + { + f"slowest_rank/{self.name}_rank": slowest_rank_id.item(), + f"slowest_rank/{self.name}_relative_perf": torch.min(data_tensor).item(), + } + ) + + for key_id, key in enumerate(report.local_section_summaries): + data_tensor = torch.tensor([v[key_id] for v in local_section_data]) + wandb_info.update( + { + f"slowest_rank/slowest_{key}_rank": torch.argmax(data_tensor).item(), + f"slowest_rank/slowest_{key}_time": torch.max(data_tensor).item(), + } + ) + if wandb.run: + wandb.log(wandb_info, step=iteration) + + import cosmos_policy._src.imaginaire.utils.launch + + if cosmos_policy._src.imaginaire.utils.launch.S3_READY and (iteration % (5 * self.report_freq) == 0): + easy_io.dump( + wandb_info, + f"s3://rundir/{self.__class__.__name__}/iter_{iteration:09d}.yaml", + ) + easy_io.dump( + report, + f"s3://rundir/{self.__class__.__name__}/report_iter_{iteration:09d}.pkl", + ) + + # Which GPUs are slower than other GPUs, based on the execution time of kernels + relative_stragglers = stragglers["straggler_gpus_relative"] + # Which GPUs are slower than itself in the past, based on the past execution time of kernels. + individual_stragglers = stragglers["straggler_gpus_individual"] + is_slowdown = relative_stragglers or individual_stragglers + if is_slowdown: + hostname = torch.ByteTensor(bytearray(os.uname().nodename, "utf-8")).cuda() + whole_hostname = all_gather_tensor(hostname) + slowest_hostname = whole_hostname[slowest_rank_id].cpu().numpy().tobytes().decode("utf-8") + logging.critical(f"Slowest rank hostname: {slowest_hostname}") + + if self.slowdown_count(is_slowdown) and self.raise_error: + raise RuntimeError( + f"Detected GPU {slowest_rank_id} to be too slow compared to other GPUs." + f" The relative performance of {slowest_rank_id} rank was {report.gpu_relative_perf_scores[slowest_rank_id]}. Terminating the training." + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/misc_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/misc_test.py new file mode 100644 index 0000000000000000000000000000000000000000..4ff03d8c0d5f47461b2fc1bd095068e52e93797b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/misc_test.py @@ -0,0 +1,61 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Usage: + pytest -v -s cosmos_policy/_src/imaginaire/utils/misc_test.py +""" + +import pytest +import torch + +from cosmos_policy._src.imaginaire.utils.misc import get_data_batch_size + + +@pytest.mark.L0 +def test_get_data_batch_size(): + """ + Test get_data_batch_size function. + + This test verifies that the function returns the correct data batch size for various inputs. + """ + data_batch = {"images": torch.zeros(1, 3, 16, 16), "tokens": torch.zeros(1, 16)} + assert get_data_batch_size(data_batch) == 1 + + data_batch = {"images": torch.zeros(2, 3, 16, 16), "tokens": torch.zeros(2, 16)} + assert get_data_batch_size(data_batch) == 2 + + # Nested dictionary "__url__" without torch tensors - should be skipped + data_batch = { + "__key__": "value", + "__url__": {"k1": "v1", "k2": "v2"}, + "images": torch.zeros(2, 3, 16, 16), + "tokens": torch.zeros(2, 16), + } + + assert get_data_batch_size(data_batch) == 2 + + # Nested dictionary "image_dict" with torch tensors - should be counted + data_batch = { + "__key__": "value", + "__url__": {"k1": "v1", "k2": "v2"}, + "image_dict": {"images": torch.zeros(2, 3, 16, 16)}, + } + assert get_data_batch_size(data_batch) == 2 + + # Invalid data_batch that should raise ValueError + invalid_data_batch = {"__key__": "value", "__url__": {"k1": "v1", "k2": "v2"}, "tokens": [0, 1, 2]} + with pytest.raises(ValueError): + get_data_batch_size(invalid_data_batch) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/object_store.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/object_store.py new file mode 100644 index 0000000000000000000000000000000000000000..898255df9fff338bb247eda72e763b1845289154 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/object_store.py @@ -0,0 +1,377 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import io +import json +import os +import pickle +from pathlib import Path +from typing import TYPE_CHECKING, Any, Callable, Optional +from urllib.parse import urlparse + +import numpy as np +import torch +import yaml +from PIL import Image + +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + +Image.MAX_IMAGE_PIXELS = None + +if TYPE_CHECKING: + from cosmos_policy._src.imaginaire.config import ObjectStoreConfig + + +class ObjectStore: + """This is the interface class for object store, used for interacting with PBSS/AWS (S3). + + **Deprecated**. Use `easy_io` directly instead. + + Attributes: + easy_io_backend: easy_io backend. + bucket (str): Object store bucket name. + """ + + def __init__(self, config_object_storage: ObjectStoreConfig): + self.easy_io_backend = easy_io.get_file_backend( + backend_args={ + "backend": "s3", + "s3_credential_path": config_object_storage.credentials, + "path_mapping": None, + } + ) + self.bucket = config_object_storage.bucket + + def _translate_key(self, key: str) -> str: + """Translate an object key to an S3 URL for easy_io. + + Args: + key (str): The key of the object. + + Returns: + str: The object's S3 URL. + """ + return f"s3://{self.bucket}/{key}" + + def load_object( + self, + key: str, + type: str | None = None, + load_func: Callable | None = None, + encoding: str = "UTF-8", + ) -> Any: + """Helper function for loading object from storage. + + Args: + key (str): The key of the object. + type (str): Specified for some common data types. If not provided, `load_func` should be specified. + The predefined types currently supported are: + - "torch": PyTorch model checkpoints, opened with torch.load(). + - "torch.jit": A JIT-compiled TorchScript model, loaded with torch.jit.load(). + - "image": Image objects, opened with PIL.Image.open(). + - "json": JSON files, opened with json.load(). + - "pickle": Picklable objects, opened with pickle.load(). + - "yaml": YAML files, opened with yaml.safe_load(). + - "text": Pure text files. + - "numpy": Numpy arrays, opened with np.load(). + - "bytes": Raw bytes. + load_func (Callable): a custom function for reading the buffer if `type` were not provided. + encoding (str): Text encoding standard (default: "UTF-8"). + + Returns: + object (Any): The downloaded object. + """ + assert type is not None or load_func is not None, "Either type or load_func should be specified." + + buffer = io.BytesIO(self.easy_io_backend.get(filepath=self._translate_key(key=key))) + buffer.seek(0) + + # Read from buffer for common data types. + if type == "torch": + return torch.load(buffer, map_location=lambda storage, loc: storage, weights_only=False) + elif type == "torch.jit": + return torch.jit.load(buffer) + elif type == "image": + image = Image.open(buffer) + image.load() + return image + elif type == "json": + return json.load(buffer) + elif type == "jsonl": + data = [] + for line in buffer: + data.append(json.loads(line)) + return {"data": data} + elif type == "pickle": + return pickle.load(buffer) + elif type == "yaml": + return yaml.safe_load(buffer) + elif type == "text": + return buffer.read().decode(encoding) + elif type == "numpy": + return np.load(buffer, allow_pickle=True) + # Read from buffer as raw bytes. + elif type == "bytes": + return buffer.read() + # Customized load_func should be provided. + else: + return load_func(buffer) + + def save_object( + self, object: Any, key: str, type: str | None = None, save_func: Callable | None = None, encoding: str = "UTF-8" + ) -> None: + """Helper function for saving object to storage. + + Args: + object (Any): The object to upload. + key (str): The key of the object. + type (str): Specified for some common data types. If not provided, `save_func` should be specified. + The predefined types currently supported are: + - "torch": PyTorch model checkpoints, saved with torch.save(). + - "torch.jit": A JIT-compiled TorchScript model, exported with torch.jit.save(). + - "image": Image objects, saved with PIL.Image.save(). + - "json": JSON files, saved with json.dumps(). + - "pickle": Picklable objects, saved with pickle.dump(). + - "yaml": YAML files, saved with yaml.safe_dump(). + - "text": Pure text files. + - "numpy": Numpy arrays, saved with np.save(). + - "bytes": Raw bytes. + save_func (Callable): a custom function for writing the buffer if `type` were not provided. + encoding (str): Text encoding standard (default: "UTF-8"). + """ + assert type is not None or save_func is not None + with io.BytesIO() as buffer: + # Write to buffer for common data types. + if type == "torch": + torch.save(object, buffer) + elif type == "torch.jit": + torch.jit.save(object, buffer) + elif type == "image": + type = os.path.basename(key).split(".")[-1] + object.save(buffer, format=type) + elif type == "json": + buffer.write(json.dumps(object).encode(encoding)) + elif type == "pickle": + pickle.dump(object, buffer) + elif type == "yaml": + buffer.write(yaml.safe_dump(object).encode(encoding)) + elif type == "text": + buffer.write(object.encode(encoding)) + elif type == "numpy": + np.save(buffer, object) + # Write to buffer as raw bytes. + elif type == "bytes": + buffer.write(bytes(object)) + # Customized save_func should be provided. + else: + save_func(object, buffer) + buffer.seek(0) + self.easy_io_backend.put(obj=buffer, filepath=self._translate_key(key=key)) + + def object_exists(self, key: str) -> bool: + """ + Check whether an object exists in the storage, with retry logic for transient errors. + + Args: + key (str): The key of the object. + + Returns: + bool: True if the object exists, False if not. + """ + return self.easy_io_backend.exists(filepath=self._translate_key(key=key)) + + +def sync_s3_dir_to_local( + s3_dir: str, + s3_credential_path: str, + cache_dir: Optional[str] = None, + rank_sync: bool = True, + local_rank_sync: bool = False, +) -> str: + """ + Download an entire directory from S3 to the local cache directory. + + Args: + s3_dir (str): The AWS S3 directory to download. + s3_credential_path (str): The path to the AWS S3 credentials file. + rank_sync (bool, optional): Whether to synchronize download across + ALL distributed workers using `distributed.barrier()`. Defaults to True. + cache_dir (str, optional): The cache folder to sync the S3 directory to. + If None, the environment variable `IMAGINAIRE_CACHE_DIR` (defaulting + to "~/.cache/imaginaire") will be used. + local_rank_sync (bool, optional): Whether to synchronize download across + workers within the same node using a node-level barrier. This is useful + when the cache directory is not shared across nodes. Defaults to False. + Note: rank_sync and local_rank_sync cannot both be True. + + Returns: + local_dir (str): The path to the local directory. + """ + if local_rank_sync and rank_sync: + raise ValueError("rank_sync and local_rank_sync cannot be True at the same time.") + + if not s3_dir.startswith("s3://"): + # If the directory exists locally, return the local path + assert os.path.exists(s3_dir), f"{s3_dir} is not a S3 path or a local path." + return s3_dir + + # Get local rank for node-level synchronization + local_rank = int(os.getenv("LOCAL_RANK", 0)) if local_rank_sync else None + + easy_io_backend = easy_io.get_file_backend( + backend_args={ + "backend": "s3", + "s3_credential_path": s3_credential_path, + "path_mapping": None, + } + ) + + # Parse the S3 URL + parsed_url = urlparse(s3_dir) + obj_prefix = parsed_url.path.lstrip("/") + + # If the local directory is not specified, use the default cache directory + cache_dir = ( + os.environ.get("IMAGINAIRE_CACHE_DIR", os.path.expanduser("~/.cache/imaginaire")) + if cache_dir is None + else cache_dir + ) + cache_dir = os.path.expanduser(cache_dir) + Path(cache_dir).mkdir(parents=True, exist_ok=True) + + for obj_suffix in easy_io_backend.list_dir_or_file(dir_path=s3_dir, list_dir=False, list_file=True): + # Create the full path for the destination file, preserving the directory structure + dest_path = os.path.join(cache_dir, obj_prefix, obj_suffix) + + # Ensure the directory exists + os.makedirs(os.path.dirname(dest_path), exist_ok=True) + + # Check if the file already exists + if os.path.exists(dest_path): + continue + else: + s3_obj = f"{s3_dir.removesuffix('/')}/{obj_suffix}" + log.info(f"Downloading {s3_obj} to {dest_path}") + # Download the file + if rank_sync: + # Only rank 0 downloads when using global rank sync + if distributed.get_rank() == 0: + easy_io_backend.copyfile_to_local(src=s3_obj, dst=dest_path, dst_type="file") + elif local_rank_sync: + # Only local rank 0 (first rank on each node) downloads when using local rank sync + if local_rank == 0: + easy_io_backend.copyfile_to_local(src=s3_obj, dst=dest_path, dst_type="file") + else: + # No synchronization - every rank downloads + easy_io_backend.copyfile_to_local(src=s3_obj, dst=dest_path, dst_type="file") + # Synchronize after downloads complete + if rank_sync or local_rank_sync: + distributed.barrier() + + local_dir = os.path.join(cache_dir, obj_prefix) + return local_dir + + +def download_from_s3_with_cache( + s3_path: str, + s3_credential_path: str, + cache_fp: Optional[str] = None, + cache_dir: Optional[str] = None, + rank_sync: bool = True, + backend_args: Optional[dict] = None, + backend_key: Optional[str] = None, +) -> str: + """download data from S3 with optional caching. + + This function first attempts to load the data from a local cache file. If + the cache file doesn't exist, it downloads the data from S3 to the cache + location. Caching is performed in a rank-aware manner + using `distributed.barrier()` to ensure only one download occurs across + distributed workers (if `rank_sync` is True). + + Args: + s3_path (str): The S3 path of the data to load. + cache_fp (str, optional): The path to the local cache file. If None, + a filename will be generated based on `s3_path` within `cache_dir`. + cache_dir (str, optional): The directory to store the cache file. If + None, the environment variable `IMAGINAIRE_CACHE_DIR` (defaulting + to "/tmp") will be used. + rank_sync (bool, optional): Whether to synchronize download across + distributed workers using `distributed.barrier()`. Defaults to True. + backend_args (dict, optional): The backend arguments passed to easy_io to construct the backend. + backend_key (str, optional): The backend key passed to easy_io to registry the backend or retrieve the backend if it is already registered. + + Returns: + cache_fp (str): The path to the local cache file. + + Raises: + FileNotFoundError: If the data cannot be found in S3 or the cache. + """ + if not s3_path.startswith("s3://"): + # If the file exists locally, return the local path + assert os.path.exists(s3_path), f"{s3_path} is not a S3 path nor a local path." + return s3_path + + easy_io_backend = easy_io.get_file_backend( + backend_args={ + "backend": "s3", + "s3_credential_path": s3_credential_path, + "path_mapping": None, + } + ) + cache_dir = ( + os.environ.get("IMAGINAIRE_CACHE_DIR", os.path.expanduser("~/.cache/imaginaire")) + if cache_dir is None + else cache_dir + ) + cache_dir = os.path.expanduser(cache_dir) + if cache_fp is None: + cache_fp = os.path.join(cache_dir, s3_path.replace("s3://", "")) + if not cache_fp.startswith("/"): + cache_fp = os.path.join(cache_dir, cache_fp) + + if rank_sync: + if distributed.get_rank() == 0: + if os.path.exists(cache_fp): + # check the size of cache_fp + if os.path.getsize(cache_fp) < 1: + os.remove(cache_fp) + log.warning(f"Removed empty cache file {cache_fp}.") + + if not os.path.exists(cache_fp): + easy_io_backend.copyfile_to_local( + s3_path, cache_fp, dst_type="file", backend_args=backend_args, backend_key=backend_key + ) + log.info(f"Downloaded {s3_path} to {cache_fp}.") + else: + log.info(f"The cache file {cache_fp} already exists.") + distributed.barrier() + else: + if os.path.exists(cache_fp): + # check the size of cache_fp + if os.path.getsize(cache_fp) < 1: + os.remove(cache_fp) + log.warning(f"Removed empty cache file {cache_fp}.") + if not os.path.exists(cache_fp): + easy_io_backend.copyfile_to_local( + s3_path, cache_fp, dst_type="file", backend_args=backend_args, backend_key=backend_key + ) + log.info(f"Downloaded {s3_path} to {cache_fp}.") + else: + log.info(f"The cache file {cache_fp} already exists") + return cache_fp diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/optim_instantiate.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/optim_instantiate.py new file mode 100644 index 0000000000000000000000000000000000000000..d6e3c15be42f7656020dd12e89c77da5b8b84898 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/optim_instantiate.py @@ -0,0 +1,86 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import hydra +import torch +from torch import nn + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.fused_adam import FusedAdam + + +def get_regular_param_group(net: nn.Module): + """ + seperate the parameters of the network into two groups: decay and no_decay. + based on nano_gpt codebase. + """ + param_dict = {pn: p for pn, p in net.named_parameters()} + param_dict = {pn: p for pn, p in param_dict.items() if p.requires_grad} + + decay_params = [p for n, p in param_dict.items() if p.dim() >= 2] + nodecay_params = [p for n, p in param_dict.items() if p.dim() < 2] + return decay_params, nodecay_params + + +def get_base_optimizer( + model: nn.Module, + lr: float, + weight_decay: float, + optim_type: str = "adamw", + sharding: bool = False, + **kwargs, +) -> torch.optim.Optimizer: + net_decay_param, net_nodecay_param = get_regular_param_group(model) + + num_decay_params = sum(p.numel() for p in net_decay_param) + num_nodecay_params = sum(p.numel() for p in net_nodecay_param) + net_param_total = num_decay_params + num_nodecay_params + log.critical(f"total num parameters : {net_param_total:,}") + + param_group = [ + { + "params": net_decay_param + net_nodecay_param, + "lr": lr, + "weight_decay": weight_decay, + }, + ] + + if optim_type == "adamw": + opt_cls = torch.optim.AdamW + elif optim_type == "fusedadam": + opt_cls = FusedAdam + else: + raise ValueError(f"Unknown optimizer type: {optim_type}") + + return opt_cls(param_group, **kwargs) + + +def get_base_scheduler( + optimizer: torch.optim.Optimizer, + model: nn.Module, + scheduler_config: dict, +): + net_scheduler = hydra.utils.instantiate(scheduler_config) + net_scheduler.model = model + + num_param_groups = len(optimizer.param_groups) + + return torch.optim.lr_scheduler.LambdaLR( + optimizer, + lr_lambda=[ + net_scheduler.schedule, + ] + * num_param_groups, + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/parallel_state_helper.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/parallel_state_helper.py new file mode 100644 index 0000000000000000000000000000000000000000..a5630ca80d68300dd1b283d9448fe7433bd5ba81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/parallel_state_helper.py @@ -0,0 +1,35 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +This module contains various helper functions designed to extend the functionality of parallel states within the MCore library. + +MCore is a third-party library that is infrequently updated and may introduce backward compatibility issues in our codebase, such as changes in function signatures or missing / new functions in new versions. + +To mitigate these issues, this module provides stable functions that ensure the cosmos_policy._src.imaginaire codebase remains compatible with different versions of MCore. +""" + +try: + from megatron.core import parallel_state +except ImportError: + print("Megatron is not installed, is_tp_cp_pp_rank0 functions will not work.") + + +def is_tp_cp_pp_rank0(): + return ( + parallel_state.get_tensor_model_parallel_rank() == 0 + and parallel_state.get_pipeline_model_parallel_rank() == 0 + and parallel_state.get_context_parallel_rank() == 0 + ) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/primitives.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/primitives.py new file mode 100644 index 0000000000000000000000000000000000000000..2b85de49645afcaa6499369ed30f519f820f093d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/primitives.py @@ -0,0 +1,29 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +def is_primitive(value): + return isinstance(value, (int, float, str, bool, type(None))) + + +def convert_to_primitive(value): + if isinstance(value, (list, tuple)): + return [convert_to_primitive(v) for v in value if is_primitive(v) or isinstance(v, (list, dict))] + elif isinstance(value, dict): + return {k: convert_to_primitive(v) for k, v in value.items() if is_primitive(v) or isinstance(v, (list, dict))} + elif is_primitive(value): + return value + else: + return "non-primitive" # Skip non-primitive types diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/profiling.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/profiling.py new file mode 100644 index 0000000000000000000000000000000000000000..af3940ef8726cc4f3614d7488230bf5d8ef76c30 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/profiling.py @@ -0,0 +1,139 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import contextlib +import os +import time + +import torch + +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + +# (qsh 2024-11-23) credits +# https://github.com/pytorch/torchtitan/blob/main/torchtitan/profiling.py + +# the number of warmup steps before the active step in each profiling cycle +TORCH_TRACE_WARMUP = 3 + +# how much memory allocation/free ops to record in memory snapshots +MEMORY_SNAPSHOT_MAX_ENTRIES = 100000 + + +@contextlib.contextmanager +def maybe_enable_profiling(config, *, global_step: int = 0): + # get user defined profiler settings + enable_profiling = config.trainer.profiling.enable_profiling + profile_freq = config.trainer.profiling.profile_freq + + if enable_profiling: + trace_dir = os.path.join(config.job.path_local, "torch_trace") + if distributed.get_rank() == 0: + os.makedirs(trace_dir, exist_ok=True) + + rank = distributed.get_rank() + + def trace_handler(prof): + curr_trace_dir_name = "iteration_" + str(prof.step_num) + curr_trace_dir = os.path.join(trace_dir, curr_trace_dir_name) + if not os.path.exists(curr_trace_dir): + os.makedirs(curr_trace_dir, exist_ok=True) + + log.info(f"Dumping traces at step {prof.step_num}") + begin = time.monotonic() + if rank in config.trainer.profiling.target_ranks: + prof.export_chrome_trace(f"{curr_trace_dir}/rank{rank}_trace.json.gz") + log.info(f"Finished dumping traces in {time.monotonic() - begin:.2f} seconds") + + log.info(f"Profiling active. Traces will be saved at {trace_dir}") + + if not os.path.exists(trace_dir): + os.makedirs(trace_dir, exist_ok=True) + + warmup, active = TORCH_TRACE_WARMUP, 1 + wait = profile_freq - (active + warmup) + assert wait >= 0, "profile_freq must be greater than or equal to warmup + active" + + with torch.profiler.profile( + activities=[ + torch.profiler.ProfilerActivity.CPU, + torch.profiler.ProfilerActivity.CUDA, + ], + schedule=torch.profiler.schedule(wait=wait, warmup=warmup, active=active), + on_trace_ready=trace_handler, + record_shapes=config.trainer.profiling.record_shape, + profile_memory=config.trainer.profiling.profile_memory, + with_stack=config.trainer.profiling.with_stack, + with_modules=config.trainer.profiling.with_modules, + ) as torch_profiler: + torch_profiler.step_num = global_step + yield torch_profiler + else: + torch_profiler = contextlib.nullcontext() + yield None + + +@contextlib.contextmanager +def maybe_enable_memory_snapshot(config, *, global_step: int = 0): + enable_snapshot = config.trainer.profiling.enable_memory_snapshot + if enable_snapshot: + if config.trainer.profiling.save_s3: + snapshot_dir = "s3://rundir" + else: + snapshot_dir = os.path.join(config.job.path_local, "memory_snapshot") + if distributed.get_rank() == 0: + os.makedirs(snapshot_dir, exist_ok=True) + + rank = torch.distributed.get_rank() + + class MemoryProfiler: + def __init__(self, step_num: int, freq: int): + torch.cuda.memory._record_memory_history(max_entries=MEMORY_SNAPSHOT_MAX_ENTRIES) + # when resume training, we start from the last step + self.step_num = step_num + self.freq = freq + + def step(self, exit_ctx: bool = False): + self.step_num += 1 + if not exit_ctx and self.step_num % self.freq != 0: + return + if not exit_ctx: + curr_step = self.step_num + dir_name = f"iteration_{curr_step}" + else: + # dump as iteration_0_exit if OOM at iter 1 + curr_step = self.step_num - 1 + dir_name = f"iteration_{curr_step}_exit" + curr_snapshot_dir = os.path.join(snapshot_dir, dir_name) + if not config.trainer.profiling.save_s3 and not os.path.exists(curr_snapshot_dir): + os.makedirs(curr_snapshot_dir, exist_ok=True) + log.info(f"Dumping memory snapshot at step {curr_step}") + begin = time.monotonic() + + if rank in config.trainer.profiling.target_ranks: + easy_io.dump( + torch.cuda.memory._snapshot(), + f"{curr_snapshot_dir}/rank{rank}_memory_snapshot.pickle", + ) + log.info(f"Finished dumping memory snapshot in {time.monotonic() - begin:.2f} seconds") + + log.info(f"Memory profiler active. Snapshot will be saved at {snapshot_dir}") + profiler = MemoryProfiler(global_step, config.trainer.profiling.profile_freq) + try: + yield profiler + except torch.cuda.OutOfMemoryError: + profiler.step(exit_ctx=True) + else: + yield None diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/registry.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/registry.py new file mode 100644 index 0000000000000000000000000000000000000000..6f27c1a0ea387a2ab0b5ac4b5503b4ce62b88de9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/registry.py @@ -0,0 +1,160 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Utilities for managing registries. +Credit: https://gitlab.com/qsh.zh/jam/-/blob/master/jammy/utils/registry.py with MIT License +""" + +import collections + +__all__ = [ + "Registry", + "DefaultRegistry", + "RegistryGroup", + "CallbackRegistry", +] + + +class Registry: + __FALLBACK_KEY__ = "__fallback__" + + _registry = None + + def __init__(self): + self._init_registry() + + def _init_registry(self): + self._registry = {} + + @property + def fallback(self): + return self._registry.get(self.__FALLBACK_KEY__, None) + + def set_fallback(self, value): + self._registry[self.__FALLBACK_KEY__] = value + return self + + def register(self, entry, value): + self._registry[entry] = value + return self + + def unregister(self, entry): + return self._registry.pop(entry, None) + + def has(self, entry): + return entry in self._registry + + def lookup(self, entry, fallback=True, default=None): + if fallback: + fallback_value = self._registry.get(self.__FALLBACK_KEY__, default) + else: + fallback_value = default + return self._registry.get(entry, fallback_value) + + def keys(self): + return list(self._registry.keys()) + + def items(self): + return list(self._registry.items()) + + +class DefaultRegistry(Registry): + __base_class__ = dict + + def _init_registry(self): + base_class = type(self).__base_class__ + self._registry = collections.defaultdict(base_class) + + def lookup(self, entry, fallback=False, default=None): + assert fallback is False and default is None + return self._registry[entry] + + def __getitem__(self, item): + return self.lookup(item) + + +class RegistryGroup: + __base_class__ = Registry + + def __init__(self): + self._init_registry_group() + + def _init_registry_group(self): + base_class = type(self).__base_class__ + self._registries = collections.defaultdict(base_class) + + def __getitem__(self, item): + return self._registries[item] + + def register(self, registry_name, entry, value, **kwargs): + return self._registries[registry_name].register(entry, value, **kwargs) + + def lookup(self, registry_name, entry, fallback=True, default=None): + return self._registries[registry_name].lookup(entry, fallback=fallback, default=default) + + +class CallbackRegistry(Registry): + """ + A callable manager utils. + + If there exists a super callback, it will block all callbacks. + A super callback will receive the called name as its first argument. + + Then the dispatcher will try to call the callback by name. + If such name does not exists, a fallback callback will be called. + + The fallback callback will also receive the called name as its first argument. + + Examples: + + >>> registry = CallbackRegistry() + >>> callback_func = print + >>> registry.register('name', callback_func) # register a callback. + >>> registry.dispatch('name', 'arg1', 'arg2', kwarg1='kwarg1') # dispatch. + """ + + def __init__(self): + super().__init__() + self._super_callback = None + + @property + def super_callback(self): + return self._super_callback + + def set_super_callback(self, callback): + self._super_callback = callback + return self + + @property + def fallback_callback(self): + return self.fallback + + def set_fallback_callback(self, callback): + return self.set_fallback(callback) + + def dispatch(self, name, *args, **kwargs): + if self._super_callback is not None: + return self._super_callback(self, name, *args, **kwargs) + return self.dispatch_direct(name, *args) + + def dispatch_direct(self, name, *args, **kwargs): + """Dispatch by name, ignoring the super callback.""" + callback = self.lookup(name, fallback=False) + if callback is None: + if self.fallback_callback is None: + raise ValueError('Unknown callback entry: "{}".'.format(name)) + return self.fallback_callback(name, *args, **kwargs) + return callback(*args, **kwargs) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/replace_bg_color.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/replace_bg_color.py new file mode 100644 index 0000000000000000000000000000000000000000..b48a810f3415a111c949ee1271fe4b08cbb7c23e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/replace_bg_color.py @@ -0,0 +1,124 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import io +import re + +import numpy as np +from PIL import Image + +_IMG_EXTENSIONS = "jpg jpeg png ppm pgm pbm pnm".split() + + +def lin2srgb(lin): + """Convert sRGB values to physically linear ones. The transformation is + uniform in RGB, so *srgb* can be of any shape. + + *srgb* values should range between 0 and 1, inclusively. + + """ + gamma = 1.055 * lin ** (1.0 / 2.4) - 0.055 + scale = 12.92 * lin + return np.where(lin > 0.0031308, gamma, scale) + + +def srgb2lin(srgb): + """Convert sRGB values to physically linear ones. The transformation is + uniform in RGB, so *srgb* can be of any shape. + + *srgb* values should range between 0 and 1, inclusively. + + """ + gamma = ((srgb + 0.055) / 1.055) ** 2.4 + scale = srgb / 12.92 + return np.where(srgb > 0.04045, gamma, scale) + + +def replace_bg_color_u8(fg: np.array, fg_mask: np.array, bg_color_old: list, bg_color_new: list): + r"""Given an image with background, as well as the foreground mask and old background color, + Replace the old background color with the new one. + Assuming everything is in uint8 + Args: + fg [..., 3] np.array + fg_mask[..., 1] np.array: 0 -> full background; 255 -> full foreground. + bg_color_old [3] RGB 0-255: Old background. + bg_color_new [3] RGB 0-255: New background + """ + assert fg.dtype == np.uint8 and fg_mask.dtype == np.uint8 + fg_mask = fg_mask.astype(np.float32) / 255.0 + fg = fg.astype(np.float32) / 255.0 + bg_color_old = np.array(bg_color_old, dtype=np.float32) / 255.0 + bg_color_new = np.array(bg_color_new, dtype=np.float32) / 255.0 + bg_mask = 1.0 - fg_mask + result = srgb2lin(fg) + bg_mask * (srgb2lin(bg_color_new) - srgb2lin(bg_color_old)) + result = lin2srgb(result) + result = np.clip((result * 255.0).round(), 0, 255).astype(np.uint8) + return result + + +def replace_bg_color_pil(fg_pil: Image.Image, fg_mask_pil: Image.Image, bg_color_old: list, bg_color_new: list): + fg = np.array(fg_pil) + fg_mask = np.array(fg_mask_pil) + if fg_mask.ndim == 2: + fg_mask = fg_mask[..., None] + else: + fg_mask = fg_mask[..., :1] + result = replace_bg_color_u8(fg, fg_mask, bg_color_old, bg_color_new) + return Image.fromarray(result) + + +def pil_loader_with_mask(key, data, background_color_new=None, background_color_old=[255, 255, 255], mask=None): + r""" + Function to load an image. + If the image is corrupt, it returns a black image. + Args: + key: Image key. + data: Image data stream. + """ + extension = re.sub(r".*[.]", "", key) + if extension.lower() not in _IMG_EXTENSIONS: + return None + + with io.BytesIO(data) as stream: + img = Image.open(stream) + img = img.convert("RGB") + if background_color_new is not None: + assert mask is not None + with io.BytesIO(mask) as stream: + mask = Image.open(stream) + mask.load() + mask = mask.convert("L") + img = replace_bg_color_pil(img, mask, background_color_old, background_color_new) + return img + + +def pil_loader(key, data, type="RGB"): + r""" + Function to load an image. + If the image is corrupt, it returns a black image. + Args: + key: Image key. + data: Image data stream. + """ + extension = re.sub(r".*[.]", "", key) + if extension.lower() not in _IMG_EXTENSIONS: + return None + + with io.BytesIO(data) as stream: + img = Image.open(stream) + img.load() + img = img.convert(type) + + return img diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/s3_utils.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/s3_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..e7816296af49da82e4861c9d6de71a048481f8af --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/s3_utils.py @@ -0,0 +1,139 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +from typing import Any, Optional + +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + + +def download_from_s3_with_cache( + s3_path: str, + cache_fp: Optional[str] = None, + cache_dir: Optional[str] = None, + rank_sync: bool = True, + backend_args: Optional[dict] = None, + backend_key: Optional[str] = None, +) -> str: + """download data from S3 with optional caching. + + This function first attempts to load the data from a local cache file. If + the cache file doesn't exist, it downloads the data from S3 to the cache + location. Caching is performed in a rank-aware manner + using `distributed.barrier()` to ensure only one download occurs across + distributed workers (if `rank_sync` is True). + + Args: + s3_path (str): The S3 path of the data to load. + cache_fp (str, optional): The path to the local cache file. If None, + a filename will be generated based on `s3_path` within `cache_dir`. + cache_dir (str, optional): The directory to store the cache file. If + None, the environment variable `IMAGINAIRE_CACHE_DIR` (defaulting + to "/tmp") will be used. + rank_sync (bool, optional): Whether to synchronize download across + distributed workers using `distributed.barrier()`. Defaults to True. + backend_args (dict, optional): The backend arguments passed to easy_io to construct the backend. + backend_key (str, optional): The backend key passed to easy_io to registry the backend or retrieve the backend if it is already registered. + + Returns: + cache_fp (str): The path to the local cache file. + + Raises: + FileNotFoundError: If the data cannot be found in S3 or the cache. + """ + cache_dir = os.environ.get("TORCH_HOME") if cache_dir is None else cache_dir + cache_dir = ( + os.environ.get("IMAGINAIRE_CACHE_DIR", os.path.expanduser("~/.cache/imaginaire")) + if cache_dir is None + else cache_dir + ) + cache_dir = os.path.expanduser(cache_dir) + if cache_fp is None: + cache_fp = os.path.join(cache_dir, s3_path.replace("s3://", "")) + if not cache_fp.startswith("/"): + cache_fp = os.path.join(cache_dir, cache_fp) + + if distributed.get_rank() == 0: + if os.path.exists(cache_fp): + # check the size of cache_fp + if os.path.getsize(cache_fp) < 1: + os.remove(cache_fp) + log.warning(f"Removed empty cache file {cache_fp}.") + + if rank_sync: + if not os.path.exists(cache_fp): + log.critical(f"Local cache {cache_fp} Not exist! Downloading {s3_path} to {cache_fp}.") + log.info(f"backend_args: {backend_args}") + log.info(f"backend_key: {backend_key}") + + easy_io.copyfile_to_local( + s3_path, cache_fp, dst_type="file", backend_args=backend_args, backend_key=backend_key + ) + log.info(f"Downloaded {s3_path} to {cache_fp}.") + else: + log.info(f"Local cache {cache_fp} already exist! {s3_path} -> {cache_fp}.") + + distributed.barrier() + else: + if not os.path.exists(cache_fp): + easy_io.copyfile_to_local( + s3_path, cache_fp, dst_type="file", backend_args=backend_args, backend_key=backend_key + ) + + log.info(f"Downloaded {s3_path} to {cache_fp}.") + return cache_fp + + +def load_from_s3_with_cache( + s3_path: str, + cache_fp: Optional[str] = None, + cache_dir: Optional[str] = None, + rank_sync: bool = True, + backend_args: Optional[dict] = None, + backend_key: Optional[str] = None, + easy_io_kwargs: Optional[dict] = None, +) -> Any: + """Loads data from S3 with optional caching. + + This function first attempts to load the data from a local cache file. If + the cache file doesn't exist, it downloads the data from S3 to the cache + location and then loads it. Caching is performed in a rank-aware manner + using `distributed.barrier()` to ensure only one download occurs across + distributed workers (if `rank_sync` is True). + + Args: + s3_path (str): The S3 path of the data to load. + cache_fp (str, optional): The path to the local cache file. If None, + a filename will be generated based on `s3_path` within `cache_dir`. + cache_dir (str, optional): The directory to store the cache file. If + None, the environment variable `IMAGINAIRE_CACHE_DIR` (defaulting + to "/tmp") will be used. + rank_sync (bool, optional): Whether to synchronize download across + distributed workers using `distributed.barrier()`. Defaults to True. + backend_args (dict, optional): The backend arguments passed to easy_io to construct the backend. + backend_key (str, optional): The backend key passed to easy_io to registry the backend or retrieve the backend if it is already registered. + + Returns: + Any: The loaded data from the S3 path or cache file. + + Raises: + FileNotFoundError: If the data cannot be found in S3 or the cache. + """ + cache_fp = download_from_s3_with_cache(s3_path, cache_fp, cache_dir, rank_sync, backend_args, backend_key) + + if easy_io_kwargs is None: + easy_io_kwargs = {} + return easy_io.load(cache_fp, **easy_io_kwargs) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/scheduler.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/scheduler.py new file mode 100644 index 0000000000000000000000000000000000000000..e45eb96a94208ad86348a785ab799bd8e71c2c7c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/scheduler.py @@ -0,0 +1,64 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math +from typing import List + +import torch + + +class WarmupLambdaLR(torch.optim.lr_scheduler.LambdaLR): + def __init__(self, optimizer, warmup, last_epoch=-1, verbose=False): + # Define the lambda function based on the warmup period + self.warmup = warmup + + def lr_lambda(epoch): + # Increase lr linearly for the first 'warmup' epochs + if epoch < warmup: + return float(epoch + 1) / warmup + # After 'warmup' epochs, keep lr constant + return 1.0 + + # Initialize the parent class with the generated lr_lambda + super(WarmupLambdaLR, self).__init__(optimizer, lr_lambda, last_epoch) + + +# cosine lr decay scheduler with warmup from https://github.com/karpathy/nanoGPT/blob/master/train.py#L228 +class WarmupCosineLR(torch.optim.lr_scheduler.LRScheduler): + def __init__( + self, + optimizer: torch.optim.Optimizer, + warmup_iters: int, + lr_decay_iters: int, + min_lr: float, + last_epoch: int = -1, + ): + self.warmup_iters = warmup_iters + self.lr_decay_iters = lr_decay_iters + self.min_lr = min_lr + super().__init__(optimizer, last_epoch) + + def get_lr(self) -> List[float]: + # 1) linear warmup for warmup_iters steps + if self.last_epoch < self.warmup_iters: + return [base_lr * self.last_epoch / self.warmup_iters for base_lr in self.base_lrs] + # 2) if it > lr_decay_iters, return min learning rate + if self.last_epoch > self.lr_decay_iters: + return [self.min_lr for _ in self.base_lrs] + # 3) in between, use cosine decay down to min learning rate + decay_ratio = (self.last_epoch - self.warmup_iters) / (self.lr_decay_iters - self.warmup_iters) + assert 0 <= decay_ratio <= 1 + coeff = 0.5 * (1.0 + math.cos(math.pi * decay_ratio)) # coeff ranges 0..1 + return [self.min_lr + coeff * (base_lr - self.min_lr) for base_lr in self.base_lrs] diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/submit_job_helper.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/submit_job_helper.py new file mode 100644 index 0000000000000000000000000000000000000000..5619aa668d2291a889eb5fee1a9b478146679b02 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/submit_job_helper.py @@ -0,0 +1,37 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +import os.path as osp + +import git +from loguru import logger as logging + + +def is_git(path): + try: + _ = git.Repo(path, search_parent_directories=True).git_dir + return True + except git.exc.InvalidGitRepositoryError: + return False + + +def git_rootdir(path=""): + if is_git(os.getcwd()): + git_repo = git.Repo(os.getcwd(), search_parent_directories=True) + root = git_repo.git.rev_parse("--show-toplevel") + return osp.join(root, path) + logging.info("not a git repo") + return osp.join(os.getcwd(), path) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/timer.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/timer.py new file mode 100644 index 0000000000000000000000000000000000000000..da18855b467f6f420dc785a20017aac1638f6279 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/timer.py @@ -0,0 +1,288 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Timer: helps measure CPU and CUDA times easily and reliably. +""" + +import time +from contextlib import ContextDecorator +from functools import wraps +from typing import Callable + +import torch + +from cosmos_policy._src.imaginaire.utils import log + + +def _autoformat_time_us(time_us: float) -> str: + """ + Automatically format time in nanoseconds. + """ + if time_us >= 1e6: + time_s = time_us * 1e-6 + return f"{time_s:.2f} s" + + if time_us >= 1e3: + time_ms = time_us * 1e-3 + return f"{time_ms:.2f} ms" + + return f"{time_us:.2f} us" + + +def format_time_str(time_us: float, unit: str | None = None) -> str: + """ + Automatically format time in nanoseconds either automatically or based on + desired unit. + """ + if unit is None: + return _autoformat_time_us(time_us) + + if unit == "us": + return f"{time_us:.2f} us" + + if unit == "ms": + return f"{time_us * 1e-3:.2f} ms" + + if unit == "s": + return f"{time_us * 1e-6:.2f} s" + + raise NotImplementedError(f"Time unit {unit} is not supported.") + + +def format_time(time_us: float, unit: str) -> float: + """ + Format time in nanoseconds based on desired unit. + """ + + if unit == "us": + return time_us + + if unit == "ms": + return time_us * 1e-3 + + if unit == "s": + return time_us * 1e-6 + + raise NotImplementedError(f"Time unit {unit} is not supported.") + + +class Timer(ContextDecorator): + """ + Reliable CPU and CUDA Timer. + + Args: + tag (str | None): Optional tag used in logs/prints. + + measure_cpu (bool): Whether to measure CPU time (using `time`). Default: `True`. + + measure_cuda (bool): Whether to measure CUDA time (using CUDA events). Default: `True`. + + unit (str | None): Optional time unit. Must be either "s" (seconds), "ms" (microseconds), + "us" (nanoseconds), or None (format automatically based on value). + + debug (bool): Whether to log results in debug mode instead of info. Default is False. + + Examples: + ```python + with Timer(measure_cpu=True, measure_cuda=True, unit="ms"): + model(x) + ``` + + ```python + @Timer(measure_cpu=True, measure_cuda=True, unit="ms") + def func(x): + return model(x) + ``` + """ + + def __init__( + self, + tag: str | None = None, + measure_cpu: bool = True, + measure_cuda: bool = True, + unit: str | None = None, + debug: bool = False, + ): + self.measure_cpu = measure_cpu + self.measure_cuda = measure_cuda + + self.measured = False + self.cpu_time_us = 0 + self.cuda_time_us = 0 + + self.busy = False + self.cpu_time_start = None + self.cuda_start_event = None + self.cuda_end_event = None + self.cuda_stream = None + + self.tag = "unknown" if tag is None else tag + self.unit = unit + if self.unit is not None and self.unit not in ["s", "ms", "us"]: + raise NotImplementedError(f"Time unit {self.unit} is not supported.") + + self.debug = debug + + def _log(self, msg: str): + if self.debug: + log.debug(msg) + else: + log.info(msg) + + def __enter__(self): + self.start() + + def __exit__(self, exc_type, exc_value, traceback): + self.end() + self.report() + + def __call__(self, func: Callable) -> Callable: + @wraps(func) + def wrapper(*args, **kwargs): # noqa: ANN202 + self.start() + result = func(*args, **kwargs) + self.end() + self.report() + return result + + return wrapper # type: ignore + + def report(self): + """ + Reports measurements. + """ + if self.measure_cpu and self.measure_cuda: + self._log(f"Time spent on {self.tag}: CPU: {self.get_cpu_time_str()}, CUDA: {self.get_cuda_time_str()}") + elif self.measure_cpu: + self._log(f"Time spent on {self.tag}: {self.get_cpu_time_str()}") + elif self.measure_cuda: + self._log(f"CUDA time spent on {self.tag}: {self.get_cuda_time_str()}") + else: + raise NotImplementedError() + + def get_cpu_time(self) -> float: + """ + Returns CPU time measurement. + """ + if not self.measure_cpu: + raise RuntimeError(f"CPU timer is disabled ({self.measure_cpu=}).") + + if not self.measured: + raise RuntimeError("No measurements were made yet!") + + if self.unit is None: + raise RuntimeError("No unit was specified. Please use get_cpu_time_str() instead.") + + assert self.unit is not None + return format_time(self.cpu_time_us, unit=self.unit) + + def get_cuda_time(self) -> float: + """ + Returns CUDA time measurement. + """ + if not self.measure_cuda: + raise RuntimeError(f"CUDA timer is disabled ({self.measure_cuda=}).") + + if not self.measured: + raise RuntimeError("No measurements were made yet!") + + if self.unit is None: + raise RuntimeError("No unit was specified. Please use get_cuda_time_str() instead.") + + assert self.unit is not None + return format_time(self.cuda_time_us, unit=self.unit) + + def get_cpu_time_str(self) -> str: + """ + Returns CPU time measurement in string format. + """ + if not self.measure_cpu: + raise RuntimeError(f"CPU timer is disabled ({self.measure_cpu=}).") + + if not self.measured: + raise RuntimeError("No measurements were made yet!") + + return format_time_str(self.cpu_time_us, unit=self.unit) + + def get_cuda_time_str(self) -> str: + """ + Returns CUDA time measurement in string format. + """ + if not self.measure_cuda: + raise RuntimeError(f"CUDA timer is disabled ({self.measure_cuda=}).") + + if not self.measured: + raise RuntimeError("No measurements were made yet!") + + return format_time_str(self.cuda_time_us, unit=self.unit) + + def reset(self): + """ + Resets recorded measurements + """ + self.measured = False + self.cpu_time_us = 0 + self.cuda_time_us = 0 + + def start(self, cuda_device: torch.device | None = None, cuda_stream: torch.cuda.Stream | None = None): + """ + Start time measurements. + + Args: + cuda_device (torch.device | None): CUDA device. Will use default CUDA device if not indicated. + + cuda_stream (torch.cuda.Stream | None): CUDA stream to use for CUDA time measurement. + Will use default stream for current CUDA device if not indicated. + """ + if self.busy: + raise RuntimeError("Already called Timer.start() once!") + + self.busy = True + + if self.measure_cuda: + self.cuda_stream = cuda_stream if cuda_stream is not None else torch.cuda.current_stream(cuda_device) + self.cuda_stream.synchronize() + + if self.measure_cpu: + self.cpu_time_start = time.time() + + if self.measure_cuda: + self.cuda_start_event = torch.cuda.Event(enable_timing=True) + self.cuda_end_event = torch.cuda.Event(enable_timing=True) + self.cuda_stream.record_event(self.cuda_start_event) + + def end(self): + """ + Ends time measurements. + + NOTE: must be done on the same CUDA device and stream as start(). + """ + if not self.busy: + raise RuntimeError("Timer.start() must be called exactly once before end()!") + + if self.measure_cuda: + self.cuda_stream.record_event(self.cuda_end_event) + self.cuda_end_event.synchronize() + + if self.measure_cpu: + self.cpu_time_end = time.time() + self.cpu_time_us = (self.cpu_time_end - self.cpu_time_start) * 1e6 + + if self.measure_cuda: + self.cuda_time_us = self.cuda_start_event.elapsed_time(self.cuda_end_event) * 1e3 + + self.busy = False + self.measured = True diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/tone_curve.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/tone_curve.py new file mode 100644 index 0000000000000000000000000000000000000000..2f394bca4a7cdfe6847a834b21382b866e7cc8e5 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/tone_curve.py @@ -0,0 +1,197 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math +from typing import Literal + +import numpy as np +from PIL import Image + + +def lin2srgb(lin): + """Convert sRGB values to physically linear ones. The transformation is + uniform in RGB, so *srgb* can be of any shape. + + *srgb* values should range between 0 and 1, inclusively. + + """ + gamma = 1.055 * lin ** (1.0 / 2.4) - 0.055 + scale = 12.92 * lin + return np.where(lin > 0.0031308, gamma, scale) + + +def srgb2lin(srgb): + """Convert sRGB values to physically linear ones. The transformation is + uniform in RGB, so *srgb* can be of any shape. + + *srgb* values should range between 0 and 1, inclusively. + + """ + gamma = ((srgb + 0.055) / 1.055) ** 2.4 + scale = srgb / 12.92 + return np.where(srgb > 0.04045, gamma, scale) + + +def commerce_tonemap(color): + startCompression = 0.8 - 0.04 + desaturation = 0.15 + + x = np.min(color, axis=-1, keepdims=True) + offset = np.where(x < 0.08, x - 6.25 * x * x, 0.04) + color -= offset + peak = np.max(color, axis=-1, keepdims=True) + uncompressed = color + + d = 1.0 - startCompression + newPeak = 1.0 - d * d / (peak + d - startCompression) + with np.errstate(divide="ignore", invalid="ignore"): # Avoid error print + color = color * (newPeak / peak) + + g = 1.0 - 1.0 / (desaturation * (peak - newPeak) + 1.0) + + compressed = color * (1 - g) + newPeak * g + + return np.where(peak < startCompression, uncompressed, compressed) + + +# https://github.com/RenderKit/oidn/blob/master/training/color.py + + +# Computes the luminance of an RGB color +def luminance(r, g, b): + return 0.212671 * r + 0.715160 * g + 0.072169 * b + + +# Computes an autoexposure value for a NumPy image +def autoexposure(image, mask, key=0.18): + maxBinSize = 16 # downsampling amount + eps = 1e-8 + + image = image * mask + # Compute the luminance of each pixel + r = image[..., 0] + g = image[..., 1] + b = image[..., 2] + L = luminance(r, g, b) + + # Center crop if the image size is not whole multiple of maxBinSize + crop_H = L.shape[0] // maxBinSize * maxBinSize + pad_top = round((L.shape[0] - crop_H) / 2) + crop_W = L.shape[1] // maxBinSize * maxBinSize + pad_left = round((L.shape[1] - crop_W) / 2) + L = L[pad_top : pad_top + crop_H, pad_left : pad_left + crop_W] + mask = mask[pad_top : pad_top + crop_H, pad_left : pad_left + crop_W] + + # Downsample the image to minimize sensitivity to noise + H = L.shape[0] # original height + W = L.shape[1] # original width + L = L.reshape(H // maxBinSize, maxBinSize, W // maxBinSize, maxBinSize) + L = np.mean(L, axis=(1, 3)) + mask = mask.reshape(H // maxBinSize, maxBinSize, W // maxBinSize, maxBinSize) + mask = np.mean(mask, axis=(1, 3)) + with np.errstate(divide="ignore", invalid="ignore"): # Avoid error print + L /= mask + L = L[mask > eps] + + # Keep only values greater than epsilon + L = L[L > eps] + if L.size == 0: + return 1.0 + + # Compute the exposure value + return float(key / np.exp2(np.log2(L).mean())) + + +# Default values changed to identity transformation, aka do nothing. +def apply_tone_curve( + imgs: list[Image.Image], + input_mapping: Literal["log", "straight"] = "log", + output_mapping: Literal["commerce", "straight", "log"] = "commerce", + exposure_bias: float = 1.5, + auto: bool = True, + ae_pregain: float = 1.0, + ae_key: float = 0.18, + ae_strength_below: float = 1.0, + ae_strength_above: float = 1.0, +) -> tuple[list[Image.Image], float]: + r"""Adjust the exposure of a list of images together. + For cam_v1 data, use input_mapping="log" + For cam_v2 data, use input_mapping="straight" + Some of the previous models are trained with output_mapping="commerce". This is a very forgiving curve. + But to match the style of PixelSquid, use output_mapping="straight" + See https://docs.google.com/document/d/1z08rWvWzqd_tNPlh7_D4aIkdaLAerSSagXK4pQlQxCk/edit for detail + + Args: + imgs: list of PIL images + + Returns: + ret: list of PIL images with exposure adjusted + """ + num_imgs = len(imgs) + img = np.concatenate([np.asarray(x) for x in imgs], axis=0).astype(np.float32) / 255.0 + mask = img[..., 3:4].astype(np.float32) # H,W,1 + img = img[..., :3] # Remove alpha + + img = srgb2lin(img) + + if input_mapping == "log": + img = np.exp(img) - 1 + elif input_mapping == "straight": + pass + else: + raise NotImplementedError(f"Unknown input_mapping: {input_mapping}") + + if auto: + img *= ae_pregain + exposure = autoexposure(img, mask, key=ae_key) + log_exposure = math.log2(exposure) + if log_exposure <= 0: + log_exposure *= ae_strength_below + else: + log_exposure *= ae_strength_above + exposure = 2.0**log_exposure + else: + exposure = 1.0 + exposure *= exposure_bias + + img = img * exposure + + if output_mapping == "commerce": + img = commerce_tonemap(img) + elif output_mapping == "log": + img = np.log(img + 1) + elif output_mapping == "straight": + pass + else: + raise NotImplementedError(f"Unknown output_mapping: {output_mapping}") + + img = lin2srgb(img) + img = np.concatenate([img, mask], axis=-1) + img = np.clip((img * 255.0).round(), 0, 255).astype(np.uint8) + return [Image.fromarray(x) for x in np.split(img, num_imgs, axis=0)], exposure + + +def apply_exposure(img: Image, exposure: float) -> Image: + r"""Apply exposure adjustment to a PIL image. + Args: + img: a PIL image, RGB or RGBA + exposure: exposure value + Returns: + img: PIL image with exposure adjusted + """ + img = np.asarray(img).astype(np.float32) / 255.0 + img[..., :3] = lin2srgb(srgb2lin(img[..., :3]) * exposure) + img = np.clip((img * 255.0).round(), 0, 255).astype(np.uint8) + return Image.fromarray(img) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/training.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/training.py new file mode 100644 index 0000000000000000000000000000000000000000..5e4a27eb43b9f1f1668adc395ddacfc61eea83a0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/training.py @@ -0,0 +1,171 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import torch + +from cosmos_policy._src.imaginaire.functional.batch_ops import batch_mul +from cosmos_policy._src.imaginaire.utils import log + + +def random_dropout(embeddings, drop_rate): + r""" + Function to perform random dropout for embeddings. + When we drop embeddings, we zero them out. + Args: + embeddings (tensor): Input embeddings + drop_rate (float): Rate of dropping the embedding. + """ + num_samples = embeddings.shape[0] + # Create a shape (num_samples, 1, 1, 1, 1, ...) depending on embeddings dim. + # This is done to ensure we can broadcast the zero_flag to the embeddings. + # embeddings.ndim is 3 for images, and 4 for videos, and the corresponding + # shapes are (num_samples, 1, 1) and (num_samples, 1, 1, 1) respectively. + tensor_shape = (num_samples,) + tuple([1] * (embeddings.ndim - 1)) + zero_flag = torch.ones(tensor_shape).to(embeddings.dtype) * (1 - drop_rate) + zero_flag = torch.bernoulli(zero_flag).to(embeddings.device) + embeddings = embeddings * zero_flag + return embeddings + + +def random_embed_replace(src_embed, tgt_embed, drop_rate, position=0): + r""" + Function to perform random embedding replacement. + With probability given by drop rate, we replace src_embed by tgt_embed + Args: + src_embed (tensor): Src embeddings + tgt_embed (tensor): Tgt embeddings + drop_rate (float): Rate of replacing the embedding. + position (int): Starting position to replace the sequence + """ + for i in range(src_embed.shape[0]): + coin_flip = torch.rand(1).item() + if coin_flip < drop_rate: + src_embed[i][position:] = tgt_embed + + return src_embed + + +def to255_round_uint8_append(vis_images, total_vis_images): + r""" + Map pixel values of vis_images to [0 255] and quantize them to 256 bins + """ + if vis_images is not None: + vis_images = ((vis_images + 1) / 2).clamp_(0, 1).mul_(255).round_().type(torch.uint8) + total_vis_images.append(vis_images) + return vis_images + + +def no_round_append(vis_images, total_vis_images): + r""" + Append the images as is without type casting + """ + if vis_images is not None: + total_vis_images.append(vis_images) + return vis_images + + +def sample_sigma_and_xt( + sde, + target_data: torch.Tensor, + data_batch: dict = None, + use_same_noise_multiview: bool = False, + use_low_noise_first_view: bool = False, +): + """Sample pertubation noise levels and generate noisy observations.""" + # Sample pertubation noise levels + tensor_kwargs = {"device": "cuda", "dtype": target_data.dtype} + t = sde.sample_t(batch_size=target_data.size()[0]).to(**tensor_kwargs) # check precision and memory_format later + if data_batch is not None and data_batch.get("num_view", None) is not None: + if use_same_noise_multiview and not use_low_noise_first_view: + t_shape = t.shape + t = t.view(-1, int(data_batch["num_view"].view(-1)[0].item())) + t[:, 1:] = t[:, 0:1] + t = t.view(t_shape) + elif use_low_noise_first_view and not use_same_noise_multiview: + t_shape = t.shape + t = t.view(-1, int(data_batch["num_view"].view(-1)[0].item())) + t[:, 0] = 0.02 + t = t.view(t_shape) + elif use_low_noise_first_view and use_same_noise_multiview: + t_shape = t.shape + t = t.view(-1, int(data_batch["num_view"].view(-1)[0].item())) + t[:, 0] = 0.02 + t[:, 2:] = t[:, 1:2] + t = t.view(t_shape) + # Generate an N(0,1) noise map. + epsilon = torch.randn_like(target_data, **tensor_kwargs) + # Get the mean and stand deviation of the marginal probability distribution. + mean, std = sde.marginal_prob(target_data, t) + # Generate noisy observations + xt = mean + batch_mul(std, epsilon) # corrupted data + + data_batch = {} + data_batch["t"] = t # between model.sde.eps to 1 + data_batch["epsilon"] = epsilon # Standard normal noise map + data_batch["mean"] = mean # mean of the marginal distribution + data_batch["std"] = std # std deviation of the marginal distribution + data_batch["xt"] = xt # corrupted data + data_batch["target"] = target_data + return data_batch + + +def form_loss_mask( + data_batch: dict, + x_shape: tuple, + dtype: torch.dtype, + device: torch.device, + loss_masking_cfg: dict = {"human_body_mask": 2, "human_face_mask": 4, "human_hand_mask": 4, "padding_mask": 0}, +) -> torch.Tensor: + r""" + Function to form a combined mask given several loss masks. + If there are overlapping region between multiple masks, we assign the max value to the + overlapping region. For the unmasked regions, we assign a value of 1. + However, if there is a mask specifying zero value, we zero it out. + Zeroing is crucial for padded loss. + Copied from i3, kaal.py form_loss_mask function. + zero_mask: mask out some region by setting them as zero + For example, + mask1: [0, 1, 1, 1, 0, 0], weight: 2 + mask2: [1, 0, 1, 0, 0, 0], weight: 4 + mask3: [0, 1, 0, 0, 0, 0], weight: 0 + + Final loss mask: [4, 0, 4, 2, 1, 1] + """ + loss_mask = torch.ones(x_shape, dtype=dtype, device=device) + zero_mask = torch.ones(x_shape, dtype=dtype, device=device) + + for key in loss_masking_cfg: + if key not in data_batch: + if loss_masking_cfg[key] > 0: + log.warning(f"You set {key} to have larger loss, but there is no such mask data") + continue + # Repeat mask along channel's dimension. ndim=4 for images. + repeat_dims = (1, 3) + tuple([1] * (data_batch[key].ndim - 2)) + mask_key = torch.tile(data_batch[key], dims=repeat_dims) + weight_key = loss_masking_cfg[key] + + assert weight_key >= 0, "Current support only for weight >= 0" + + if key == "zero_mask": + zero_mask = zero_mask * mask_key + elif weight_key == 0: + zero_mask = zero_mask * (1 - mask_key) + else: + no_mask_region = (mask_key == 0).float() + loss_mask = mask_key * weight_key + no_mask_region * loss_mask + # loss_mask = torch.max(loss_mask_new, loss_mask) + + loss_mask_final = loss_mask * zero_mask + return loss_mask_final diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/validator.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/validator.py new file mode 100644 index 0000000000000000000000000000000000000000..5d32d6f8cee464d51f8f073df6221bbeddbc9dd6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/validator.py @@ -0,0 +1,514 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import base64 +import binascii +import itertools +import json +import os +from abc import ABC, abstractmethod +from io import BytesIO +from typing import Any + +# Sentinel value to indicate that no default was explicitly set by the user +# we want to mimic usage of function parameters: if no default is provided, the parameter is mandatory +_UNSET = object() + + +# from https://docs.python.org/3/howto/descriptor.html#validator-class +# For usage of hidden flag see the ModelParams class in apis/utils/model_params.py + + +# validators can be customized to very specific needs, e.g. see HumanAttributes below +class Validator(ABC): + def __init__(self, default=_UNSET, hidden=False): + self.default = default + self.hidden = hidden + + # set name is called when the validator is created as class variable + # name is the name of the variable in the owner class, so here we create the name for the backing variable + def __set_name__(self, owner, name): + self.private_name = "_" + name + + def __get__(self, obj, objtype=None): + value = getattr(obj, self.private_name, self.default) + if value is _UNSET: + # If we reach here, it means a mandatory parameter was accessed without being set + attr_name = getattr(self, "private_name", "unknown").lstrip("_") + raise ValueError( + f"Parameter '{attr_name}' is mandatory but has not been set. " + f"No default value was provided and no value was assigned." + ) + return value + + def __set__(self, obj, value): + value = self.validate(value) + setattr(obj, self.private_name, value) + + @abstractmethod + def validate(self, value): + pass + + def json(self): + pass + + +class Bool(Validator): + def __init__(self, default=_UNSET, hidden=False, tooltip=None): + super().__init__(default, hidden) + self.default = default + self.hidden = hidden + self.tooltip = tooltip + + def validate(self, value): + if isinstance(value, int): + value = value != 0 + elif isinstance(value, str): + value = value.lower() + if value in ["true", "1"]: + value = True + elif value in ["false", "0"]: + value = False + else: + raise ValueError(f"Expected {value!r} to be one of ['True', 'False', '1', '0']") + elif not isinstance(value, bool): + raise TypeError(f"Expected {value!r} to be an bool") + + return value + + def get_range_iterator(self): + return [True, False] + + def __repr__(self) -> str: + return f"Bool({self.private_name=} {self.default=} {self.hidden=})" + + def json(self): + return { + "type": bool.__name__, + "default": self.default, + "tooltip": self.tooltip, + } + + +class Int(Validator): + def __init__(self, default=_UNSET, min=None, max=None, step=1, hidden=False, tooltip=None): + self.min = min + self.max = max + self.default = default + self.step = step + self.hidden = hidden + self.tooltip = tooltip + + def validate(self, value): + if isinstance(value, str): + value = int(value) + elif not isinstance(value, int): + raise TypeError(f"Expected {value!r} to be an int") + + if self.min is not None and value < self.min: + raise ValueError(f"Expected {value!r} to be at least {self.min!r}") + if self.max is not None and value > self.max: + raise ValueError(f"Expected {value!r} to be no more than {self.max!r}") + return value + + def get_range_iterator(self): + if self.default is _UNSET: + default_val = 0 + else: + default_val = int(self.default) if isinstance(self.default, (int, float, str)) else 0 + iter_min = self.min if self.min is not None else default_val + iter_max = self.max if self.max is not None else (default_val + 100) + return itertools.takewhile(lambda x: x <= iter_max, itertools.count(iter_min, self.step)) + + def __repr__(self) -> str: + return f"Int({self.private_name=} {self.default=}, {self.min=}, {self.max=} {self.hidden=})" + + def json(self): + return { + "type": int.__name__, + "default": self.default, + "min": self.min, + "max": self.max, + "step": self.step, + "tooltip": self.tooltip, + } + + +class Float(Validator): + def __init__(self, default=_UNSET, min=None, max=None, step=0.5, hidden=False, tooltip=None): + self.min = min + self.max = max + self.default = default + self.step = step + self.hidden = hidden + self.tooltip = tooltip + + def validate(self, value): + if isinstance(value, str) or isinstance(value, int): + value = float(value) + elif not isinstance(value, float): + raise TypeError(f"Expected {value!r} to be float") + + if self.min is not None and value < self.min: + raise ValueError(f"Expected {value!r} to be at least {self.min!r}") + if self.max is not None and value > self.max: + raise ValueError(f"Expected {value!r} to be no more than {self.max!r}") + return value + + def get_range_iterator(self): + if self.default is _UNSET: + default_val = 0.0 + else: + default_val = float(self.default) if isinstance(self.default, (int, float, str)) else 0.0 + iter_min = self.min if self.min is not None else default_val + iter_max = self.max if self.max is not None else (default_val + 100.0) + return itertools.takewhile(lambda x: x <= iter_max, itertools.count(iter_min, self.step)) + + def __repr__(self) -> str: + return f"Float({self.private_name=} {self.default=}, {self.min=}, {self.max=} {self.hidden=})" + + def json(self): + return { + "type": float.__name__, + "default": self.default, + "min": self.min, + "max": self.max, + "step": self.step, + "tooltip": self.tooltip, + } + + +class String(Validator): + def __init__(self, default=_UNSET, min=None, max=None, predicate=None, hidden=False, tooltip=None): + self.min = min + self.max = max + self.predicate = predicate + self.default = default + self.hidden = hidden + self.tooltip = tooltip + + def validate(self, value): + if value is None: + return value # Allow None as a valid value to be compatible with existing code + # this breaks strict typing, so do this only for strings + if not isinstance(value, str): + raise TypeError(f"Expected {value!r} to be an str or None") + if self.min is not None and len(value) < self.min: + raise ValueError(f"Expected {value!r} to be no smaller than {self.min!r}") + if self.max is not None and len(value) > self.max: + raise ValueError(f"Expected {value!r} to be no bigger than {self.max!r}") + if self.predicate is not None and not self.predicate(value): + raise ValueError(f"Expected {self.predicate} to be true for {value!r}") + return value + + def get_range_iterator(self): + return iter([self.default]) + + def __repr__(self) -> str: + return f"String({self.private_name=} {self.default=}, {self.min=}, {self.max=} {self.hidden=})" + + def json(self): + return { + "type": str.__name__, + "default": self.default, + "tooltip": self.tooltip, + } + + +class Path(Validator): + def __init__(self, default=_UNSET, hidden=False, tooltip=None): + self.default = default + self.hidden = hidden + self.tooltip = tooltip + + def validate(self, value): + if value is None: + return value + if not isinstance(value, str): + raise TypeError(f"{self.private_name} validator: Expected {value!r} to be an str") + if not os.path.exists(value): + raise ValueError(f"{self.private_name} validator: Expected {value!r} to be a valid path") + + return value + + def get_range_iterator(self): + return iter([self.default]) + + def __repr__(self) -> str: + return f"String({self.private_name=} {self.default=}, {self.hidden=})" + + +class InputImage(Validator): + def __init__( + self, default=_UNSET, hidden=False, tooltip=None, supported_formats=["jpeg", "jpg", "png", "bmp", "gif"] + ): + self.default = default + self.hidden = hidden + self.tooltip = tooltip + self.supported_formats = supported_formats + + def validate(self, value): + ext = os.path.splitext(value)[1].lower() + + if ext not in self.supported_formats: + raise ValueError(f"Unsupported image format: {ext}") + + if not isinstance(value, str): + raise TypeError(f"Expected {value!r} to be an str") + if not os.path.exists(value): + raise ValueError(f"Expected {value!r} to be a valid path") + return value + + def get_range_iterator(self): + return iter([self.default]) + + def __repr__(self) -> str: + return f"String({self.private_name=} {self.default=} {self.hidden=})" + + def json(self): + return { + "type": InputImage.__name__, + "default": self.default, + "values": self.supported_formats, + "tooltip": self.tooltip, + } + + +class JsonDict(Validator): + """ + JSON stringified version of a python dict. + Example: '{"ema_customization_iter.pt": "ema_customization_iter.pt"}' + """ + + def __init__(self, default=_UNSET, hidden=False): + self.default = default + self.hidden = hidden + + def validate(self, value): + if not value: + return {} + try: + dict = json.loads(value) + return dict + except json.JSONDecodeError as e: + raise ValueError(f"Expected {value!r} to be json stringified dict. Error: {str(e)}") + + def __repr__(self) -> str: + return f"Dict({self.default=} {self.hidden=})" + + +class Dict(Validator): + """ + Python dict. + Example: {'key': 'value'} + + This allows a single level of parameter nesting, but not a full nested dict. + For now we validate the individual keys here and store the dict as is. + Alternatively, we could have a validator that gets/sets another ValidatorParams class. + """ + + def __init__(self, default=_UNSET, hidden=False): + self.default = default + self.hidden = hidden + + def validate(self, value): + if not isinstance(value, dict): + raise TypeError(f"Expected {value!r} to be an dict") + return value + + def __repr__(self) -> str: + value = getattr(self, self.private_name, self.default) + + return f"Dict({self.private_name=} {self.default=} {self.hidden=} value={json.dumps(value, indent=4)})" + + +class OneOf(Validator): + def __init__(self, default=_UNSET, options=None, type_cast=None, hidden=False, tooltip=None): + self.options = set(options) if options is not None else set() + self.default = default + self.type_cast = type_cast # Cast the value to this type before checking if it's in options + self.tooltip = tooltip + self.hidden = hidden + + def validate(self, value): + if self.type_cast: + try: + value = self.type_cast(value) + except ValueError: + raise ValueError(f"Expected {value!r} to be castable to {self.type_cast!r}") + + if value not in self.options: + raise ValueError(f"Expected {value!r} to be one of {self.options!r}") + + return value + + def get_range_iterator(self): + return self.options + + def __repr__(self) -> str: + return f"OneOf({self.private_name=} {self.options=} {self.hidden=})" + + def json(self): + return { + "type": OneOf.__name__, + "default": self.default, + "values": list(self.options), + "tooltip": self.tooltip, + } + + +class MultipleOf(Validator): + def __init__(self, default=_UNSET, multiple_of: int = 1, type_cast=None, hidden=False, tooltip=None): + if type(multiple_of) is not int: + raise ValueError(f"Expected {multiple_of!r} to be an int") + self.multiple_of = multiple_of + self.default = default + self.type_cast = type_cast + + # For usage of hidden flag see the ModelParams class in apis/utils/model_params.py + # if a parameter is hidden then probe() can't expose the param + # and the param can't be set anymore + self.hidden = hidden + self.tooltip = tooltip + + def validate(self, value): + if self.type_cast: + try: + value = self.type_cast(value) + except ValueError: + raise ValueError(f"Expected {value!r} to be castable to {self.type_cast!r}") + + if value % self.multiple_of != 0: + raise ValueError(f"Expected {value!r} to be a multiple of {self.multiple_of!r}") + + return value + + def get_range_iterator(self): + return itertools.count(0, self.multiple_of) + + def __repr__(self) -> str: + return f"MultipleOf({self.private_name=} {self.multiple_of=} {self.hidden=})" + + def json(self): + return { + "type": MultipleOf.__name__, + "default": self.default, + "multiple_of": self.multiple_of, + "tooltip": self.tooltip, + } + + +class HumanAttributes(Validator): + def __init__(self, default=_UNSET, hidden=False, tooltip=None): + self.default = default + self.hidden = hidden + self.tooltip = tooltip + + # hard code the options for now + # we extend this to init parameter as needed + valid_attributes = { + "emotion": ["angry", "contemptful", "disgusted", "fearful", "happy", "neutral", "sad", "surprised"], + "race": ["asian", "indian", "black", "white", "middle eastern", "latino hispanic"], + "gender": ["male", "female"], + "age group": [ + "young", + "teen", + "adult early twenties", + "adult late twenties", + "adult early thirties", + "adult late thirties", + "adult middle aged", + "older adult", + ], + } + + def get_range_iterator(self): + # create a list of all possible combinations + l1 = self.valid_attributes["emotion"] + l2 = self.valid_attributes["race"] + l3 = self.valid_attributes["gender"] + l4 = self.valid_attributes["age group"] + all_combinations = list(itertools.product(l1, l2, l3, l4)) + return iter(all_combinations) + + def validate(self, value): + human_attributes = value.lower() + if human_attributes not in ["none", "random"]: + # In this case, we need for custom attribute string + + attr_string = human_attributes + for attr_key in ["emotion", "race", "gender", "age group"]: + attr_detected = False + for attr_label in self.valid_attributes[attr_key]: + if attr_string.startswith(attr_label): + attr_string = attr_string[len(attr_label) + 1 :] # noqa: E203 + attr_detected = True + break + + if attr_detected is False: + raise ValueError(f"Expected {value!r} to be one of {self.valid_attributes!r}") + + return value + + def __repr__(self) -> str: + return f"HumanAttributes({self.private_name=} {self.hidden=})" + + def json(self): + return { + "type": HumanAttributes.__name__, + "default": self.default, + "values": self.valid_attributes, + "tooltip": self.tooltip, + } + + +class BytesIOType(Validator): + """ + Validator class for BytesIO. Valid inputs are either: + - bytes + - objects of class BytesIO + - str which can be successfully decoded into BytesIO + """ + + def __init__(self, default=_UNSET, hidden=False, tooltip=None): + self.default = default + self.hidden = hidden + self.tooltip = tooltip + + def validate(self, value: Any) -> BytesIO: + if isinstance(value, str): + try: + # Decode the Base64 string + decoded_bytes = base64.b64decode(value) + # Create a BytesIO stream from the decoded bytes + return BytesIO(decoded_bytes) + except (binascii.Error, ValueError) as e: + raise ValueError(f"Invalid Base64 encoded string: {e}") + elif isinstance(value, bytes): + return BytesIO(value) + elif isinstance(value, BytesIO): + return value + else: + raise TypeError(f"Expected {value!r} to be a Base64 encoded string, bytes, or BytesIO") + + def __repr__(self) -> str: + return f"BytesIOValidator({self.default=}, {self.hidden=})" + + def json(self): + return { + "type": BytesIO.__name__, + "default": self.default, + "tooltip": self.tooltip, + } diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/validator_params.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/validator_params.py new file mode 100644 index 0000000000000000000000000000000000000000..b0fa48ef88e92ce35269aa1ed6ab4e18e5392b1e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/validator_params.py @@ -0,0 +1,184 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import json +import pprint +import shlex + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.validator import _UNSET, Validator + +""" +Base class for all model parameter classes. + +The primary purpose is to fully validate any input parameter including type, range, etc. +By using custom validators, we can additionally validate complex parameters such as images, text, etc. +Additioally, the class can parse command line arguments into a dictionary of parameters +and create a model parameter class from a dictionary of parameters. + +if default of a validator is _UNSET, the parameter is mandatory and must be provided by the user. +Hence validators without explicit defaults require user input. +""" + + +class ValidatorParams: + """ + factory method to create a model params class from a given api and a dictionary of args + in comparison to createFromCmd, the server can first parse and modify some args, + finally use this factory method to create the model params + """ + + @classmethod + def create(cls, kwargs): + instance = cls() + log.info(f"creating model params class={cls}") + instance.from_kwargs(kwargs) + + val_dict = cls.get_val_dict() + + for key, validator in val_dict.items(): + # Check if validator has no user-provided default (_UNSET) and no value was assigned + if validator.default is _UNSET: + value = getattr(instance, key, _UNSET) + if value is _UNSET: + raise ValueError( + f"mandatory parameter {key} is missing - no default provided and no value assigned by user" + ) + + return instance + + """ + factory method to create a model params class from a command string + """ + + @classmethod + def createFromCmd(cls, cmd: str) -> object: + kwargs = cls.parse(cmd) + return cls.create(kwargs) + + def from_kwargs(self, kwargs): + # most attributes of this class are validators, + # but dervied class could add non-validators + # or some validators might be hidden + # therefore only allow exposed params to be set + for key, value in kwargs.items(): + if key in self.get_exposed_params(): + setattr(self, key, value) + else: + raise ValueError(f"unknown parameter {key} in command line") + + def to_kwargs(self) -> dict: + """for a given config return a dictionary of all the parameters and their values""" + param_names = self.get_exposed_params() + return {key: getattr(self, key) for key in param_names} + + @classmethod + def validate_kwargs(cls, kwargs) -> dict: + """validate a dictionary of args and return the validated dictionary""" + instance = cls.create(kwargs) + return instance.to_kwargs() + + @staticmethod + def parse(cmd: str) -> dict: + """parse a command string into an api command (e.g. text2image) and a dictionary of args""" + args = {} + pairs = shlex.split(cmd) + + for arg in pairs: + key, value = arg.split("=", 1) # Split only on the first '=' + value = value.strip().strip("'") + key = key.strip("--") + args[key.strip()] = value + + log.debug(f"parsed cmd-line: {args}") + return args + + @classmethod + def probe(cls) -> list[str]: + params = cls.get_exposed_params() + log.info(f"exposed params for {cls}: {params}") + return params + + """ + extened version of probe will query from each validator extended information. + This will include default parameters, min, max, step, etc. + """ + + @classmethod + def probe_ex(cls) -> dict: + validator_dict = cls.get_val_dict() + + parameter_info = {key: value.json() for key, value in validator_dict.items() if not value.hidden} + log.info(f"exposed params for {cls}: {json.dumps(parameter_info, indent=4)}") + return parameter_info + + # a model parameter class can also have non exposed parameters: + # we can hide parameters as needed from public API (compare to former exposed_params list in yaml configs in imaginaire3) + # class can also have non-validator attributes + @classmethod + def get_exposed_params(cls) -> list[str]: + # log.debug(f"getting exposed params of {cls.__name__}") + + # the exposed params are repeatedly used for parsing so we cache them + # note that we are caching the exposed params per class in the class hierarchy! + # each class has its own set of exposed params. + # instances of the class will have the same set of exposed params + if "_exposed_params" not in cls.__dict__: + # log.debug(f"creating cache exposed params of {cls.__name__}") + validator_dict = cls.get_val_dict() + + # if a parameter is hidden then probe() can't expose the param + # and the param can't be set anymore + cls._exposed_params = [key for key, value in validator_dict.items() if not value.hidden] + return cls._exposed_params + + def exposed_params_dict(self): + keys = self.get_exposed_params() + out_dict = {key: getattr(self, key) for key in keys} + return out_dict + + """ + returns a dictionary of all validators in the class hierarchy, e.g. for a string validator: + + prompt_validator = String() + + so prompt_validator is the instance of the String validator. the dictionary will be: + + {'prompt_validator': prompt_validator} + """ + + @classmethod + def get_val_dict(cls) -> dict[str, Validator]: + # log.debug(f"getting val dict of {cls.__name__}") + val_dict = {} + if cls is not ValidatorParams: + val_dict.update(cls.__bases__[0].get_val_dict()) + + val_dict.update({key: value for key, value in cls.__dict__.items() if isinstance(value, Validator)}) + + return val_dict + + @classmethod + def debug_print(cls): + pp = pprint.PrettyPrinter(indent=4) + print(f"*********** validator dict for {cls.__name__} ***********") + val_dict = cls.get_val_dict() + pp.pprint(val_dict) + + def __str__(self): + return ", ".join(f"{key}={value}" for key, value in self.__dict__.items()) + + def __repr__(self): + return ", ".join(f"{key}={value}" for key, value in self.__dict__.items()) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/validator_test.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/validator_test.py new file mode 100644 index 0000000000000000000000000000000000000000..f1245108067e9a4eb9c8d037be714dd7cd3e9efb --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/validator_test.py @@ -0,0 +1,166 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import pytest + +from cosmos_policy._src.imaginaire.utils.validator import Bool, Dict, Float, Int, OneOf, String +from cosmos_policy._src.imaginaire.utils.validator_params import ValidatorParams + +param_dict = { + "prompt": "a cat", + "num_samples": 2, + "guidance": 6.5, + "media_type": "illustration", +} + + +class SampleParams(ValidatorParams): + """All the required values to generate image from text at a given resolution.""" + + # no default, so it is mandatory + prompt = String() + + negative_prompt = String( + "ugly, blurry, childish, flat, malformed, poorly drawn, old, dated, 80s, 90s, photoshop, post process, " + "collage, fake" + ) + num_samples = Int(4, min=1, max=4) + media_type = OneOf("photography", ["photography", "illustration", "film"]) + guidance = Float(7.5, min=5, max=10) + + +class ChildParams(SampleParams): + """Child class might want to remove some parameters if certain model paramters shouldn't be exposed to the user.""" + + num_samples = Int(1, hidden=True) + nsfw_flag = Bool(False) + + +nested_dict_param = { + "dict_param": { + "path": "some/file/path", + }, +} + + +class ParamsWithNesting(ValidatorParams): + """Test dict parameter.""" + + seed = Int(0) + dict_param = Dict(default={}) + + +@pytest.mark.L0 +def test_from_kwargs(): + params = SampleParams.create(param_dict) + assert params.prompt == "a cat" + assert params.num_samples == 2 + assert params.guidance == 6.5 + assert params.media_type == "illustration" + print("✅ test_from_kwargs: PASSED") + + +@pytest.mark.L0 +def test_from_kwargs_dict(): + params = ParamsWithNesting.create(nested_dict_param) + assert params.dict_param == nested_dict_param["dict_param"] + assert params.dict_param["path"] == nested_dict_param["dict_param"]["path"] + params.debug_print() + print(params) + print("✅ test_from_kwargs_dict: PASSED") + + +@pytest.mark.L0 +def test_from_cmd_line(): + # input is the legacy command line format + cmd = "--prompt='cat' --num_samples=1 --guidance=5.5 --media_type='illustration'" + params = SampleParams.createFromCmd(cmd) + params.debug_print() + # access descriptors same as regular variables + assert params.num_samples == 1 + assert params.guidance == 5.5 + print("✅ test_from_cmd_line: PASSED") + + +@pytest.mark.L0 +def test_unknown_parameter(): + cmd = "--prompt='cat' --unkown_param=1" + with pytest.raises(ValueError): + params_config = ValidatorParams.createFromCmd(cmd) + print("✅ test_unknown_parameter: PASSED") + + +@pytest.mark.L0 +def test_no_default(): + with pytest.raises(ValueError): + params = SampleParams.create({}) + print("✅ test_no_default: PASSED") + + +@pytest.mark.L0 +def test_freeze(): + test_config = SampleParams() + # todo the class dict isn't frozen + # so unfortunately following code is still allowed + test_config.nsamplesss = "some_param" + print("⚠️ test_freeze: PASSED (freezing not fully implemented yet)") + + +@pytest.mark.L0 +def test_out_of_range(): + cmd = "--prompt='cat' --human_attributes='out of range param'" + with pytest.raises(ValueError): + params_config = ValidatorParams.createFromCmd(cmd) + print("✅ test_out_of_range: PASSED") + + +@pytest.mark.L0 +def test_range_iter(): + range_config = SampleParams() + range_config.prompt = "a fat cat" + val_dict = range_config.get_val_dict() + descriptors = list(val_dict.values()) + + # simple test over parameter ranges (not permutation of all parameters) + # we interate over the value range of each parameter while keeping the rest at default + iteration_count = 0 + for desc in descriptors: + param_iterations = 0 + for i in desc.get_range_iterator(): + setattr(range_config, desc.private_name, i) + param_iterations += 1 + iteration_count += 1 + # Limit iterations to prevent excessive output + if param_iterations > 5: + break + # todo run test with upated values + # t2i(range_config) + setattr(range_config, desc.private_name, desc.default) + + print(f"✅ test_range_iter: PASSED ({iteration_count} iterations)") + + +if __name__ == "__main__": + print("🚀 Running validator tests...\n") + test_no_default() + test_from_kwargs_dict() + + # test_from_kwargs() + # test_from_cmd_line() + # test_probe() + # test_unknown_parameter() + # test_freeze() + # test_out_of_range() + # test_range_iter() diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/utils/wandb_util.py b/REGEN-main/cosmos_policy/_src/imaginaire/utils/wandb_util.py new file mode 100644 index 0000000000000000000000000000000000000000..4d39c97dbdaf3dc96750105cfd8430b36a5b70d6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/utils/wandb_util.py @@ -0,0 +1,118 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import os +from typing import TYPE_CHECKING + +import attrs +import wandb +import wandb.util +from omegaconf import DictConfig + +from cosmos_policy._src.imaginaire.lazy_config.lazy import LazyConfig +from cosmos_policy._src.imaginaire.utils import distributed, log, object_store +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + +if TYPE_CHECKING: + from cosmos_policy._src.imaginaire.config import CheckpointConfig, Config, JobConfig + from cosmos_policy._src.imaginaire.model import ImaginaireModel + + +@distributed.rank0_only +def init_wandb(config: Config, model: ImaginaireModel) -> None: + """Initialize Weights & Biases (wandb) logger. + + Args: + config (Config): The config object for the Imaginaire codebase. + model (ImaginaireModel): The PyTorch model. + """ + if isinstance(config.job, DictConfig): + from cosmos_policy._src.imaginaire.config import JobConfig + + config_job = JobConfig(**config.job) + else: + config_job = config.job + config_checkpoint = config.checkpoint + # Try to fetch the W&B job ID for resuming training. + wandb_id = _read_wandb_id(config_job, config_checkpoint) + if wandb_id is None: + # Generate a new W&B job ID. + wandb_id = wandb.util.generate_id() + _write_wandb_id(config_job, config_checkpoint, wandb_id=wandb_id) + log.info(f"Generating new wandb ID: {wandb_id}") + else: + log.info(f"Resuming with existing wandb ID: {wandb_id}") + # refactor config so that wandb better understands it + local_safe_yaml_fp = LazyConfig.save_yaml(config, os.path.join(config_job.path_local, "config.yaml")) + if os.path.exists(local_safe_yaml_fp): + config_resolved = easy_io.load(local_safe_yaml_fp) + else: + config_resolved = attrs.asdict(config) + # Initialize the wandb library. + wandb.init( + force=True, + id=wandb_id, + project=config_job.project, + group=config_job.group, + name=config_job.name, + config=config_resolved, + dir=config_job.path_local, + resume="allow", + mode=config_job.wandb_mode, + ) + + +def _read_wandb_id(config_job: JobConfig, config_checkpoint: CheckpointConfig) -> str | None: + """Read the W&B job ID. If it doesn't exist, return None. + + Args: + config_wandb (JobConfig): The config object for the W&B logger. + config_checkpoint (CheckpointConfig): The config object for the checkpointer. + + Returns: + wandb_id (str | None): W&B job ID. + """ + wandb_id = None + if config_checkpoint.load_from_object_store.enabled: + object_store_loader = object_store.ObjectStore(config_checkpoint.load_from_object_store) + wandb_id_path = f"{config_job.path}/wandb_id.txt" + if object_store_loader.object_exists(key=wandb_id_path): + wandb_id = object_store_loader.load_object(key=wandb_id_path, type="text").strip() + else: + wandb_id_path = f"{config_job.path_local}/wandb_id.txt" + if os.path.isfile(wandb_id_path): + wandb_id = open(wandb_id_path).read().strip() + return wandb_id + + +def _write_wandb_id(config_job: JobConfig, config_checkpoint: CheckpointConfig, wandb_id: str) -> None: + """Write the generated W&B job ID. + + Args: + config_wandb (JobConfig): The config object for the W&B logger. + config_checkpoint (CheckpointConfig): The config object for the checkpointer. + wandb_id (str): The W&B job ID. + """ + content = f"{wandb_id}\n" + if config_checkpoint.save_to_object_store.enabled: + object_store_saver = object_store.ObjectStore(config_checkpoint.save_to_object_store) + wandb_id_path = f"{config_job.path}/wandb_id.txt" + object_store_saver.save_object(content, key=wandb_id_path, type="text") + else: + wandb_id_path = f"{config_job.path_local}/wandb_id.txt" + with open(wandb_id_path, "w") as file: + file.write(content) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/visualize/__init__.py b/REGEN-main/cosmos_policy/_src/imaginaire/visualize/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..ce4bac789b462bb83bade1389bf3aea5c7d50fe0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/visualize/__init__.py @@ -0,0 +1,16 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from cosmos_policy._src.imaginaire.visualize.img import save_batch_img, show_batch_img # noqa: F401 diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/visualize/img.py b/REGEN-main/cosmos_policy/_src/imaginaire/visualize/img.py new file mode 100644 index 0000000000000000000000000000000000000000..8b1a988e1cacd78090715fa7cf663d1eeb216858 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/visualize/img.py @@ -0,0 +1,148 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""image visualization utilities. + +based on https://gitlab.com/qsh.zh/jam/-/blob/master/jamviz/img.py MIT License +""" + +import os +from typing import Union + +import matplotlib.pyplot as plt +import numpy as np +import torch +from einops import rearrange +from PIL import Image +from torchvision.utils import make_grid + +__all__ = [ + "show_batch_img", + "save_batch_img", +] + + +def _reshape_viz_batch_img(img_data: torch.Tensor | np.ndarray, shape: int | str = 7) -> tuple: + """ + Reshapes a batch of images for visualization, organizing them into a grid format. + + Args: + img_data (torch.Tensor | np.ndarray): The image data to be reshaped, can be either a PyTorch tensor or a NumPy array. + shape (int | str, optional): Defines the layout of the grid. If an integer is provided, it specifies both the number of rows and columns. If a string is provided in the format 'nrowxncol', it parses to individual row and column numbers. Defaults to 7. + + Returns: + tuple: A tuple containing: + img (np.ndarray | torch.Tensor): The image data arranged in grid format. + nrow (int): Number of rows in the grid. + ncol (int): Number of columns in the grid. + + Raises: + RuntimeError: If the shape parameter is neither an int nor a string, or if it's a string that doesn't contain 'x'. + + Example: + >>> tensor_images = torch.rand(64, 3, 28, 28) # Example tensor of 64 images + >>> img_grid, rows, cols = _reshape_viz_batch_img(tensor_images, '8x8') + >>> img_grid.shape + (224, 224, 3) + """ + if isinstance(shape, int): + nrow, ncol = shape, shape + elif isinstance(shape, str): + if "x" not in shape: + nrow, ncol = int(shape), int(shape) + else: + shape = shape.split("x") + nrow, ncol = int(shape[0]), int(shape[1]) + else: + raise RuntimeError(f"shape {shape} not support") + if isinstance(img_data, torch.Tensor): + assert img_data.shape[1] in [1, 3] + grid_img = make_grid(img_data[: nrow * ncol].detach().cpu(), ncol) + img = grid_img.permute(1, 2, 0) + elif isinstance(img_data, np.ndarray): + if img_data.shape[1] in [1, 3]: + img = rearrange(img_data[: nrow * ncol], "(b t) c h w -> (b h) (t w) c", b=nrow) + else: + img = rearrange(img_data[: nrow * ncol], "(b t) h w c -> (b h) (t w) c", b=nrow) + return img, nrow, ncol + + +def show_batch_img( + img_data: torch.Tensor | np.ndarray, + shape: int | str = 7, + grid: int = 3, + is_n1p1: bool = False, + auto_n1p1: bool = True, +) -> None: + """ + Displays a batch of images using matplotlib after arranging them into a specified grid layout. + + Args: + img_data (torch.Tensor | np.ndarray): The image data to be displayed. + shape (int | str, optional): The grid shape to organize the images. Defaults to 7. + grid (int, optional): Scaling factor for each image in the grid, affecting the overall size of the displayed figure. Defaults to 3. + is_n1p1 (bool, optional): Whether to normalize the images from [-1, 1] to [0, 1] for visualization. Defaults to False. + auto_n1p1 (bool, optional): If true, automatically adjusts images from [-1, 1] to [0, 1] based on minimum pixel value detection. Defaults to True. + + Returns: + None: This function does not return anything but displays the image grid using matplotlib. + + Example: + >>> tensor_images = torch.rand(64, 3, 28, 28) # Example tensor of 64 images + >>> show_batch_img(tensor_images, '8x8') + """ + if is_n1p1: + img_data = (img_data + 1) / 2 + else: + if auto_n1p1: + if isinstance(img_data, torch.Tensor): + if img_data.min().item() < -0.5: + img_data = (img_data + 1) / 2 + elif isinstance(img_data, np.ndarray): + if np.min(img_data) < -0.5: + img_data = (img_data + 1) / 2 + img, nrow, ncol = _reshape_viz_batch_img(img_data, shape) + plt.figure(figsize=(ncol * grid, nrow * grid)) + plt.axis("off") + plt.imshow(img) + + +def save_batch_img(fpath: str, img_data: Union[torch.Tensor, np.ndarray], shape: Union[int, str] = 7) -> None: + """ + Saves a batch of images to a file after arranging them into a grid format. Handles both PyTorch tensors and NumPy arrays as input. + + Args: + fpath (str): File path where the image will be saved. + img_data (Union[torch.Tensor, np.ndarray]): The image data to be saved. Can be a PyTorch tensor or a NumPy array. + shape (Union[int, str], optional): The grid shape to organize the images. Can be an integer specifying equal number of rows and columns, or a string specifying 'nrowxncol'. Defaults to 7. + + Returns: + None: This function does not return anything but saves the image to the specified file path. + + Raises: + RuntimeError: If the input shape is neither an integer nor a string, or it does not include 'x' when provided as a string. + + Example: + >>> tensor_images = torch.rand(64, 3, 28, 28) # Example tensor of 64 images + >>> save_batch_img('path/to/save/image.png', tensor_images, '8x8') + # This saves the image grid to 'path/to/save/image.png' + """ + img, _, _ = _reshape_viz_batch_img(img_data, shape) + if isinstance(img, np.ndarray): + img = torch.from_numpy(img) + ndarr = img.mul(255).add_(0.5).clamp_(0, 255).to("cpu", torch.uint8).numpy() + im = Image.fromarray(ndarr) + os.makedirs(os.path.dirname(fpath), exist_ok=True) + im.save(fpath) diff --git a/REGEN-main/cosmos_policy/_src/imaginaire/visualize/video.py b/REGEN-main/cosmos_policy/_src/imaginaire/visualize/video.py new file mode 100644 index 0000000000000000000000000000000000000000..fe0a6992316dda7fb685ca0a540d272760e950ca --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/imaginaire/visualize/video.py @@ -0,0 +1,97 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import IO, Any, Union + +import cv2 +import numpy as np +import torch +from einops import rearrange +from PIL import Image as PILImage +from torch import Tensor + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + +try: + import ffmpegcv +except Exception as e: # ImportError cannot catch all problems + log.info(e) + ffmpegcv = None + + +def save_video(grid, video_name, fps=30): + grid = (grid * 255).astype(np.uint8) + grid = np.transpose(grid, (1, 2, 3, 0)) + with ffmpegcv.VideoWriter(video_name, "h264", fps) as writer: + for frame in grid: + frame = cv2.cvtColor(frame, cv2.COLOR_RGB2BGR) + + writer.write(frame) + + +def save_img_or_video( + sample_C_T_H_W_in01: Tensor, save_fp_wo_ext: Union[str, IO[Any]], fps: int = 24, quality=None, ffmpeg_params=None +) -> None: + """ + Save a tensor as an image or video file based on shape + + Args: + sample_C_T_H_W_in01 (Tensor): Input tensor with shape (C, T, H, W) in [0, 1] range. + save_fp_wo_ext (Union[str, IO[Any]]): File path without extension or file-like object. + fps (int): Frames per second for video. Default is 24. + """ + assert sample_C_T_H_W_in01.ndim == 4, "Only support 4D tensor" + assert isinstance(save_fp_wo_ext, str) or hasattr(save_fp_wo_ext, "write"), ( + "save_fp_wo_ext must be a string or file-like object" + ) + + if torch.is_floating_point(sample_C_T_H_W_in01): + sample_C_T_H_W_in01 = sample_C_T_H_W_in01.clamp(0, 1) + else: + assert sample_C_T_H_W_in01.dtype == torch.uint8, "Only support uint8 tensor" + sample_C_T_H_W_in01 = sample_C_T_H_W_in01.float().div(255) + + kwargs = {} + if quality is not None: + kwargs["quality"] = quality + if ffmpeg_params is not None: + kwargs["ffmpeg_params"] = ffmpeg_params + + if sample_C_T_H_W_in01.shape[1] == 1: + save_obj = PILImage.fromarray( + rearrange((sample_C_T_H_W_in01.cpu().float().numpy() * 255), "c 1 h w -> h w c").astype(np.uint8), + mode="RGB", + ) + ext = ".jpg" if isinstance(save_fp_wo_ext, str) else "" + easy_io.dump( + save_obj, + f"{save_fp_wo_ext}{ext}" if isinstance(save_fp_wo_ext, str) else save_fp_wo_ext, + file_format="jpg", + format="JPEG", + quality=85, + **kwargs, + ) + else: + save_obj = rearrange((sample_C_T_H_W_in01.cpu().float().numpy() * 255), "c t h w -> t h w c").astype(np.uint8) + ext = ".mp4" if isinstance(save_fp_wo_ext, str) else "" + easy_io.dump( + save_obj, + f"{save_fp_wo_ext}{ext}" if isinstance(save_fp_wo_ext, str) else save_fp_wo_ext, + file_format="mp4", + format="mp4", + fps=fps, + **kwargs, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/conditioner.py b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/conditioner.py new file mode 100644 index 0000000000000000000000000000000000000000..03650a9ea218353d7d666c80b4086a4e31ddb0c4 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/conditioner.py @@ -0,0 +1,302 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import random +from dataclasses import dataclass +from typing import Dict, Optional + +import torch +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.utils.context_parallel import broadcast_split_tensor +from cosmos_policy._src.predict2.conditioner import ( + BooleanFlag, + GeneralConditioner, + ReMapkey, + Text2WorldCondition, + TextAttr, +) + + +@dataclass(frozen=True) +class Video2WorldCondition(Text2WorldCondition): + use_video_condition: bool = False + # the following two attributes are used to set the video condition; during training, inference + gt_frames: Optional[torch.Tensor] = None + condition_video_input_mask_B_C_T_H_W: Optional[torch.Tensor] = None + + def set_video_condition( + self, + gt_frames: torch.Tensor, + random_min_num_conditional_frames: int, + random_max_num_conditional_frames: int, + num_conditional_frames: Optional[int] = None, + conditional_frames_probs: Optional[Dict[int, float]] = None, + ) -> "Video2WorldCondition": + """ + Sets the video conditioning frames for video-to-video generation. + + This method creates a conditioning mask for the input video frames that determines + which frames will be used as context frames for generating new frames. The method + handles both image batches (T=1) and video batches (T>1) differently. + + Args: + gt_frames: A tensor of ground truth frames with shape [B, C, T, H, W], where: + B = batch size + C = number of channels + T = number of frames + H = height + W = width + + random_min_num_conditional_frames: Minimum number of frames to use for conditioning + when randomly selecting a number of conditioning frames. + + random_max_num_conditional_frames: Maximum number of frames to use for conditioning + when randomly selecting a number of conditioning frames. + + num_conditional_frames: Optional; If provided, all examples in the batch will use + exactly this many frames for conditioning. If None, a random number of frames + between random_min_num_conditional_frames and random_max_num_conditional_frames + will be selected for each example in the batch. + + conditional_frames_probs: Optional; Dictionary mapping number of frames to probabilities. + If provided, overrides the random_min/max_num_conditional_frames with weighted sampling. + Example: {0: 0.5, 1: 0.25, 2: 0.25} for 50% chance of 0 frames, 25% for 1, 25% for 2. + + Returns: + A new Video2WorldCondition object with the gt_frames and conditioning mask set. + The conditioning mask (condition_video_input_mask_B_C_T_H_W) is a binary tensor + of shape [B, 1, T, H, W] where 1 indicates frames used for conditioning and 0 + indicates frames to be generated. + + Notes: + - For image batches (T=1), no conditioning frames are used (num_conditional_frames_B = 0). + - For video batches: + - If num_conditional_frames is provided, all examples use that fixed number of frames. + - Otherwise, each example randomly uses between random_min_num_conditional_frames and + random_max_num_conditional_frames frames. + - The mask marks the first N frames as conditioning frames (set to 1) for each example. + """ + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = gt_frames + + # condition_video_input_mask_B_C_T_H_W + B, _, T, H, W = gt_frames.shape + condition_video_input_mask_B_C_T_H_W = torch.zeros( + B, 1, T, H, W, dtype=gt_frames.dtype, device=gt_frames.device + ) + if T == 1: # handle image batch + num_conditional_frames_B = torch.zeros(B, dtype=torch.int32) + else: # handle video batch + if num_conditional_frames is not None: + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames + elif conditional_frames_probs is not None: + # Use weighted sampling based on provided probabilities + frames_options = list(conditional_frames_probs.keys()) + weights = list(conditional_frames_probs.values()) + num_conditional_frames_B = torch.tensor( + random.choices(frames_options, weights=weights, k=B), dtype=torch.int32 + ) + else: + num_conditional_frames_B = torch.randint( + random_min_num_conditional_frames, random_max_num_conditional_frames + 1, size=(B,) + ) + for idx in range(B): + condition_video_input_mask_B_C_T_H_W[idx, :, : num_conditional_frames_B[idx], :, :] += 1 + + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + return type(self)(**kwargs) + + def edit_for_inference( + self, is_cfg_conditional: bool = True, num_conditional_frames: int = 1 + ) -> "Video2WorldCondition": + _condition = self.set_video_condition( + gt_frames=self.gt_frames, + random_min_num_conditional_frames=0, + random_max_num_conditional_frames=0, + num_conditional_frames=num_conditional_frames, + ) + if not is_cfg_conditional: + # Do not use classifier free guidance on conditional frames. + # YB found that it leads to worse results. + _condition.use_video_condition.fill_(True) + return _condition + + def broadcast(self, process_group: torch.distributed.ProcessGroup) -> "Video2WorldCondition": + if self.is_broadcasted: + return self + # extra efforts + gt_frames = self.gt_frames + condition_video_input_mask_B_C_T_H_W = self.condition_video_input_mask_B_C_T_H_W + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = None + kwargs["condition_video_input_mask_B_C_T_H_W"] = None + new_condition = Text2WorldCondition.broadcast( + type(self)(**kwargs), + process_group, + ) + + kwargs = new_condition.to_dict(skip_underscore=False) + _, _, T, _, _ = gt_frames.shape + if process_group is not None: + if T > 1 and process_group.size() > 1: + gt_frames = broadcast_split_tensor(gt_frames, seq_dim=2, process_group=process_group) + condition_video_input_mask_B_C_T_H_W = broadcast_split_tensor( + condition_video_input_mask_B_C_T_H_W, seq_dim=2, process_group=process_group + ) + kwargs["gt_frames"] = gt_frames + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + return type(self)(**kwargs) + + +class Video2WorldConditionV2(Video2WorldCondition): + """ + compared to Video2WorldCondition, this class apply zero frames when use_video_condition is False~(unconditional generation in cfg) + in the case, we do zero-out conditional frames in the video condition + """ + + def set_video_condition( + self, + gt_frames: torch.Tensor, + random_min_num_conditional_frames: int, + random_max_num_conditional_frames: int, + num_conditional_frames: Optional[int] = None, + ) -> "Video2WorldConditionV2": + num_conditional_frames = 0 if not self.use_video_condition else num_conditional_frames + return super().set_video_condition( + gt_frames=gt_frames, + random_min_num_conditional_frames=random_min_num_conditional_frames, + random_max_num_conditional_frames=random_max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + ) + + def edit_for_inference( + self, is_cfg_conditional: bool = True, num_conditional_frames: int = 1 + ) -> "Video2WorldConditionV2": + del is_cfg_conditional + _condition = super().set_video_condition( + gt_frames=self.gt_frames, + random_min_num_conditional_frames=0, + random_max_num_conditional_frames=0, + num_conditional_frames=num_conditional_frames, + ) + return _condition + + +class Video2WorldConditioner(GeneralConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> Video2WorldCondition: + output = super()._forward(batch, override_dropout_rate) + return Video2WorldCondition(**output) + + +class Video2WorldConditionerV2(GeneralConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> Video2WorldConditionV2: + output = super()._forward(batch, override_dropout_rate) + return Video2WorldConditionV2(**output) + + +@dataclass(frozen=True) +class ActionConditionedCondition(Video2WorldCondition): + action: Optional[torch.Tensor] = None + + +class ActionConditionedConditioner(Video2WorldConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> ActionConditionedCondition: + output = super()._forward(batch, override_dropout_rate) + assert "action" in batch, "ActionConditionalConditioner requires 'action' in batch" + output["action"] = batch["action"] + return ActionConditionedCondition(**output) + + +_SHARED_CONFIG = dict( + fps=L(ReMapkey)( + input_key="fps", + output_key="fps", + dropout_rate=0.0, + dtype=None, + ), + padding_mask=L(ReMapkey)( + input_key="padding_mask", + output_key="padding_mask", + dropout_rate=0.0, + dtype=None, + ), + text=L(TextAttr)( + input_key=["t5_text_embeddings"], + dropout_rate=0.2, + use_empty_string=False, + ), + use_video_condition=L(BooleanFlag)( + input_key="fps", + output_key="use_video_condition", + dropout_rate=0.2, + ), +) + +VideoPredictionConditioner: LazyDict = L(Video2WorldConditioner)( + **_SHARED_CONFIG, +) + +VideoPredictionConditionerV2: LazyDict = L(Video2WorldConditionerV2)( + **_SHARED_CONFIG, +) + +ActionConditionedConditionerConfig: LazyDict = L(ActionConditionedConditioner)( + **_SHARED_CONFIG, + action=L(ReMapkey)( + input_key="action", + output_key="action", + dropout_rate=0.0, + dtype=None, + ), +) + + +def register_conditioner(): + cs = ConfigStore.instance() + cs.store( + group="conditioner", + package="model.config.conditioner", + name="video_prediction_conditioner", + node=VideoPredictionConditioner, + ) + + cs.store( + group="conditioner", + package="model.config.conditioner", + name="video_prediction_conditioner_v2", + node=VideoPredictionConditionerV2, + ) + + cs.store( + group="conditioner", + package="model.config.conditioner", + name="action_conditioned_video_conditioner", + node=ActionConditionedConditionerConfig, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/config.py b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/config.py new file mode 100644 index 0000000000000000000000000000000000000000..ea76be7dedc5b6c023441faef47e0d6a7fdb8a88 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/config.py @@ -0,0 +1,106 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any, List + +import attrs + +from cosmos_policy._src.imaginaire import config +from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer as Trainer +from cosmos_policy._src.imaginaire.utils.config_helper import import_all_modules_from_package +from cosmos_policy._src.predict2.action.configs.action_conditioned.conditioner import register_conditioner +from cosmos_policy._src.predict2.action.configs.action_conditioned.data import register_training_and_val_data +from cosmos_policy._src.predict2.action.configs.action_conditioned.model import register_model +from cosmos_policy._src.predict2.action.configs.action_conditioned.net import register_net +from cosmos_policy._src.predict2.configs.common.defaults.checkpoint import register_checkpoint +from cosmos_policy._src.predict2.configs.common.defaults.ckpt_type import register_ckpt_type +from cosmos_policy._src.predict2.configs.common.defaults.ema import register_ema +from cosmos_policy._src.predict2.configs.common.defaults.optimizer import register_optimizer +from cosmos_policy._src.predict2.configs.common.defaults.scheduler import register_scheduler +from cosmos_policy._src.predict2.configs.common.defaults.tokenizer import register_tokenizer +from cosmos_policy._src.predict2.configs.video2world.defaults.callbacks import register_callbacks + + +@attrs.define(slots=False) +class Config(config.Config): + # default config groups that will be used unless overwritten + # see config groups in registry.py + defaults: List[Any] = attrs.field( + factory=lambda: [ + "_self_", + {"data_train": "mock"}, + {"data_val": "mock"}, + {"optimizer": "fusedadamw"}, + {"scheduler": "lambdalinear"}, + {"model": "action_conditioned_video2world_fsdp_rectified_flow"}, + {"callbacks": "basic"}, + {"net": None}, + {"conditioner": "video_prediction_conditioner"}, + {"ema": "power"}, + {"tokenizer": "wan2pt2_tokenizer"}, + {"checkpoint": "s3"}, + {"ckpt_type": "dummy"}, + # the list is with order, we need global experiment to be the last one + {"experiment": None}, + ] + ) + + +def make_config() -> Config: + c = Config( + model=None, + optimizer=None, + scheduler=None, + dataloader_train=None, + dataloader_val=None, + ) + + # Specifying values through instances of attrs + c.job.project = "cosmos_diffusion_v2" + c.job.group = "debug" + c.job.name = "delete_${now:%Y-%m-%d}_${now:%H-%M-%S}" + + c.trainer.type = Trainer + c.trainer.straggler_detection.enabled = False + c.trainer.max_iter = 400_000 + c.trainer.logging_iter = 10 + c.trainer.validation_iter = 100 + c.trainer.run_validation = False + c.trainer.callbacks = None + + # Call this function to register config groups for advanced overriding. the order follows the default config groups + register_optimizer() + register_scheduler() + register_model() + register_callbacks() + register_ema() + register_tokenizer() + register_checkpoint() + register_ckpt_type() + + register_training_and_val_data() + + register_net() + register_conditioner() + + import_all_modules_from_package("cosmos_predict2.experiments", reload=True) + import_all_modules_from_package("cosmos_policy._src.predict2.configs.video2world.experiment", reload=True) + + # experiment config are defined in the experiment folder + # call import_all_modules_from_package to register them + import_all_modules_from_package( + "cosmos_policy._src.predict2.action.configs.action_conditioned.experiment", reload=True + ) + return c diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/data.py b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/data.py new file mode 100644 index 0000000000000000000000000000000000000000..caef3896c7553c2d29367e5af16b707031d06780 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/data.py @@ -0,0 +1,204 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os + +from hydra.core.config_store import ConfigStore +from megatron.core import parallel_state +from torch.utils.data import DataLoader, DistributedSampler + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.action.datasets.dataset_local import Dataset_3D + +try: + from cosmos_policy._src.predict2.action.configs.action_conditioned.experiment.gr00t_customized_gr1 import ( + register_gr00t_customized_gr1_data, + ) +except ImportError: + register_gr00t_customized_gr1_data = None + +# bridge dataset path +base_path = "datasets/bridge/" + +train_annotation_path = os.path.join(base_path, "annotation/train") +val_annotation_path = os.path.join(base_path, "annotation/val") +test_annotation_path = os.path.join(base_path, "annotation/test") + + +# experiment for next-frame prediction +bridge_train_dataset = L(Dataset_3D)( + train_annotation_path=train_annotation_path, + val_annotation_path=val_annotation_path, + test_annotation_path=test_annotation_path, + video_path=base_path, + fps_downsample_ratio=1, + num_action_per_chunk=1, + cam_ids=[0], + accumulate_action=False, + video_size=[256, 320], + val_start_frame_interval=1, + mode="train", +) +bridge_val_dataset = L(Dataset_3D)( + train_annotation_path=train_annotation_path, + val_annotation_path=val_annotation_path, + test_annotation_path=test_annotation_path, + video_path=base_path, + fps_downsample_ratio=1, + num_action_per_chunk=1, + cam_ids=[0], + accumulate_action=False, + video_size=[256, 320], + val_start_frame_interval=1, + mode="val", +) + +# experiment for action-sequence video prediction +bridge_13frame_480_640_train_dataset = L(Dataset_3D)( + train_annotation_path=train_annotation_path, + val_annotation_path=val_annotation_path, + test_annotation_path=test_annotation_path, + video_path=base_path, + fps_downsample_ratio=1, + num_action_per_chunk=12, + cam_ids=[0], + accumulate_action=False, + video_size=[480, 640], + val_start_frame_interval=1, + mode="train", +) +bridge_13frame_480_640_val_dataset = L(Dataset_3D)( + train_annotation_path=train_annotation_path, + val_annotation_path=val_annotation_path, + test_annotation_path=test_annotation_path, + video_path=base_path, + fps_downsample_ratio=1, + num_action_per_chunk=12, + cam_ids=[0], + accumulate_action=False, + video_size=[480, 640], + val_start_frame_interval=1, + mode="val", +) + + +# ------------------------------------------------------------ + + +# create dataloader for each dataset +def get_sampler(dataset): + return DistributedSampler( + dataset, + num_replicas=parallel_state.get_data_parallel_world_size(), + rank=parallel_state.get_data_parallel_rank(), + shuffle=True, + seed=0, + ) + + +def build_webdataset(webdataset_instance, **kwargs): + """Helper function to build WebDataset from a WebDataset instance. + + WebDatasets need to call build_dataset() to get the actual iterable dataset + that can be used with DataLoader. + + Args: + webdataset_instance: An instantiated WebDataset object. + **kwargs: Additional parameters to override on the webdataset instance + before building. This allows experiment configs to override parameters + like gripper_rescale_factor, num_action_per_chunk, etc. + """ + # Apply any parameter overrides to the webdataset instance + for key, value in kwargs.items(): + if hasattr(webdataset_instance, key): + setattr(webdataset_instance, key, value) + return webdataset_instance.build_dataset() + + +bridge_train_dataloader = L(DataLoader)( + dataset=bridge_train_dataset, + sampler=L(get_sampler)(dataset=bridge_train_dataset), + batch_size=1, + drop_last=True, +) +bridge_val_dataloader = L(DataLoader)( + dataset=bridge_val_dataset, + sampler=L(get_sampler)(dataset=bridge_val_dataset), + batch_size=1, + drop_last=True, +) + +bridge_13frame_480_640_train_dataloader = L(DataLoader)( + dataset=bridge_13frame_480_640_train_dataset, + sampler=L(get_sampler)(dataset=bridge_13frame_480_640_train_dataset), + batch_size=1, + drop_last=True, +) +bridge_13frame_480_640_val_dataloader = L(DataLoader)( + dataset=bridge_13frame_480_640_val_dataset, + sampler=L(get_sampler)(dataset=bridge_13frame_480_640_val_dataset), + batch_size=1, + drop_last=True, +) + + +def register_training_and_val_data(): + cs = ConfigStore.instance() + from cosmos_policy._src.predict2.configs.common.mock_data import MOCK_DATA_INTERLEAVE_CONFIG + + # Always register mock dataloaders to satisfy defaults when not overridden + cs.store( + group="data_train", + package="dataloader_train", + name="mock", + node=MOCK_DATA_INTERLEAVE_CONFIG, + ) + cs.store( + group="data_val", + package="dataloader_val", + name="mock", + node=MOCK_DATA_INTERLEAVE_CONFIG, + ) + + cs.store( + group="data_train", + package="dataloader_train", + name="bridge_train", + node=bridge_train_dataloader, + ) + cs.store( + group="data_val", + package="dataloader_val", + name="bridge_val", + node=bridge_val_dataloader, + ) + + # 13 frame 480 640 + cs.store( + group="data_train", + package="dataloader_train", + name="bridge_13frame_480_640_train", + node=bridge_13frame_480_640_train_dataloader, + ) + cs.store( + group="data_val", + package="dataloader_val", + name="bridge_13frame_480_640_val", + node=bridge_13frame_480_640_val_dataloader, + ) + + # Register gr00t_customized_gr1 data + if register_gr00t_customized_gr1_data is not None: + register_gr00t_customized_gr1_data() diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/experiment/exp_2B_action_conditioned_rectify_flow.py b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/experiment/exp_2B_action_conditioned_rectify_flow.py new file mode 100644 index 0000000000000000000000000000000000000000..d53918f0aef76f27825f72be6773f0cb07d25781 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/experiment/exp_2B_action_conditioned_rectify_flow.py @@ -0,0 +1,700 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +import functools + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import ( + duplicate_batches, + duplicate_batches_random, + get_cached_replay_dataloader, +) +from cosmos_policy._src.predict2.datasets.dataset_provider import get_image_dataset, get_video_dataset +from cosmos_policy._src.predict2.datasets.joint_dataloader import IterativeJointDataLoader +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy +from cosmos_policy._src.predict2.text_encoders.text_encoder import EmbeddingConcatStrategy + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=1000, + logging_iter=50, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000000000000, + ), + every_n_sample_ema=dict( + every_n=1000000000000, + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_STANDALONE = LazyDict( + dict( + defaults=[ + {"override /data_train": "mock"}, + {"override /model": "fsdp"}, + {"override /net": "cosmos_v1_2B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "fusedadamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_standalone", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.001, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100000], + ), + model=dict( + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + loss_scale=10.0, + adjust_video_noise=False, + scaling="rectified_flow", + sigma_data=1.0, + fsdp_shard_size=8, + resolution="720", + state_t=24, + resize_online=True, + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + rectified_flow_loss_weight_uniform=False, + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_2b_720_aggressive", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + use_empty_string=False, + ), + ), + sde=dict( + p_mean=1.6094379124341003, # math.log(5.0) + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + ), + ) + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000065000/", + load_training_state=False, + strict_resume=True, + ), + model_parallel=dict( + context_parallel_size=2, + ), + trainer=dict( + max_iter=100000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + callbacks=dict( + every_n_sample_reg=dict( + every_n=5000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + every_n_sample_ema=dict( + every_n=5000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + use_native_fps=True, + ), + ), + ratio=3, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW = LazyDict( + dict( + defaults=[ + {"override /data_train": "mock"}, + {"override /model": "fsdp_rectified_flow"}, + {"override /net": "cosmos_v1_2B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "adamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only", + ), + optimizer=dict( + lr=3e-5, # 2**(-14.5) = 3.0517578125e-05 + weight_decay=1e-3, + betas=[0.9, 0.999], + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[100], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + fsdp_shard_size=8, + resolution="720", + state_t=24, + shift=5, + use_dynamic_shift=False, + train_time_weight="reweighting", + train_time_distribution="logitnormal", + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + timestep_scale=0.001, + sac_config=dict( + mode="predict2_2b_720_aggressive", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + use_wan_fp32_strategy=True, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + use_empty_string=False, # (TODO: hanzim): check + ), + ), + tokenizer=dict( + temporal_window=16, + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + ), + ) + ), + checkpoint=dict( + save_iter=1000, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + ), + model_parallel=dict( + context_parallel_size=2, + ), + trainer=dict( + max_iter=150_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + callbacks=dict( + grad_clip=dict( + clip_norm=0.1, + ), + manual_gc=dict( + every_n=200, + ), + every_n_sample_reg=dict( + every_n=1000000000000, + ), + every_n_sample_ema=dict( + every_n=1000000000000, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + use_native_fps=True, + ), + ), + ratio=3, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + # {"override /data_train": None}, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_improved", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only/checkpoints/iter_000037000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + dataloader_train=L(IterativeJointDataLoader)( + dataloaders={ + "image_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_pretrain_and_synthetic_photoreal_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="qwen2p5_7b_v4", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=12, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + ), + ratio=1, + ), + "image_data_prompt_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_synthetic_filtered_combined_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="prompts", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=12, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + ), + ratio=1, + ), + "video_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name="cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_video_whole", + object_store="s3", + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + # will use the augmentor to filter out frame drop + # so min and max fps can just use generic ones + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + # does not touch on low res data that will have jittering + dataset_resolution_type="gt720p", + use_native_fps=True, + embedding_type=None, + is_train=True, + chunk_size=256, + ), + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + num_workers=2, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + ), + ratio=2, + ), + }, + ), + ), + flags={"allow_objects": True}, +) + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --config=cosmos_policy/_src/predict2/action/configs/action_conditioned/config.py -- experiment=cosmos_predict2p5_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_480_640_ ~dataloader_train.dataloaders +""" +AC_REASON_EMBEDDINGS_RECTIFIED_FLOW_2B = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only", + {"override /model": "action_conditioned_video2world_fsdp_rectified_flow"}, + {"override /net": "cosmos_v1_2B_action_conditioned"}, + {"override /conditioner": "action_conditioned_video_conditioner"}, + {"override /data_train": "mock"}, + {"override /data_val": "mock"}, + ], + job=dict( + group="official_runs_vid2vid", + name="cosmos_predict2p5_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_480_640_", + project="cosmos_predict2_action_conditioned", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + callbacks=dict( + every_n_sample_reg=dict( + every_n=500, + do_x0_prediction=False, + guidance=[0], + fps=16, + ), + every_n_sample_ema=dict( + every_n=500, + do_x0_prediction=False, + guidance=[0], + fps=16, + ), + ), + ), + model_parallel=dict( + context_parallel_size=1, + ), + model=dict( + config=dict( + # NOTE: this should be 1 for the action conditioned model + min_num_conditional_frames=1, + max_num_conditional_frames=1, + # overwrite the probs to disable random num of conditional frames + conditional_frames_probs=None, + state_t=1 + 12 // 4, + net=dict( + action_dim=7, + num_action_per_chunk=12, + ), + ), + ), + dataloader_train=dict( + batch_size=2, + ), + ), + flags={"allow_objects": True}, +) + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --config=cosmos_policy/_src/predict2/action/configs/action_conditioned/config.py -- experiment=cosmos_predict2p5_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_256x320 ~dataloader_train.dataloaders +""" +AC_CHUNK_MULTI_VIEW_REASON_EMBEDDINGS_RECTIFIED_FLOW_2B_BRIDGE_13FRAME_256X320 = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2p5_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_480_640_", + {"override /net": "cosmos_v1_2B_action_chunk_conditioned"}, + {"override /data_train": "bridge_13frame_480_640_train"}, + {"override /data_val": "bridge_13frame_480_640_val"}, + ], + job=dict( + group="official_runs_vid2vid", + name="cosmos_predict2p5_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_256x320", + project="cosmos_predict2_action_conditioned", + ), + optimizer=dict( + lr=32e-5, + weight_decay=0.1, + ), + model=dict( + config=dict( + state_t=1 + 12 // 4, + net=dict( + action_dim=7, + temporal_compression_ratio=4, + ), + ), + ), + dataloader_train=dict( + batch_size=8, + sampler=dict( + dataset=dict( + gripper_rescale_factor=1, num_action_per_chunk=12, fps_downsample_ratio=1, video_size=[256, 320] + ), + ), + dataset=dict( + gripper_rescale_factor=1, num_action_per_chunk=12, fps_downsample_ratio=1, video_size=[256, 320] + ), + ), + ), + flags={"allow_objects": True}, +) + + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_STANDALONE, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_STANDALONE), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED + ), + ], + [ + AC_REASON_EMBEDDINGS_RECTIFIED_FLOW_2B, + *build_debug_runs(AC_REASON_EMBEDDINGS_RECTIFIED_FLOW_2B), + ], + [ + AC_CHUNK_MULTI_VIEW_REASON_EMBEDDINGS_RECTIFIED_FLOW_2B_BRIDGE_13FRAME_256X320, + *build_debug_runs(AC_CHUNK_MULTI_VIEW_REASON_EMBEDDINGS_RECTIFIED_FLOW_2B_BRIDGE_13FRAME_256X320), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/experiment/exp_2B_action_conditioned_rectify_flow_gr00t.py b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/experiment/exp_2B_action_conditioned_rectify_flow_gr00t.py new file mode 100644 index 0000000000000000000000000000000000000000..fb588138306e8e988ff8061b027f632c18023d80 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/experiment/exp_2B_action_conditioned_rectify_flow_gr00t.py @@ -0,0 +1,892 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +import functools + +# pyrefly: ignore [missing-import] +from cosmos_predict2.config import MODEL_CHECKPOINTS, ModelKey +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.utils.checkpoint_db import get_checkpoint_path +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import ( + duplicate_batches, + duplicate_batches_random, + get_cached_replay_dataloader, +) +from cosmos_policy._src.predict2.datasets.dataset_provider import get_image_dataset, get_video_dataset +from cosmos_policy._src.predict2.datasets.joint_dataloader import IterativeJointDataLoader +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy +from cosmos_policy._src.predict2.text_encoders.text_encoder import EmbeddingConcatStrategy + +DEFAULT_CHECKPOINT = MODEL_CHECKPOINTS[ModelKey()] # This uses post_trained=True by default + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=1000, + logging_iter=50, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000000000000, + ), + every_n_sample_ema=dict( + every_n=1000000000000, + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_STANDALONE = LazyDict( + dict( + defaults=[ + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + {"override /model": "fsdp"}, + {"override /net": "cosmos_v1_2B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "fusedadamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_standalone", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.001, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100000], + ), + model=dict( + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + loss_scale=10.0, + adjust_video_noise=False, + scaling="rectified_flow", + sigma_data=1.0, + fsdp_shard_size=8, + resolution="720", + state_t=24, + resize_online=True, + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + rectified_flow_loss_weight_uniform=False, + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_2b_720_aggressive", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + use_empty_string=False, + ), + ), + sde=dict( + p_mean=1.6094379124341003, # math.log(5.0) + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + ), + ) + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000065000/", + load_training_state=False, + strict_resume=True, + ), + model_parallel=dict( + context_parallel_size=2, + ), + trainer=dict( + max_iter=100000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + callbacks=dict( + every_n_sample_reg=dict( + every_n=5000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + every_n_sample_ema=dict( + every_n=5000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + use_native_fps=True, + ), + ), + ratio=3, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW = LazyDict( + dict( + defaults=[ + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + {"override /model": "fsdp_rectified_flow"}, + {"override /net": "cosmos_v1_2B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "adamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only", + ), + optimizer=dict( + lr=3e-5, # 2**(-14.5) = 3.0517578125e-05 + weight_decay=1e-3, + betas=[0.9, 0.999], + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[100], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + fsdp_shard_size=8, + resolution="720", + state_t=24, + shift=5, + use_dynamic_shift=False, + train_time_weight="reweighting", + train_time_distribution="logitnormal", + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + timestep_scale=0.001, + sac_config=dict( + mode="predict2_2b_720_aggressive", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + use_wan_fp32_strategy=True, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + use_empty_string=False, # (TODO: hanzim): check + ), + ), + tokenizer=dict( + temporal_window=16, + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + ), + ) + ), + checkpoint=dict( + save_iter=1000, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + ), + model_parallel=dict( + context_parallel_size=2, + ), + trainer=dict( + max_iter=150_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + callbacks=dict( + grad_clip=dict( + clip_norm=0.1, + ), + manual_gc=dict( + every_n=200, + ), + every_n_sample_reg=dict( + every_n=1000000000000, + ), + every_n_sample_ema=dict( + every_n=1000000000000, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + use_native_fps=True, + ), + ), + ratio=3, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + # {"override /data_train": None}, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_improved", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only/checkpoints/iter_000037000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + dataloader_train=L(IterativeJointDataLoader)( + dataloaders={ + "image_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_pretrain_and_synthetic_photoreal_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="qwen2p5_7b_v4", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=12, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + ), + ratio=1, + ), + "image_data_prompt_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_synthetic_filtered_combined_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="prompts", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=12, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + ), + ratio=1, + ), + "video_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name="cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_video_whole", + object_store="s3", + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + # will use the augmentor to filter out frame drop + # so min and max fps can just use generic ones + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + # does not touch on low res data that will have jittering + dataset_resolution_type="gt720p", + use_native_fps=True, + embedding_type=None, + is_train=True, + chunk_size=256, + ), + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + num_workers=2, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + ), + ratio=2, + ), + }, + ), + ), + flags={"allow_objects": True}, +) + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --config=cosmos_policy/_src/predict2/action/configs/config.py -- experiment=cosmos_predict2p1_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_480_640_ ~dataloader_train.dataloaders +""" +AC_REASON_EMBEDDINGS_RECTIFIED_FLOW_2B = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only", + {"override /model": "action_conditioned_video2world_fsdp_rectified_flow"}, + {"override /net": "cosmos_v1_2B_action_conditioned"}, + {"override /conditioner": "action_conditioned_video_conditioner"}, + {"override /data_train": "bridge_13frame_480_640_train"}, + {"override /data_val": "bridge_13frame_480_640_val"}, + ], + job=dict( + group="official_runs_vid2vid", + name="cosmos_predict2p1_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_480_640_", + project="cosmos_predict2_action_conditioned", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + callbacks=dict( + every_n_sample_reg=dict( + every_n=500, + do_x0_prediction=False, + guidance=[0], + fps=16, + ), + every_n_sample_ema=dict( + every_n=500, + do_x0_prediction=False, + guidance=[0], + fps=16, + ), + ), + ), + model_parallel=dict( + context_parallel_size=1, + ), + model=dict( + config=dict( + # NOTE: this should be 1 for the action conditioned model + min_num_conditional_frames=1, + max_num_conditional_frames=1, + # overwrite the probs to disable random num of conditional frames + conditional_frames_probs=None, + state_t=1 + 12 // 4, + net=dict( + action_dim=7, + num_action_per_chunk=12, + ), + ), + ), + dataloader_train=dict( + batch_size=2, + ), + ), + flags={"allow_objects": True}, +) + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --config=cosmos_policy/_src/predict2/action/configs/config.py -- experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame ~dataloader_train.dataloaders +""" +AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2p1_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_480_640_", + {"override /net": "cosmos_v1_2B_action_chunk_conditioned"}, + {"override /data_train": "gr00t_customiezed_gr1_train"}, + {"override /data_val": "gr00t_customiezed_gr1_val"}, + ], + job=dict( + group="official_runs_vid2vid", + name="cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame", + project="cosmos_predict2_action_conditioned", + ), + model=dict( + config=dict( + state_t=1 + 12 // 4, + net=dict( + action_dim=29, + ), + ), + ), + dataloader_train=dict( + batch_size=4, + ), + ), + flags={"allow_objects": True}, +) + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --config=cosmos_policy/_src/predict2/action/configs/config.py -- experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full ~dataloader_train.dataloaders +""" +AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME_FULL = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame", + ], + job=dict( + group="official_runs_vid2vid", + name="cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full", + project="cosmos_predict2_action_conditioned", + ), + dataloader_train=dict( + batch_size=4, + sampler=dict( + dataset=dict(num_frames=13, data_split="full"), + ), + dataset=dict(num_frames=13, data_split="full"), + ), + ), + flags={"allow_objects": True}, +) + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --config=cosmos_policy/_src/predict2/action/configs/action_conditioned/config.py -- experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes_release ~dataloader_train.dataloaders +""" +AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME_FULL_16NODES = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full", + ], + job=dict( + group="official_runs_vid2vid", + name="cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes_release", + project="cosmos_predict2_action_conditioned", + ), + optimizer=dict( + lr=16e-5, + weight_decay=0.1, + ), + ), + flags={"allow_objects": True}, +) + +default_experiment = DEFAULT_CHECKPOINT.experiment +load_path = get_checkpoint_path(DEFAULT_CHECKPOINT.s3.uri) +ac_reason_embeddings_rectified_flow_2b_oss = LazyDict( + dict( + defaults=[ + default_experiment, + {"override /model": "action_conditioned_video2world_fsdp_rectified_flow"}, + {"override /net": "cosmos_v1_2B_action_conditioned"}, + {"override /conditioner": "action_conditioned_video_conditioner"}, + {"override /data_train": "bridge_13frame_480_640_train"}, + {"override /data_val": "bridge_13frame_480_640_val"}, + "_self_", + ], + job=dict( + project="cosmos_predict2_action_conditioned", + group="cosmos_predict_v2p5", + name="2b_bridge_action_conditioned_oss", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + checkpoint=dict( + save_iter=2_000, + load_path=load_path, + load_training_state=False, + strict_resume=False, + load_from_object_store=dict( + enabled=False, + ), + save_to_object_store=dict( + enabled=False, + ), + ), + trainer=dict( + straggler_detection=dict(enabled=False), + callbacks=dict( + every_n_sample_reg=dict( + every_n=5000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + save_s3=False, + ), + every_n_sample_ema=dict( + every_n=5000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + save_s3=False, + ), + heart_beat=dict( + save_s3=False, + ), + iter_speed=dict( + hit_thres=100, + save_s3=False, + ), + device_monitor=dict( + save_s3=False, + ), + wandb=dict( + save_s3=False, + ), + wandb_10x=dict( + save_s3=False, + ), + dataloader_speed=dict( + save_s3=False, + ), + ), + ), + model_parallel=dict( + context_parallel_size=1, + ), + model=dict( + config=dict( + # NOTE: this should be 1 for the action conditioned model + min_num_conditional_frames=1, + max_num_conditional_frames=1, + # overwrite the probs to disable random num of conditional frames + conditional_frames_probs=None, + state_t=1 + 12 // 4, + net=dict( + action_dim=7, + num_action_per_chunk=12, + ), + ), + ), + dataloader_train=dict( + batch_size=2, + ), + ), + flags={"allow_objects": True}, +) + + +AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME_FULL_16NODES_OSS = LazyDict( + dict( + defaults=[ + "/experiment/2b_bridge_action_conditioned_oss", + {"override /net": "cosmos_v1_2B_action_chunk_conditioned"}, + {"override /data_train": "gr00t_customiezed_gr1_train"}, + {"override /data_val": "gr00t_customiezed_gr1_val"}, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes_release_oss", + project="cosmos_predict2_action_conditioned", + ), + model=dict( + config=dict( + state_t=1 + 12 // 4, + net=dict( + action_dim=29, + ), + ), + ), + dataloader_train=dict( + batch_size=4, + sampler=dict( + dataset=dict(num_frames=13, data_split="full"), + ), + dataset=dict(num_frames=13, data_split="full"), + ), + optimizer=dict( + lr=16e-5, + weight_decay=0.1, + ), + ), + flags={"allow_objects": True}, +) + + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_STANDALONE, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_STANDALONE), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED + ), + ], + [ + AC_REASON_EMBEDDINGS_RECTIFIED_FLOW_2B, + *build_debug_runs(AC_REASON_EMBEDDINGS_RECTIFIED_FLOW_2B), + ], + [ + AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME, + *build_debug_runs(AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME), + ], + [ + AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME_FULL_16NODES, + *build_debug_runs(AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME_FULL_16NODES), + ], + [ + AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME_FULL, + *build_debug_runs(AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME_FULL), + ], + [ + ac_reason_embeddings_rectified_flow_2b_oss, + *build_debug_runs(ac_reason_embeddings_rectified_flow_2b_oss), + ], + [ + AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME_FULL_16NODES_OSS, + *build_debug_runs(AC_CHUNK_MULTI_VIEW_2B_GR00T_GR1_CUSTOMIZED_13FRAME_FULL_16NODES_OSS), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/experiment/gr00t_customized_gr1.py b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/experiment/gr00t_customized_gr1.py new file mode 100644 index 0000000000000000000000000000000000000000..63a90cdd80485e93280b7ee0c36e1eea3e6ccb87 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/experiment/gr00t_customized_gr1.py @@ -0,0 +1,114 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os + +from hydra.core.config_store import ConfigStore +from megatron.core import parallel_state +from torch.utils.data import DataLoader, DistributedSampler + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.dataset import LeRobotDataset +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.groot_configs import ( + construct_modality_config_and_transforms, +) + +# [local] gr00t_gr1 dataset path +base_path_gr00t_gr1_local = "/project/cosmos/user/datasets/gr1_unified/gr1_unified.RU0226RemoveStaticFreq20" +train_annotation_path_gr00t_gr1_local = os.path.join(base_path_gr00t_gr1_local, "annotation/train") +val_annotation_path_gr00t_gr1_local = os.path.join(base_path_gr00t_gr1_local, "annotation/train") + +# Construct modality configs and transforms +modality_configs, train_transform, test_transform = construct_modality_config_and_transforms( + num_frames=13, embodiment="gr1", downscaled_res=False +) + + +gr00t_customiezed_gr1_train_dataset = L(LeRobotDataset)( + num_frames=13, + time_division_factor=4, + time_division_remainder=1, + max_pixels=1920 * 1080, + data_file_keys=("video",), + image_file_extension=("jpg", "jpeg", "png", "webp"), + video_file_extension=("mp4", "avi", "mov", "wmv", "mkv", "flv", "webm"), + repeat=1, + args=None, + dataset_path=base_path_gr00t_gr1_local, + data_split="train", + embodiment="gr1", + downscaled_res=False, +) + +gr00t_customiezed_gr1_val_dataset = L(LeRobotDataset)( + num_frames=13, + time_division_factor=4, + time_division_remainder=1, + max_pixels=1920 * 1080, + data_file_keys=("video",), + image_file_extension=("jpg", "jpeg", "png", "webp"), + video_file_extension=("mp4", "avi", "mov", "wmv", "mkv", "flv", "webm"), + repeat=1, + args=None, + dataset_path=base_path_gr00t_gr1_local, + data_split="test", + embodiment="gr1", + downscaled_res=False, +) + + +# Dataloader helper function +def get_sampler(dataset): + return DistributedSampler( + dataset, + num_replicas=parallel_state.get_data_parallel_world_size(), + rank=parallel_state.get_data_parallel_rank(), + shuffle=True, + seed=0, + ) + + +# Dataloader definitions +gr00t_customiezed_gr1_train_dataloader = L(DataLoader)( + dataset=gr00t_customiezed_gr1_train_dataset, + sampler=L(get_sampler)(dataset=gr00t_customiezed_gr1_train_dataset), + batch_size=1, + drop_last=True, +) + +gr00t_customiezed_gr1_val_dataloader = L(DataLoader)( + dataset=gr00t_customiezed_gr1_val_dataset, + sampler=L(get_sampler)(dataset=gr00t_customiezed_gr1_val_dataset), + batch_size=1, + drop_last=True, +) + + +# Registration function +def register_gr00t_customized_gr1_data(): + cs = ConfigStore.instance() + + cs.store( + group="data_train", + package="dataloader_train", + name="gr00t_customiezed_gr1_train", + node=gr00t_customiezed_gr1_train_dataloader, + ) + cs.store( + group="data_val", + package="dataloader_val", + name="gr00t_customiezed_gr1_val", + node=gr00t_customiezed_gr1_val_dataloader, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/model.py b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/model.py new file mode 100644 index 0000000000000000000000000000000000000000..2ea2260b3ef9aae358700d1e91b1eecd61c4a0e8 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/model.py @@ -0,0 +1,75 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.action.models.action_conditioned_video2world_model import ( + ActionConditionedVideo2WorldConfig, + ActionConditionedVideo2WorldModel, +) +from cosmos_policy._src.predict2.action.models.action_conditioned_video2world_rectified_flow_model import ( + ActionVideo2WorldModelRectifiedFlow, + Video2WorldModelRectifiedFlowConfig, +) + +# EDM model +DDP_CONFIG = dict( + trainer=dict( + distributed_parallelism="ddp", + ), + model=L(ActionConditionedVideo2WorldModel)( + config=ActionConditionedVideo2WorldConfig(), + _recursive_=False, + ), +) + +FSDP_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(ActionConditionedVideo2WorldModel)( + config=ActionConditionedVideo2WorldConfig( + fsdp_shard_size=8, + ), + _recursive_=False, + ), +) + +# rectified flow model +FSDP_RECTIFIED_FLOW_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(ActionVideo2WorldModelRectifiedFlow)( + config=Video2WorldModelRectifiedFlowConfig( + fsdp_shard_size=8, + state_t=24, + ), + _recursive_=False, + ), +) + + +def register_model(): + cs = ConfigStore.instance() + cs.store(group="model", package="_global_", name="action_conditioned_video2world_ddp", node=DDP_CONFIG) + cs.store(group="model", package="_global_", name="action_conditioned_video2world_fsdp", node=FSDP_CONFIG) + cs.store( + group="model", + package="_global_", + name="action_conditioned_video2world_fsdp_rectified_flow", + node=FSDP_RECTIFIED_FLOW_CONFIG, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/net.py b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/net.py new file mode 100644 index 0000000000000000000000000000000000000000..1bf98822e6a2a12cd1501cda338295f7aa489a6f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/configs/action_conditioned/net.py @@ -0,0 +1,139 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import copy + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.action.networks.action_conditioned_minimal_v1_lvg_dit import ( + ActionChunkConditionedMinimalV1LVGDiT, + ActionConditionedMinimalV1LVGDiT, +) + +# from cosmos_policy._src.predict2.networks.minimal_v1_lvg_dit import MinimalV1LVGDiT +from cosmos_policy._src.predict2.networks.minimal_v4_dit import SACConfig + +COSMOS_V1_7B_NET_MININET: LazyDict = L(ActionConditionedMinimalV1LVGDiT)( + max_img_h=240, + max_img_w=240, + max_frames=128, + in_channels=16, + out_channels=16, + patch_spatial=2, + patch_temporal=1, + model_channels=4096, + num_blocks=28, + num_heads=32, + concat_padding_mask=True, + pos_emb_cls="rope3d", + pos_emb_learnable=True, + pos_emb_interpolation="crop", + use_adaln_lora=True, + adaln_lora_dim=256, + atten_backend="minimal_a2a", + extra_per_block_abs_pos_emb=True, + rope_h_extrapolation_ratio=1.0, + rope_w_extrapolation_ratio=1.0, + rope_t_extrapolation_ratio=2.0, + sac_config=SACConfig(), +) +COSMOS_V1_2B_NET_MININET = copy.deepcopy(COSMOS_V1_7B_NET_MININET) +COSMOS_V1_2B_NET_MININET.model_channels = 2048 +COSMOS_V1_2B_NET_MININET.num_heads = 16 +COSMOS_V1_2B_NET_MININET.num_blocks = 28 +COSMOS_V1_2B_NET_MININET.extra_per_block_abs_pos_emb = False +COSMOS_V1_2B_NET_MININET.rope_t_extrapolation_ratio = 1.0 + +COSMOS_V1_14B_NET_MININET = copy.deepcopy(COSMOS_V1_7B_NET_MININET) +COSMOS_V1_14B_NET_MININET.model_channels = 5120 +COSMOS_V1_14B_NET_MININET.num_heads = 40 +COSMOS_V1_14B_NET_MININET.num_blocks = 36 +COSMOS_V1_14B_NET_MININET.extra_per_block_abs_pos_emb = False +COSMOS_V1_14B_NET_MININET.rope_t_extrapolation_ratio = 1.0 + +mini_net = copy.deepcopy(COSMOS_V1_7B_NET_MININET) +mini_net.model_channels = 1024 +mini_net.num_heads = 8 +mini_net.num_blocks = 2 +mini_net.rope_t_extrapolation_ratio = 1.0 + + +COSMOS_V1_7B_NET_MININET_ACTION_CHUNK: LazyDict = L(ActionChunkConditionedMinimalV1LVGDiT)( + max_img_h=240, + max_img_w=240, + max_frames=128, + in_channels=16, + out_channels=16, + patch_spatial=2, + patch_temporal=1, + model_channels=4096, + num_blocks=28, + num_heads=32, + concat_padding_mask=True, + pos_emb_cls="rope3d", + pos_emb_learnable=True, + pos_emb_interpolation="crop", + use_adaln_lora=True, + adaln_lora_dim=256, + atten_backend="minimal_a2a", + extra_per_block_abs_pos_emb=True, + rope_h_extrapolation_ratio=1.0, + rope_w_extrapolation_ratio=1.0, + rope_t_extrapolation_ratio=2.0, + sac_config=SACConfig(), +) +COSMOS_V1_2B_NET_MININET_ACTION_CHUNK = copy.deepcopy(COSMOS_V1_7B_NET_MININET_ACTION_CHUNK) +COSMOS_V1_2B_NET_MININET_ACTION_CHUNK.model_channels = 2048 +COSMOS_V1_2B_NET_MININET_ACTION_CHUNK.num_heads = 16 +COSMOS_V1_2B_NET_MININET_ACTION_CHUNK.num_blocks = 28 +COSMOS_V1_2B_NET_MININET_ACTION_CHUNK.extra_per_block_abs_pos_emb = False +COSMOS_V1_2B_NET_MININET_ACTION_CHUNK.rope_t_extrapolation_ratio = 1.0 + +COSMOS_V1_14B_NET_MININET_ACTION_CHUNK = copy.deepcopy(COSMOS_V1_7B_NET_MININET_ACTION_CHUNK) +COSMOS_V1_14B_NET_MININET_ACTION_CHUNK.model_channels = 5120 +COSMOS_V1_14B_NET_MININET_ACTION_CHUNK.num_heads = 40 +COSMOS_V1_14B_NET_MININET_ACTION_CHUNK.num_blocks = 36 +COSMOS_V1_14B_NET_MININET_ACTION_CHUNK.extra_per_block_abs_pos_emb = False +COSMOS_V1_14B_NET_MININET_ACTION_CHUNK.rope_t_extrapolation_ratio = 1.0 + + +def register_net(): + cs = ConfigStore.instance() + cs.store(group="net", package="model.config.net", name="mini_net", node=mini_net) + cs.store( + group="net", package="model.config.net", name="cosmos_v1_2B_action_conditioned", node=COSMOS_V1_2B_NET_MININET + ) + cs.store( + group="net", package="model.config.net", name="cosmos_v1_7B_action_conditioned", node=COSMOS_V1_7B_NET_MININET + ) + cs.store( + group="net", package="model.config.net", name="cosmos_v1_14B_action_conditioned", node=COSMOS_V1_14B_NET_MININET + ) + + # action chunk conditioned + cs.store( + group="net", + package="model.config.net", + name="cosmos_v1_2B_action_chunk_conditioned", + node=COSMOS_V1_2B_NET_MININET_ACTION_CHUNK, + ) + cs.store( + group="net", + package="model.config.net", + name="cosmos_v1_14B_action_chunk_conditioned", + node=COSMOS_V1_14B_NET_MININET_ACTION_CHUNK, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..907183fbedc6ddb60751363d313fb946c098c8e0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/__init__.py @@ -0,0 +1,16 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Empty __init__.py file to make this directory a Python package diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/dataset_local.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/dataset_local.py new file mode 100644 index 0000000000000000000000000000000000000000..f0072c319d863afa6f8d33c2cfe698655aced222 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/dataset_local.py @@ -0,0 +1,432 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Run this command to interactively debug: +PYTHONPATH=. python cosmos_policy/_src/predict2/action/datasets/dataset_local.py + +Adapted from: +https://github.com/bytedance/IRASim/blob/main/dataset/dataset_3D.py +""" + +import json +import os +import random +import time +import traceback +import warnings +from concurrent.futures import ThreadPoolExecutor, as_completed + +import imageio +import numpy as np +import torch +from decord import VideoReader, cpu +from einops import rearrange +from torch.utils.data import Dataset +from torchvision import transforms as T +from tqdm import tqdm + +from cosmos_policy._src.imaginaire.flags import INTERNAL +from cosmos_policy._src.imaginaire.utils.dataset_utils import Resize_Preprocess, ToTensorVideo, euler2rotm, rotm2euler + + +class Dataset_3D(Dataset): + def __init__( + self, + train_annotation_path, + val_annotation_path, + test_annotation_path, + video_path, + fps_downsample_ratio, + num_action_per_chunk, + cam_ids, + accumulate_action, + video_size, + val_start_frame_interval, + debug=False, + normalize=False, + pre_encode=False, + do_evaluate=False, + load_t5_embeddings=False, + load_action=True, + mode="train", + state_key="state", + gripper_key="continuous_gripper_state", + gripper_rescale_factor=1.0, + is_rollout=None, + ): + """Dataset class for loading 3D robot action-conditional data. + + This dataset loads robot trajectories consisting of RGB video frames, robot states (arm positions and gripper states), + and computes relative actions between consecutive frames. + + Args: + train_annotation_path (str): Path to training annotation files + val_annotation_path (str): Path to validation annotation files + test_annotation_path (str): Path to test annotation files + video_path (str): Base path to video files + fps_downsample_ratio (int): Interval between sampled frames in a sequence + num_action_per_chunk (int): Number of frames to load per sequence + cam_ids (list): List of camera IDs to sample from + accumulate_action (bool): Whether to accumulate actions relative to first frame + video_size (list): Target size [H,W] for video frames + val_start_frame_interval (int): Frame sampling interval for validation/test + debug (bool, optional): If True, only loads subset of data. Defaults to False. + normalize (bool, optional): Whether to normalize video frames. Defaults to False. + pre_encode (bool, optional): Whether to pre-encode video frames. Defaults to False. + do_evaluate (bool, optional): Whether in evaluation mode. Defaults to False. + load_t5_embeddings (bool, optional): Whether to load T5 embeddings. Defaults to False. + load_action (bool, optional): Whether to load actions. Defaults to True. + mode (str, optional): Dataset mode - 'train', 'val' or 'test'. Defaults to 'train'. + + The dataset loads robot trajectories and computes: + - RGB video frames from specified camera views + - Robot arm states (xyz position + euler angles) + - Gripper states (binary open/closed) + - Relative actions between consecutive frames + + Actions are computed as relative transforms between frames: + - Translation: xyz offset in previous frame's coordinate frame + - Rotation: euler angles of relative rotation + - Gripper: binary gripper state + + Returns dict with: + - video: RGB frames tensor [T,C,H,W] + - action: Action tensor [T-1,7] + - video_name: Dict with episode/frame metadata + - latent: Pre-encoded video features if pre_encode=True + """ + + super().__init__() + if mode == "train": + self.data_path = train_annotation_path + self.start_frame_interval = 1 + elif mode == "val": + self.data_path = val_annotation_path + self.start_frame_interval = val_start_frame_interval + elif mode == "test": + self.data_path = test_annotation_path + self.start_frame_interval = val_start_frame_interval + self.video_path = video_path + self.fps_downsample_ratio = fps_downsample_ratio + self.mode = mode + + # self.sequence_length = num_frames + self.sequence_length = 1 + num_action_per_chunk + self.normalize = normalize + self.pre_encode = pre_encode + self.load_t5_embeddings = load_t5_embeddings + self.load_action = load_action + + self.cam_ids = cam_ids + self.accumulate_action = accumulate_action + self.is_rollout = is_rollout + + self.action_dim = 7 # ee xyz (3) + ee euler (3) + gripper(1) + self.c_act_scaler = [20.0, 20.0, 20.0, 20.0, 20.0, 20.0, gripper_rescale_factor] + self.c_act_scaler = np.array(self.c_act_scaler, dtype=float) + self.ann_files = self._init_anns(self.data_path) + self._filter_rollout() + + self._state_key = state_key + self._gripper_key = gripper_key + + print(f"{len(self.ann_files)} trajectories in total") + self.samples = self._init_sequences(self.ann_files) + + self.samples = sorted(self.samples, key=lambda x: (x["ann_file"], x["frame_ids"][0])) + if debug and not do_evaluate: + self.samples = self.samples[0:10] + print(f"{len(self.ann_files)} trajectories in total") + print(f"{len(self.samples)} samples in total") + # with open('./samples_16.pkl','wb') as file: + # pickle.dump(self.samples,file) + self.wrong_number = 0 + self.transform = T.Compose([T.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True)]) + self.training = False + self.preprocess = T.Compose( + [ + ToTensorVideo(), + Resize_Preprocess(tuple(video_size)), # 288 512 + T.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True), + ] + ) + self.not_norm_preprocess = T.Compose([ToTensorVideo(), Resize_Preprocess(tuple(video_size))]) + + def __str__(self): + return f"{len(self.ann_files)} samples from {self.data_path}" + + def _init_anns(self, data_dir): + ann_files = [os.path.join(data_dir, f) for f in os.listdir(data_dir) if f.endswith(".json")] + return ann_files + + def _init_sequences(self, ann_files): + samples = [] + with ThreadPoolExecutor(32) as executor: + future_to_ann_file = { + executor.submit(self._load_and_process_ann_file, ann_file): ann_file for ann_file in ann_files + } + for future in tqdm(as_completed(future_to_ann_file), total=len(ann_files)): + samples.extend(future.result()) + return samples + + def _filter_rollout(self): + if self.is_rollout is None: + return + + print(f"Filtering rollout: {self.is_rollout}") + ann_files = [] + # Check if any file in self.ann_files has "is_eval" set to True + for ann_file in self.ann_files: + with open(ann_file, "r") as f: + ann_data = json.load(f) + is_eval = ann_data["episode_metadata"]["is_eval"] + if self.is_rollout and is_eval: + ann_files.append(ann_file) + elif not self.is_rollout and not is_eval: + ann_files.append(ann_file) + + self.ann_files = ann_files + print(f"Filtered {len(ann_files)} rollout: {self.is_rollout}") + return + + def _load_and_process_ann_file(self, ann_file): + samples = [] + with open(ann_file, "r") as f: + ann = json.load(f) + + n_frames = len(ann[self._state_key]) + for frame_i in range(0, n_frames, self.start_frame_interval): + sample = dict() + sample["ann_file"] = ann_file + sample["frame_ids"] = [] + curr_frame_i = frame_i + while True: + if curr_frame_i > (n_frames - 1): + break + sample["frame_ids"].append(curr_frame_i) + if len(sample["frame_ids"]) == self.sequence_length: + break + curr_frame_i += self.fps_downsample_ratio + # make sure there are sequence_length number of frames + if len(sample["frame_ids"]) == self.sequence_length: + samples.append(sample) + return samples + + def __len__(self): + return len(self.samples) + + def _load_video(self, video_path, frame_ids): + vr = VideoReader(video_path, ctx=cpu(0), num_threads=2) + assert (np.array(frame_ids) < len(vr)).all() + assert (np.array(frame_ids) >= 0).all() + vr.seek(0) + frame_data = vr.get_batch(frame_ids).asnumpy() + return frame_data + + def _get_frames(self, label, frame_ids, cam_id, pre_encode): + if pre_encode: + raise NotImplementedError("Pre-encoded videos are not supported for this dataset.") + else: + video_path = label["videos"][cam_id]["video_path"] + video_path = os.path.join(self.video_path, video_path) + frames = self._load_video(video_path, frame_ids) + frames = frames.astype(np.uint8) + frames = torch.from_numpy(frames).permute(0, 3, 1, 2) # (l, c, h, w) + + def printvideo(videos, filename): + t_videos = rearrange(videos, "f c h w -> f h w c") + t_videos = ( + ((t_videos / 2.0 + 0.5).clamp(0, 1) * 255).detach().to(dtype=torch.uint8).cpu().contiguous().numpy() + ) + print(t_videos.shape) + writer = imageio.get_writer(filename, fps=4) # fps 是帧率 + for frame in t_videos: + writer.append_data(frame) # 1 4 13 23 # fp16 24 76 456 688 + + if self.normalize: + frames = self.preprocess(frames) + else: + frames = self.not_norm_preprocess(frames) + frames = torch.clamp(frames * 255.0, 0, 255).to(torch.uint8) + return frames + + def _get_obs(self, label, frame_ids, cam_id, pre_encode): + if cam_id is None: + temp_cam_id = random.choice(self.cam_ids) + else: + temp_cam_id = cam_id + frames = self._get_frames(label, frame_ids, cam_id=temp_cam_id, pre_encode=pre_encode) + return frames, temp_cam_id + + def _get_robot_states(self, label, frame_ids): + all_states = np.array(label[self._state_key]) + all_cont_gripper_states = np.array(label[self._gripper_key]) + states = all_states[frame_ids] + cont_gripper_states = all_cont_gripper_states[frame_ids] + arm_states = states[:, :6] + return arm_states, cont_gripper_states + + def _get_actions(self, arm_states, gripper_states, accumulate_action): + action = np.zeros((self.sequence_length - 1, self.action_dim)) + if accumulate_action: + base_xyz = arm_states[0, 0:3] + base_rpy = arm_states[0, 3:6] + base_rotm = euler2rotm(base_rpy) + for k in range(1, self.sequence_length): + curr_xyz = arm_states[k, 0:3] + curr_rpy = arm_states[k, 3:6] + curr_gripper = gripper_states[k] + curr_rotm = euler2rotm(curr_rpy) + rel_xyz = np.dot(base_rotm.T, curr_xyz - base_xyz) + rel_rotm = base_rotm.T @ curr_rotm + rel_rpy = rotm2euler(rel_rotm) + action[k - 1, 0:3] = rel_xyz + action[k - 1, 3:6] = rel_rpy + action[k - 1, 6] = curr_gripper + if k % 4 == 0: + base_xyz = arm_states[k, 0:3] + base_rpy = arm_states[k, 3:6] + base_rotm = euler2rotm(base_rpy) + else: + for k in range(1, self.sequence_length): + prev_xyz = arm_states[k - 1, 0:3] + prev_rpy = arm_states[k - 1, 3:6] + prev_rotm = euler2rotm(prev_rpy) + curr_xyz = arm_states[k, 0:3] + curr_rpy = arm_states[k, 3:6] + curr_gripper = gripper_states[k] + curr_rotm = euler2rotm(curr_rpy) + rel_xyz = np.dot(prev_rotm.T, curr_xyz - prev_xyz) + rel_rotm = prev_rotm.T @ curr_rotm + rel_rpy = rotm2euler(rel_rotm) + action[k - 1, 0:3] = rel_xyz + action[k - 1, 3:6] = rel_rpy + action[k - 1, 6] = curr_gripper + return torch.from_numpy(action) # (l - 1, act_dim) + + def __getitem__(self, index, cam_id=None, return_video=False): + if self.mode != "train": + np.random.seed(index) + random.seed(index) + + try: + sample = self.samples[index] + ann_file = sample["ann_file"] + frame_ids = sample["frame_ids"] + with open(ann_file, "r") as f: + label = json.load(f) + arm_states, gripper_states = self._get_robot_states(label, frame_ids) + actions = self._get_actions(arm_states, gripper_states, self.accumulate_action) + actions *= self.c_act_scaler + + data = dict() + if self.load_action: + data["action"] = actions.float() + + if self.pre_encode: + raise NotImplementedError("Pre-encoded videos are not supported for this dataset.") + else: + video, cam_id = self._get_obs(label, frame_ids, cam_id, pre_encode=False) + video = video.permute(1, 0, 2, 3) # Rearrange from [T, C, H, W] to [C, T, H, W] + data["video"] = video.to(dtype=torch.uint8) + + data["annotation_file"] = ann_file + + # NOTE: __key__ is used to uniquely identify the sample, required for callback functions + if "episode_id" in label: + data["__key__"] = label["episode_id"] + else: + try: + data["__key__"] = label["original_path"] + except Exception: + try: + data["__key__"] = label["episode_metadata"]["episode_id"] + except Exception: + data["__key__"] = label["episode_metadata"]["segment_id"] + + # Just add these to fit the interface + if self.load_t5_embeddings: + t5_embeddings = np.squeeze(np.load(ann_file.replace(".json", ".npy"))) + data["t5_text_embeddings"] = torch.from_numpy(t5_embeddings).cuda() + else: + data["t5_text_embeddings"] = torch.zeros(512, 1024, dtype=torch.bfloat16).cuda() + data["ai_caption"] = "" + data["t5_text_mask"] = torch.ones(512, dtype=torch.int64).cuda() + data["fps"] = 4 + data["image_size"] = 256 * torch.ones(4).cuda() + data["num_frames"] = self.sequence_length + data["padding_mask"] = torch.zeros(1, 256, 256).cuda() + + return data + except Exception: + warnings.warn( + f"Invalid data encountered: {self.samples[index]['ann_file']}. Skipped " + f"(by randomly sampling another sample in the same dataset)." + ) + warnings.warn("FULL TRACEBACK:") + warnings.warn(traceback.format_exc()) + self.wrong_number += 1 + print(self.wrong_number) + return self[np.random.randint(len(self.samples))] + + +if INTERNAL: + """ Run this command to interactively debug: + PYTHONPATH=. python cosmos_policy/_src/predict2/action/datasets/dataset_local.py + """ + if __name__ == "__main__": + """ + PYTHONPATH=. python cosmos_policy/_src/predict2/action/datasets/dataset_local.py + """ + + base_path_pi_benchmark_local = "/project/cosmos/user/nvidia-cosmos-raw-data/ur5-data/video-evals-raw-data/datasets/action_dataset/single_chunk/" + train_annotation_path_pi_benchmark_local = os.path.join(base_path_pi_benchmark_local, "annotation/val") + val_annotation_path_pi_benchmark_local = os.path.join(base_path_pi_benchmark_local, "annotation/val") + test_annotation_path_pi_benchmark_local = os.path.join(base_path_pi_benchmark_local, "annotation/test") + dataset = Dataset_3D( + train_annotation_path=train_annotation_path_pi_benchmark_local, + val_annotation_path=val_annotation_path_pi_benchmark_local, + test_annotation_path=test_annotation_path_pi_benchmark_local, + video_path=base_path_pi_benchmark_local, + fps_downsample_ratio=1, + num_action_per_chunk=1, + cam_ids=["base_0"], + accumulate_action=False, + video_size=[480, 640], + val_start_frame_interval=1, + mode="train", + state_key="ee_pose", + is_rollout=None, + ) + + indices = [0, 13, 200, -1] + for idx in indices: + start_time = time.time() + print( + ( + f"{idx=} " + f"{dataset[idx]['video'].sum()=}\n" + f"{dataset[idx]['video'].shape=}\n" + # f"{dataset[idx]['video_name']=}\n" + f"{dataset[idx]['action'].sum()=}\n" + "---" + ) + ) + end_time = time.time() + print(f"Time taken: {end_time - start_time} seconds") + from IPython import embed + + embed() diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/dataset_mv_local.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/dataset_mv_local.py new file mode 100644 index 0000000000000000000000000000000000000000..368102b11e2e6bff00e1f079d812521496706a43 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/dataset_mv_local.py @@ -0,0 +1,112 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Run this command to interactively debug: +PYTHONPATH=. python cosmos_policy/_src/predict2/action/datasets/dataset_mv_local.py + +Adapted from: +https://github.com/bytedance/IRASim/blob/main/dataset/dataset_3D.py +""" + +import json +import random +import time + +import torch + +from cosmos_policy._src.predict2.action.datasets.dataset_local import Dataset_3D + + +class ActionConditionedMultiViewDataset(Dataset_3D): + def _get_obs(self, label, frame_ids, cam_id, pre_encode): + if cam_id is None: + temp_cam_id_0 = random.choice(self.cam_ids[0]) + temp_cam_id_1 = self.cam_ids[1] + else: + temp_cam_id_0 = cam_id[0] + temp_cam_id_1 = cam_id[1] + frames_0 = self._get_frames(label, frame_ids, cam_id=temp_cam_id_0, pre_encode=pre_encode) + frames_1 = self._get_frames(label, frame_ids, cam_id=temp_cam_id_1, pre_encode=pre_encode) + frames = torch.cat([frames_0, frames_1], dim=3) + return frames, [temp_cam_id_0, temp_cam_id_1] + + def _load_and_process_ann_file(self, ann_file): + samples = [] + with open(ann_file, "r") as f: + ann = json.load(f) + + n_frames = len(ann[self._state_key]) + + if isinstance(self.fps_downsample_ratio, int): + fps_downsample_ratio_list = [self.fps_downsample_ratio] + else: + fps_downsample_ratio_list = self.fps_downsample_ratio + + for fps_downsample_ratio in fps_downsample_ratio_list: + for frame_i in range(0, n_frames, self.start_frame_interval): + sample = dict() + sample["ann_file"] = ann_file + sample["frame_ids"] = [] + curr_frame_i = frame_i + while True: + if curr_frame_i > (n_frames - 1): + break + sample["frame_ids"].append(curr_frame_i) + if len(sample["frame_ids"]) == self.sequence_length: + break + # curr_frame_i += self.fps_downsample_ratio + curr_frame_i += fps_downsample_ratio + # make sure there are sequence_length number of frames + if len(sample["frame_ids"]) == self.sequence_length: + samples.append(sample) + return samples + + +if __name__ == "__main__": + dataset = ActionConditionedMultiViewDataset( + train_annotation_path="/project/cosmos/user/nvidia-cosmos-raw-data/ur5-data/video-evals-raw-data/datasets/action_dataset/single_chunk/annotation/train", + val_annotation_path="/project/cosmos/user/nvidia-cosmos-raw-data/ur5-data/video-evals-raw-data/datasets/action_dataset/single_chunk/annotation/val", + test_annotation_path="/project/cosmos/user/nvidia-cosmos-raw-data/ur5-data/video-evals-raw-data/datasets/action_dataset/single_chunk/annotation/test", + video_path="/project/cosmos/user/nvidia-cosmos-raw-data/ur5-data/video-evals-raw-data/datasets/action_dataset/single_chunk/", + fps_downsample_ratio=1, + num_action_per_chunk=1, + cam_ids=[["base_0", "base_1"], "wrist"], + accumulate_action=False, + video_size=[480, 640], + val_start_frame_interval=1, + mode="train", + load_t5_embeddings=False, + state_key="ee_pose", + ) + + indices = [0, 13, 200, -1] + for idx in indices: + start_time = time.time() + print( + ( + f"{idx=} " + f"{dataset[idx]['video'].sum()=}\n" + f"{dataset[idx]['video'].shape=}\n" + # f"{dataset[idx]['video_name']=}\n" + f"{dataset[idx]['action'].sum()=}\n" + "---" + ) + ) + end_time = time.time() + print(f"Time taken: {end_time - start_time} seconds") + from IPython import embed + + embed() diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/dataset_utils.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/dataset_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..197cb0cba7df0b892d775c427f3bbb0f4d2c15d5 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/dataset_utils.py @@ -0,0 +1,350 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Adapted from: +https://github.com/bytedance/IRASim/blob/main/dataset/dataset_util.py +""" + +import base64 +import math +import os +from io import BytesIO + +import numpy as np +import torch +import torch.distributed as dist +import torchvision.transforms.functional as F +from PIL import Image + + +def is_dist_avail_and_initialized(): + if not dist.is_available(): + return False + if not dist.is_initialized(): + return False + return True + + +def get_rank(): + if not is_dist_avail_and_initialized(): + return 0 + return dist.get_rank() + + +def get_1d_sincos_pos_embed_from_grid(embed_dim, pos): + """ + embed_dim: output dimension for each position + pos: a list of positions to be encoded: size (M,) + out: (M, D) + """ + assert embed_dim % 2 == 0 + omega = np.arange(embed_dim // 2, dtype=np.float32) + omega /= embed_dim / 2.0 + omega = 1.0 / 10000**omega # (D/2,) + + pos = pos.reshape(-1) # (M,) + out = np.einsum("m,d->md", pos, omega) # (M, D/2), outer product + + emb_sin = np.sin(out) # (M, D/2) + emb_cos = np.cos(out) # (M, D/2) + + emb = np.concatenate([emb_sin, emb_cos], axis=1) # (M, D) + return emb + + +def get_2d_sincos_pos_embed_from_grid(embed_dim, grid): + assert embed_dim % 2 == 0 + + # use half of dimensions to encode grid_h + emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0]) # (H*W, D/2) + emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1]) # (H*W, D/2) + + emb = np.concatenate([emb_h, emb_w], axis=1) # (H*W, D) + return emb + + +def get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False): + """ + grid_size: int of the grid height and width + return: + pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token) + """ + grid_h = np.arange(grid_size, dtype=np.float32) + grid_w = np.arange(grid_size, dtype=np.float32) + grid = np.meshgrid(grid_w, grid_h) # here w goes first + grid = np.stack(grid, axis=0) + + grid = grid.reshape([2, 1, grid_size, grid_size]) + pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid) + if cls_token: + pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0) + return pos_embed + + +def b64_2_img(data: str): + image_b64 = base64.b64decode(data) + img = Image.open(BytesIO(image_b64)).convert("RGB") + return img + + +def get_continuous_action(d_acts, c_act_max, c_act_min, n_bins): + c_act_max = c_act_max.to(d_acts.device) + c_act_min = c_act_min.to(d_acts.device) + c_acts = d_acts / (n_bins - 1) * (c_act_max - c_act_min) + c_act_min + return c_acts + + +def alpha2rotm(a): + """Alpha euler angle to rotation matrix.""" + rotm = np.array([[1, 0, 0], [0, np.cos(a), -np.sin(a)], [0, np.sin(a), np.cos(a)]]) + return rotm + + +def beta2rotm(b): + """Beta euler angle to rotation matrix.""" + rotm = np.array([[np.cos(b), 0, np.sin(b)], [0, 1, 0], [-np.sin(b), 0, np.cos(b)]]) + return rotm + + +def gamma2rotm(c): + """Gamma euler angle to rotation matrix.""" + rotm = np.array([[np.cos(c), -np.sin(c), 0], [np.sin(c), np.cos(c), 0], [0, 0, 1]]) + return rotm + + +def euler2rotm(euler_angles): + """Euler angle (ZYX) to rotation matrix.""" + alpha = euler_angles[0] + beta = euler_angles[1] + gamma = euler_angles[2] + + rotm_a = alpha2rotm(alpha) + rotm_b = beta2rotm(beta) + rotm_c = gamma2rotm(gamma) + + rotm = rotm_c @ rotm_b @ rotm_a + + return rotm + + +def isRotm(R): + # Checks if a matrix is a valid rotation matrix. + # Forked from Andy Zeng + Rt = np.transpose(R) + shouldBeIdentity = np.dot(Rt, R) + I = np.identity(3, dtype=R.dtype) + n = np.linalg.norm(I - shouldBeIdentity) + return n < 1e-6 + + +def rotm2euler(R): + # Forked from: https://learnopencv.com/rotation-matrix-to-euler-angles/ + # R = Rz * Ry * Rx + assert isRotm(R) + sy = math.sqrt(R[0, 0] * R[0, 0] + R[1, 0] * R[1, 0]) + singular = sy < 1e-6 + + if not singular: + x = math.atan2(R[2, 1], R[2, 2]) + y = math.atan2(-R[2, 0], sy) + z = math.atan2(R[1, 0], R[0, 0]) + else: + x = math.atan2(-R[1, 2], R[1, 1]) + y = math.atan2(-R[2, 0], sy) + z = 0 + + # (-pi , pi] + while x > np.pi: + x -= 2 * np.pi + while x <= -np.pi: + x += 2 * np.pi + while y > np.pi: + y -= 2 * np.pi + while y <= -np.pi: + y += 2 * np.pi + while z > np.pi: + z -= 2 * np.pi + while z <= -np.pi: + z += 2 * np.pi + return np.array([x, y, z]) + + +def get_converted_fp32_paths(deepspeed_ckpt_path): + deepspeed_ckpt_path = deepspeed_ckpt_path.rstrip("/") + ckpt_dir = os.path.dirname(deepspeed_ckpt_path) + ckpt_name = os.path.basename(deepspeed_ckpt_path) + fp32_ckpt_name = f"{ckpt_name}.fp32.pt" + converted_path = os.path.join(ckpt_dir, fp32_ckpt_name) + return converted_path + + +def quat2rotm(quat): + """Quaternion to rotation matrix. + + Args: + quat (4, numpy array): quaternion x, y, z, w + Returns: + rotm (3x3 numpy array): rotation matrix + """ + w = quat[3] + x = quat[0] + y = quat[1] + z = quat[2] + + s = w * w + x * x + y * y + z * z + + rotm = np.array( + [ + [1 - 2 * (y * y + z * z) / s, 2 * (x * y - z * w) / s, 2 * (x * z + y * w) / s], + [2 * (x * y + z * w) / s, 1 - 2 * (x * x + z * z) / s, 2 * (y * z - x * w) / s], + [2 * (x * z - y * w) / s, 2 * (y * z + x * w) / s, 1 - 2 * (x * x + y * y) / s], + ] + ) + + return rotm + + +def rotm2quat(R): + """Convert 3x3 rotation matrix to quaternion (w, x, y, z).""" + R = np.array(R, dtype=float) + trace = np.trace(R) + + if trace > 0: + s = 0.5 / np.sqrt(trace + 1.0) + w = 0.25 / s + x = (R[2, 1] - R[1, 2]) * s + y = (R[0, 2] - R[2, 0]) * s + z = (R[1, 0] - R[0, 1]) * s + else: + if R[0, 0] > R[1, 1] and R[0, 0] > R[2, 2]: + s = 2.0 * np.sqrt(1.0 + R[0, 0] - R[1, 1] - R[2, 2]) + w = (R[2, 1] - R[1, 2]) / s + x = 0.25 * s + y = (R[0, 1] + R[1, 0]) / s + z = (R[0, 2] + R[2, 0]) / s + elif R[1, 1] > R[2, 2]: + s = 2.0 * np.sqrt(1.0 + R[1, 1] - R[0, 0] - R[2, 2]) + w = (R[0, 2] - R[2, 0]) / s + x = (R[0, 1] + R[1, 0]) / s + y = 0.25 * s + z = (R[1, 2] + R[2, 1]) / s + else: + s = 2.0 * np.sqrt(1.0 + R[2, 2] - R[0, 0] - R[1, 1]) + w = (R[1, 0] - R[0, 1]) / s + x = (R[0, 2] + R[2, 0]) / s + y = (R[1, 2] + R[2, 1]) / s + z = 0.25 * s + + return np.array([w, x, y, z]) + + +class Resize_Preprocess: + """Video preprocessing class for resizing frames to a specific size.""" + + def __init__(self, size): + """ + Initialize the preprocessing class with the target size. + Args: + size (tuple): The target height and width as a tuple (height, width). + """ + self.size = size + + def __call__(self, video_frames): + """ + Apply the transformation to each frame in the video. + Args: + video_frames (torch.Tensor): A tensor representing a batch of video frames. + Returns: + torch.Tensor: The transformed video frames. + """ + # Resize each frame in the video + resized_frames = torch.stack([F.resize(frame, self.size, antialias=True) for frame in video_frames]) + return resized_frames + + +class Preprocess: + """Video preprocessing class for resizing clips while maintaining aspect ratio.""" + + def __init__(self, size): + self.size = size + + def __call__(self, clip): + clip = Preprocess.resize_scale(clip, self.size[0], self.size[1], interpolation_mode="bilinear") + return clip + + def __repr__(self) -> str: + return f"{self.__class__.__name__}(size={self.size})" + + @staticmethod + def resize_scale(clip, target_height, target_width, interpolation_mode): + """Resize video clip while maintaining aspect ratio.""" + target_ratio = target_height / target_width + H = clip.size(-2) + W = clip.size(-1) + clip_ratio = H / W + if clip_ratio > target_ratio: + scale_ = target_width / W + else: + scale_ = target_height / H + return torch.nn.functional.interpolate(clip, scale_factor=scale_, mode=interpolation_mode, align_corners=False) + + +class ToTensorVideo: + """ + Convert tensor data type from uint8 to float, divide value by 255.0 and + permute the dimensions of clip tensor + """ + + def __init__(self): + pass + + def __call__(self, clip): + """ + Args: + clip (torch.tensor, dtype=torch.uint8): Size is (T, C, H, W) + Return: + clip (torch.tensor, dtype=torch.float): Size is (T, C, H, W) + """ + return to_tensor(clip) + + def __repr__(self) -> str: + return self.__class__.__name__ + + +def to_tensor(clip): + """ + Convert tensor data type from uint8 to float, divide value by 255.0 and + permute the dimensions of clip tensor + Args: + clip (torch.tensor, dtype=torch.uint8): Size is (T, C, H, W) + Return: + clip (torch.tensor, dtype=torch.float): Size is (T, C, H, W) + """ + _is_tensor_video_clip(clip) + if not clip.dtype == torch.uint8: + raise TypeError("clip tensor should have data type uint8. Got %s" % str(clip.dtype)) + # return clip.float().permute(3, 0, 1, 2) / 255.0 + return clip.float() / 255.0 + + +def _is_tensor_video_clip(clip): + if not torch.is_tensor(clip): + raise TypeError("clip should be Tensor. Got %s" % type(clip)) + + if not clip.ndimension() == 4: + raise ValueError("clip should be 4D. Got %dD" % clip.dim()) + + return True diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/dataset.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/dataset.py new file mode 100644 index 0000000000000000000000000000000000000000..7d69b7c981b93df5cbf17970ece0d75aecfe2cec --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/dataset.py @@ -0,0 +1,1145 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import json +from collections import defaultdict +from pathlib import Path +from random import randint + +import imageio +import numpy as np +import pandas as pd +import torch +from einops import rearrange +from PIL import Image +from pydantic import BaseModel, ValidationError +from torch.utils.data import Dataset +from tqdm import tqdm + +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.embodiment_tags import EmbodimentTag +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.schema import ( + DatasetMetadata, + DatasetStatisticalValues, + LeRobotModalityMetadata, + LeRobotStateActionMetadata, +) + +# from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.transform. import ComposedModalityTransform +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.transform.base import ComposedModalityTransform +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.utils.video import ( + get_all_frames, + get_frames_by_timestamps, +) + +# from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.groot_configs import construct_modality_config_and_transforms + +LE_ROBOT_MODALITY_FILENAME = "meta/modality.json" +LE_ROBOT_EPISODE_FILENAME = "meta/episodes.jsonl" +LE_ROBOT_TASKS_FILENAME = "meta/tasks.jsonl" +LE_ROBOT_INFO_FILENAME = "meta/info.json" +LE_ROBOT_STATS_FILENAME = "meta/stats.json" +LE_ROBOT_DATA_FILENAME = "data/*/*.parquet" + + +def calculate_dataset_statistics(parquet_paths: list[Path]) -> dict: + """Calculate the dataset statistics of all columns for a list of parquet files.""" + # Dataset statistics + all_low_dim_data_list = [] + # Collect all the data + for parquet_path in tqdm( + sorted(list(parquet_paths)), + desc="Collecting all parquet files...", + ): + # Load the parquet file + parquet_data = pd.read_parquet(parquet_path) + parquet_data = parquet_data + all_low_dim_data_list.append(parquet_data) + all_low_dim_data = pd.concat(all_low_dim_data_list, axis=0) + # Compute dataset statistics + dataset_statistics = {} + for le_modality in all_low_dim_data.columns: + print(f"Computing statistics for {le_modality}...") + np_data = np.vstack([np.asarray(x, dtype=np.float32) for x in all_low_dim_data[le_modality]]) + dataset_statistics[le_modality] = { + "mean": np.mean(np_data, axis=0).tolist(), + "std": np.std(np_data, axis=0).tolist(), + "min": np.min(np_data, axis=0).tolist(), + "max": np.max(np_data, axis=0).tolist(), + "q01": np.quantile(np_data, 0.01, axis=0).tolist(), + "q99": np.quantile(np_data, 0.99, axis=0).tolist(), + } + return dataset_statistics + + +class ModalityConfig(BaseModel): + """Configuration for a modality.""" + + delta_indices: list[int] + """Delta indices to sample relative to the current index. The returned data will correspond to the original data at a sampled base index + delta indices.""" + modality_keys: list[str] + """The keys to load for the modality in the dataset.""" + + +class LeRobotSingleDataset(Dataset): + """ + Base dataset class for LeRobot that supports sharding. + """ + + def __init__( + self, + dataset_path: Path | str, + modality_configs: dict[str, ModalityConfig], + embodiment_tag: str | EmbodimentTag, + video_backend: str = "decord", + video_backend_kwargs: dict | None = None, + transforms: ComposedModalityTransform | None = None, + single_base_index: bool = False, + ): + """ + Initialize the dataset. + + Args: + dataset_path (Path | str): The path to the dataset. + modality_configs (dict[str, ModalityConfig]): The configuration for each modality. The keys are the modality names, and the values are the modality configurations. + See `ModalityConfig` for more details. + video_backend (str): Backend for video reading. + video_backend_kwargs (dict): Keyword arguments for the video backend when initializing the video reader. + transforms (ComposedModalityTransform): The transforms to apply to the dataset. + embodiment_tag (EmbodimentTag): Overload the embodiment tag for the dataset. e.g. define it as "new_embodiment" + """ + # first check if the path directory exists + if not Path(dataset_path).exists(): + raise FileNotFoundError(f"Dataset path {dataset_path} does not exist") + + self.modality_configs = modality_configs + self.video_backend = video_backend + self.video_backend_kwargs = video_backend_kwargs if video_backend_kwargs is not None else {} + self.transforms = transforms if transforms is not None else ComposedModalityTransform(transforms=[]) + + self._dataset_path = Path(dataset_path) + self._dataset_name = self._dataset_path.name + if isinstance(embodiment_tag, EmbodimentTag): + self.tag = embodiment_tag.value + else: + self.tag = embodiment_tag + + self._metadata = self._get_metadata(EmbodimentTag(self.tag)) + self._trajectory_ids, self._trajectory_lengths = self._get_trajectories() + self._all_steps = self._get_all_steps(single_base_index=single_base_index) + self._modality_keys = self._get_modality_keys() + self._delta_indices = self._get_delta_indices() + self.set_transforms_metadata(self.metadata) + self.set_epoch(0) + + print(f"Initialized dataset {self.dataset_name} with {embodiment_tag}") + + # LeRobot-specific config + self._lerobot_modality_meta = self._get_lerobot_modality_meta() + self._lerobot_info_meta = self._get_lerobot_info_meta() + self._data_path_pattern = self._get_data_path_pattern() + self._video_path_pattern = self._get_video_path_pattern() + self._chunk_size = self._get_chunk_size() + self._tasks = self._get_tasks() + self.curr_traj_data = None + self.curr_traj_id = None + + # Check if the dataset is valid + self._check_integrity() + + @property + def dataset_path(self) -> Path: + """The path to the dataset that contains the METADATA_FILENAME file.""" + return self._dataset_path + + @property + def metadata(self) -> DatasetMetadata: + """The metadata for the dataset, loaded from metadata.json in the dataset directory""" + return self._metadata + + @property + def trajectory_ids(self) -> np.ndarray: + """The trajectory IDs in the dataset, stored as a 1D numpy array of strings.""" + return self._trajectory_ids + + @property + def trajectory_lengths(self) -> np.ndarray: + """The trajectory lengths in the dataset, stored as a 1D numpy array of integers. + The order of the lengths is the same as the order of the trajectory IDs. + """ + return self._trajectory_lengths + + @property + def all_steps(self) -> list[tuple[int, int]]: + """The trajectory IDs and base indices for all steps in the dataset. + Example: + self.trajectory_ids: [0, 1, 2] + self.trajectory_lengths: [3, 2, 4] + return: [ + ("traj_0", 0), ("traj_0", 1), ("traj_0", 2), + ("traj_1", 0), ("traj_1", 1), + ("traj_2", 0), ("traj_2", 1), ("traj_2", 2), ("traj_2", 3) + ] + """ + return self._all_steps + + @property + def modality_keys(self) -> dict: + """The modality keys for the dataset. The keys are the modality names, and the values are the keys for each modality. + + Example: { + "video": ["video.image_side_0", "video.image_side_1"], + "state": ["state.eef_position", "state.eef_rotation"], + "action": ["action.eef_position", "action.eef_rotation"], + "language": ["language.human.task"], + "timestamp": ["timestamp"], + "reward": ["reward"], + } + """ + return self._modality_keys + + @property + def delta_indices(self) -> dict[str, np.ndarray]: + """The delta indices for the dataset. The keys are the modality.key, and the values are the delta indices for each modality.key.""" + return self._delta_indices + + @property + def dataset_name(self) -> str: + """The name of the dataset.""" + return self._dataset_name + + @property + def lerobot_modality_meta(self) -> LeRobotModalityMetadata: + """The metadata for the LeRobot dataset.""" + return self._lerobot_modality_meta + + @property + def lerobot_info_meta(self) -> dict: + """The metadata for the LeRobot dataset.""" + return self._lerobot_info_meta + + @property + def data_path_pattern(self) -> str: + """The path pattern for the LeRobot dataset.""" + return self._data_path_pattern + + @property + def video_path_pattern(self) -> str: + """The path pattern for the LeRobot dataset.""" + return self._video_path_pattern + + @property + def chunk_size(self) -> int: + """The chunk size for the LeRobot dataset.""" + return self._chunk_size + + @property + def tasks(self) -> pd.DataFrame: + """The tasks for the dataset.""" + return self._tasks + + def _get_metadata(self, embodiment_tag: EmbodimentTag) -> DatasetMetadata: + """Get the metadata for the dataset. + + Returns: + dict: The metadata for the dataset. + """ + + # 1. Modality metadata + modality_meta_path = self.dataset_path / LE_ROBOT_MODALITY_FILENAME + if not (modality_meta_path.exists()): + modality_meta_path = Path( + "/mnt/amlfs-03/shared/datasets/agibot-beta-converted-0512/agibotworld/modality.json" + ) + print( + "WARNING: Could not find modality.json in dataset path, falling back to /mnt/amlfs-03/shared/datasets/agibot-beta-converted-0512/agibotworld/modality.json" + ) + assert modality_meta_path.exists(), f"Please provide a {LE_ROBOT_MODALITY_FILENAME} file in {self.dataset_path}" + + # 1.1. State and action modalities + simplified_modality_meta: dict[str, dict] = {} + with open(modality_meta_path, "r") as f: + le_modality_meta = LeRobotModalityMetadata.model_validate(json.load(f)) + for modality in ["state", "action"]: + simplified_modality_meta[modality] = {} + le_state_action_meta: dict[str, LeRobotStateActionMetadata] = getattr(le_modality_meta, modality) + for subkey in le_state_action_meta: + state_action_dtype = np.dtype(le_state_action_meta[subkey].dtype) + if np.issubdtype(state_action_dtype, np.floating): + continuous = True + else: + continuous = False + simplified_modality_meta[modality][subkey] = { + "absolute": le_state_action_meta[subkey].absolute, + "rotation_type": le_state_action_meta[subkey].rotation_type, + "shape": [le_state_action_meta[subkey].end - le_state_action_meta[subkey].start], + "continuous": continuous, + } + + # 1.2. Video modalities + le_info_path = self.dataset_path / LE_ROBOT_INFO_FILENAME + assert le_info_path.exists(), f"Please provide a {LE_ROBOT_INFO_FILENAME} file in {self.dataset_path}" + with open(le_info_path, "r") as f: + le_info = json.load(f) + simplified_modality_meta["video"] = {} + for new_key in le_modality_meta.video: + original_key = le_modality_meta.video[new_key].original_key + if original_key is None: + original_key = new_key + + if original_key in le_info["features"]: + le_video_meta = le_info["features"][original_key] + else: + le_video_meta = le_info["features"][f"observation.images.{original_key}"] + # try: + # le_video_meta = le_info["features"][original_key] + # except: + # le_video_meta = le_info["features"][f"observation.images.{original_key}"] + height = le_video_meta["shape"][le_video_meta["names"].index("height")] + width = le_video_meta["shape"][le_video_meta["names"].index("width")] + # NOTE(FH): different lerobot dataset versions have different keys for the number of channels and fps + try: + channels = le_video_meta["shape"][le_video_meta["names"].index("channel")] + fps = le_video_meta["video_info"]["video.fps"] + except ValueError: + channels = le_video_meta["shape"][le_video_meta["names"].index("channels")] + fps = le_video_meta["info"]["video.fps"] + simplified_modality_meta["video"][new_key] = { + "resolution": [width, height], + "channels": channels, + "fps": fps, + } + + # 2. Dataset statistics + stats_path = self.dataset_path / LE_ROBOT_STATS_FILENAME + if "agibot" in str(stats_path): + # if not stats_path.exists(): + # print("WARNING: Could not find stats.json in dataset path, falling back to /mnt/amlfs-03/shared/datasets/agibot-beta-converted-0512/agibotworld/stats.json") + print( + "NOTE: Using standard action normalization at /mnt/amlfs-03/shared/datasets/agibot-beta-converted-0512/agibotworld/stats.json" + ) + stats_path = Path("/mnt/amlfs-03/shared/datasets/agibot-beta-converted-0512/agibotworld/stats.json") + try: + with open(stats_path, "r") as f: + le_statistics = json.load(f) + for stat in le_statistics.values(): + if isinstance(stat, int): + continue + DatasetStatisticalValues.model_validate(stat) + except (FileNotFoundError, ValidationError) as e: + print(f"Failed to load dataset statistics: {e}") + print(f"Calculating dataset statistics for {self.dataset_name}") + # Get all parquet files in the dataset paths + parquet_files = list((self.dataset_path).glob(LE_ROBOT_DATA_FILENAME)) + le_statistics = calculate_dataset_statistics(parquet_files) + dataset_statistics = {} + for our_modality in ["state", "action"]: + dataset_statistics[our_modality] = {} + for subkey in simplified_modality_meta[our_modality]: + dataset_statistics[our_modality][subkey] = {} + state_action_meta = le_modality_meta.get_key_meta(f"{our_modality}.{subkey}") + assert isinstance(state_action_meta, LeRobotStateActionMetadata) + le_modality = state_action_meta.original_key + for stat_name in le_statistics[le_modality]: + indices = np.arange( + state_action_meta.start, + state_action_meta.end, + ) + stat = np.array(le_statistics[le_modality][stat_name]) + dataset_statistics[our_modality][subkey][stat_name] = stat[indices].tolist() + + # 3. Full dataset metadata + metadata = DatasetMetadata( + statistics=dataset_statistics, # type: ignore + modalities=simplified_modality_meta, # type: ignore + embodiment_tag=embodiment_tag, + ) + + return metadata + + def _get_trajectories(self) -> tuple[np.ndarray, np.ndarray]: + """Get the trajectories in the dataset.""" + # Get trajectory lengths, IDs, and whitelist from dataset metadata + episode_path = self.dataset_path / LE_ROBOT_EPISODE_FILENAME + with open(episode_path, "r") as f: + episode_metadata = [json.loads(line) for line in f] + trajectory_ids = [] + trajectory_lengths = [] + for episode in episode_metadata: + trajectory_ids.append(episode["episode_index"]) + trajectory_lengths.append(episode["length"]) + return np.array(trajectory_ids), np.array(trajectory_lengths) + + def _get_all_steps(self, single_base_index=False) -> list[tuple[int, int]]: + """Get the trajectory IDs and base indices for all steps in the dataset. + + Returns: + list[tuple[str, int]]: A list of (trajectory_id, base_index) tuples. + + Example: + self.trajectory_ids: [0, 1, 2] + self.trajectory_lengths: [3, 2, 4] + return: [ + ("traj_0", 0), ("traj_0", 1), ("traj_0", 2), + ("traj_1", 0), ("traj_1", 1), + ("traj_2", 0), ("traj_2", 1), ("traj_2", 2), ("traj_2", 3) + ] + """ + all_steps: list[tuple[int, int]] = [] + for trajectory_id, trajectory_length in zip(self.trajectory_ids, self.trajectory_lengths): + if single_base_index: + all_steps.append((trajectory_id, 0)) + else: + for base_index in range(trajectory_length): + all_steps.append((trajectory_id, base_index)) + return all_steps + + def _get_modality_keys(self) -> dict: + """Get the modality keys for the dataset. + The keys are the modality names, and the values are the keys for each modality. + See property `modality_keys` for the expected format. + """ + modality_keys = defaultdict(list) + for modality, config in self.modality_configs.items(): + modality_keys[modality] = config.modality_keys + return modality_keys + + def _get_delta_indices(self) -> dict[str, np.ndarray]: + """Restructure the delta indices to use modality.key as keys instead of just the modalities.""" + delta_indices: dict[str, np.ndarray] = {} + for config in self.modality_configs.values(): + for key in config.modality_keys: + delta_indices[key] = np.array(config.delta_indices) + return delta_indices + + def _get_lerobot_modality_meta(self) -> LeRobotModalityMetadata: + """Get the metadata for the LeRobot dataset.""" + modality_meta_path = self.dataset_path / LE_ROBOT_MODALITY_FILENAME + if not (modality_meta_path.exists()): + modality_meta_path = Path( + "/mnt/amlfs-03/shared/datasets/agibot-beta-converted-0512/agibotworld/modality.json" + ) + print( + "WARNING: Could not find modality.json in dataset path, falling back to /mnt/amlfs-03/shared/datasets/agibot-beta-converted-0512/agibotworld/modality.json" + ) + assert modality_meta_path.exists(), f"Please provide a {LE_ROBOT_MODALITY_FILENAME} file in {self.dataset_path}" + with open(modality_meta_path, "r") as f: + modality_meta = LeRobotModalityMetadata.model_validate(json.load(f)) + return modality_meta + + def _get_lerobot_info_meta(self) -> dict: + """Get the metadata for the LeRobot dataset.""" + info_meta_path = self.dataset_path / LE_ROBOT_INFO_FILENAME + with open(info_meta_path, "r") as f: + info_meta = json.load(f) + return info_meta + + def _get_data_path_pattern(self) -> str: + """Get the data path pattern for the LeRobot dataset.""" + return self.lerobot_info_meta["data_path"] + + def _get_video_path_pattern(self) -> str: + """Get the video path pattern for the LeRobot dataset.""" + return self.lerobot_info_meta["video_path"] + + def _get_chunk_size(self) -> int: + """Get the chunk size for the LeRobot dataset.""" + return self.lerobot_info_meta["chunks_size"] + + def _get_tasks(self) -> pd.DataFrame: + """Get the tasks for the dataset.""" + tasks_path = self.dataset_path / LE_ROBOT_TASKS_FILENAME + with open(tasks_path, "r") as f: + tasks = [json.loads(line) for line in f] + df = pd.DataFrame(tasks) + return df.set_index("task_index") + + def _check_integrity(self): + """Use the config to check if the keys are valid and detect silent data corruption.""" + ERROR_MSG_HEADER = f"Error occurred in initializing dataset {self.dataset_name}:\n" + + for modality_config in self.modality_configs.values(): + for key in modality_config.modality_keys: + if key == "lapa_action" or key == "dream_actions": + continue # no need for any metadata for lapa actions because it comes normalized + # Check if the key is valid + try: + self.lerobot_modality_meta.get_key_meta(key) + except Exception as e: + raise ValueError(ERROR_MSG_HEADER + f"Unable to find key {key} in modality metadata:\n{e}") + + def set_transforms_metadata(self, metadata: DatasetMetadata): + """Set the metadata for the transforms. This is useful for transforms that need to know the metadata, such as the normalization values.""" + self.transforms.set_metadata(metadata) + + def set_epoch(self, epoch: int): + """Set the epoch for the dataset. + + Args: + epoch (int): The epoch to set. + """ + self.epoch = epoch + + def __len__(self) -> int: + """Get the total number of data points in the dataset. + + Returns: + int: the total number of data points in the dataset. + """ + return len(self.all_steps) + + def __str__(self) -> str: + """Get the description of the dataset.""" + return f"{self.dataset_name} ({len(self)} steps)" + + def __getitem__(self, index: int) -> dict: + """Get the data for a single step in a trajectory. + + Args: + index (int): The index of the step to get. + + Returns: + dict: The data for the step. + """ + trajectory_id, base_index = self.all_steps[index] + return self.transforms(self.get_step_data(trajectory_id, base_index)) + + def get_step_data(self, trajectory_id: int, base_index: int) -> dict: + """Get the RAW data for a single step in a trajectory. No transforms are applied. + + Args: + trajectory_id (int): The name of the trajectory. + base_index (int): The base step index in the trajectory. + + Returns: + dict: The RAW data for the step. + + Example return: + { + "video": { + "video.image_side_0": [B, T, H, W, C], + "video.image_side_1": [B, T, H, W, C], + }, + "state": { + "state.eef_position": [B, T, state_dim], + "state.eef_rotation": [B, T, state_dim], + }, + "action": { + "action.eef_position": [B, T, action_dim], + "action.eef_rotation": [B, T, action_dim], + }, + } + """ + data = {} + # Get the data for all modalities + self.curr_traj_data = self.get_trajectory_data(trajectory_id) + for modality in self.modality_keys: + # Get the data corresponding to each key in the modality + for key in self.modality_keys[modality]: + data[key] = self.get_data_by_modality(trajectory_id, modality, key, base_index) + return data + + def get_trajectory_data(self, trajectory_id: int) -> pd.DataFrame: + """Get the data for a trajectory.""" + if self.curr_traj_id == trajectory_id and self.curr_traj_data is not None: + return self.curr_traj_data + else: + chunk_index = self.get_episode_chunk(trajectory_id) + parquet_path = self.dataset_path / self.data_path_pattern.format( + episode_chunk=chunk_index, episode_index=trajectory_id + ) + assert parquet_path.exists(), f"Parquet file not found at {parquet_path}" + return pd.read_parquet(parquet_path) + + def get_trajectory_index(self, trajectory_id: int) -> int: + """Get the index of the trajectory in the dataset by the trajectory ID. + This is useful when you need to get the trajectory length or sampling weight corresponding to the trajectory ID. + + Args: + trajectory_id (str): The ID of the trajectory. + + Returns: + int: The index of the trajectory in the dataset. + """ + trajectory_indices = np.where(self.trajectory_ids == trajectory_id)[0] + if len(trajectory_indices) != 1: + raise ValueError(f"Error finding trajectory index for {trajectory_id}, found {trajectory_indices=}") + return trajectory_indices[0] + + def get_episode_chunk(self, ep_index: int) -> int: + """Get the chunk index for an episode index.""" + return ep_index // self.chunk_size + + def retrieve_data_and_pad( + self, + array: np.ndarray, + step_indices: np.ndarray, + max_length: int, + padding_strategy: str = "first_last", + ) -> np.ndarray: + """Retrieve the data from the dataset and pad it if necessary. + Args: + array (np.ndarray): The array to retrieve the data from. + step_indices (np.ndarray): The step indices to retrieve the data for. + max_length (int): The maximum length of the data. + padding_strategy (str): The padding strategy, either "first" or "last". + """ + # Get the padding indices + front_padding_indices = step_indices < 0 + end_padding_indices = step_indices >= max_length + padding_positions = np.logical_or(front_padding_indices, end_padding_indices) + # Retrieve the data with the non-padding indices + # If there exists some padding, Given T step_indices, the shape of the retrieved data will be (T', ...) where T' < T + raw_data = array[step_indices[~padding_positions]] + assert isinstance(raw_data, np.ndarray), f"{type(raw_data)=}" + # This is the shape of the output, (T, ...) + if raw_data.ndim == 1: + expected_shape = (len(step_indices),) + else: + expected_shape = (len(step_indices), *array.shape[1:]) + + # Pad the data + output = np.zeros(expected_shape) + # Assign the non-padded data + output[~padding_positions] = raw_data + # If there exists some padding, pad the data + if padding_positions.any(): + if padding_strategy == "first_last": + # Use first / last step data to pad + front_padding_data = array[0] + end_padding_data = array[-1] + output[front_padding_indices] = front_padding_data + output[end_padding_indices] = end_padding_data + elif padding_strategy == "zero": + # Use zero padding + output[padding_positions] = 0 + else: + raise ValueError(f"Invalid padding strategy: {padding_strategy}") + return output + + def get_video_path(self, trajectory_id: int, key: str) -> Path: + chunk_index = self.get_episode_chunk(trajectory_id) + original_key = self.lerobot_modality_meta.video[key].original_key + if original_key is None: + original_key = key + video_filename = self.video_path_pattern.format( + episode_chunk=chunk_index, episode_index=trajectory_id, video_key=original_key + ) + if not (self.dataset_path / video_filename).exists(): + original_key = f"observation.images.{original_key}" + video_filename = self.video_path_pattern.format( + episode_chunk=chunk_index, episode_index=trajectory_id, video_key=original_key + ) + return self.dataset_path / video_filename + + def get_video( + self, + trajectory_id: int, + key: str, + base_index: int, + ) -> np.ndarray: + """Get the video frames for a trajectory by a base index. + + Args: + dataset (BaseSingleDataset): The dataset to retrieve the data from. + trajectory_id (str): The ID of the trajectory. + key (str): The key of the video. + base_index (int): The base index of the trajectory. + + Returns: + np.ndarray: The video frames for the trajectory and frame indices. Shape: (T, H, W, C) + """ + # Get the step indices + step_indices = self.delta_indices[key] + base_index + # print(f"{step_indices=}") + # Get the trajectory index + trajectory_index = self.get_trajectory_index(trajectory_id) + # Ensure the indices are within the valid range + # This is equivalent to padding the video with extra frames at the beginning and end + step_indices = np.maximum(step_indices, 0) + # if step_indices[-1] >= self.trajectory_lengths[trajectory_index]: + # step_indices -= (self.trajectory_lengths[trajectory_index] - step_indices[-1] + 1) + step_indices = np.minimum(step_indices, self.trajectory_lengths[trajectory_index] - 1) + assert key.startswith("video."), f"Video key must start with 'video.', got {key}" + # Get the sub-key + key = key.replace("video.", "") + video_path = self.get_video_path(trajectory_id, key) + # Get the action/state timestamps for each frame in the video + assert self.curr_traj_data is not None, f"No data found for {trajectory_id=}" + assert "timestamp" in self.curr_traj_data.columns, f"No timestamp found in {trajectory_id=}" + timestamp: np.ndarray = self.curr_traj_data["timestamp"].to_numpy() + # Get the corresponding video timestamps from the step indices + video_timestamp = timestamp[step_indices] + + try: + return get_frames_by_timestamps( + video_path.as_posix(), + video_timestamp, + video_backend=self.video_backend, + video_backend_kwargs=self.video_backend_kwargs, + ) + except Exception: + self.video_backend = "torchvision_av" + return get_frames_by_timestamps( + video_path.as_posix(), + video_timestamp, + video_backend=self.video_backend, + video_backend_kwargs=self.video_backend_kwargs, + ) + + def get_state_or_action( + self, + trajectory_id: int, + modality: str, + key: str, + base_index: int, + ) -> np.ndarray: + """Get the state or action data for a trajectory by a base index. + If the step indices are out of range, pad with the data: + if the data is stored in absolute format, pad with the first or last step data; + otherwise, pad with zero. + + Args: + dataset (BaseSingleDataset): The dataset to retrieve the data from. + trajectory_id (int): The ID of the trajectory. + modality (str): The modality of the data. + key (str): The key of the data. + base_index (int): The base index of the trajectory. + + Returns: + np.ndarray: The data for the trajectory and step indices. + """ + # Get the step indices + step_indices = self.delta_indices[key] + base_index + # Get the trajectory index + trajectory_index = self.get_trajectory_index(trajectory_id) + # Get the maximum length of the trajectory + max_length = self.trajectory_lengths[trajectory_index] + assert key.startswith(modality + "."), f"{key} must start with {modality + '.'}, got {key}" + # Get the sub-key, e.g. state.joint_angles -> joint_angles + key = key.replace(modality + ".", "") + # Get the lerobot key + le_state_or_action_cfg = getattr(self.lerobot_modality_meta, modality) + le_key = le_state_or_action_cfg[key].original_key + if le_key is None: + le_key = key + # Get the data array, shape: (T, D) + assert self.curr_traj_data is not None, f"No data found for {trajectory_id=}" + assert le_key in self.curr_traj_data.columns, f"No {le_key} found in {trajectory_id=}" + data_array: np.ndarray = np.stack(self.curr_traj_data[le_key]) # type: ignore + assert data_array.ndim == 2, f"Expected 2D array, got {data_array.shape} array" + le_indices = np.arange( + le_state_or_action_cfg[key].start, + le_state_or_action_cfg[key].end, + ) + data_array = data_array[:, le_indices] + # Get the state or action configuration + state_or_action_cfg = getattr(self.metadata.modalities, modality)[key] + + # Pad the data + return self.retrieve_data_and_pad( + array=data_array, + step_indices=step_indices, + max_length=max_length, + padding_strategy="first_last" if state_or_action_cfg.absolute else "zero", + ) + + def get_language( + self, + trajectory_id: int, + key: str, + base_index: int, + ) -> list[str]: + """Get the language annotation data for a trajectory by step indices. + + Args: + dataset (BaseSingleDataset): The dataset to retrieve the data from. + trajectory_id (int): The ID of the trajectory. + key (str): The key of the annotation. + base_index (int): The base index of the trajectory. + + Returns: + list[str]: The annotation data for the trajectory and step indices. If no matching data is found, return empty strings. + """ + assert self.curr_traj_data is not None, f"No data found for {trajectory_id=}" + # Get the step indices + step_indices = self.delta_indices[key] + base_index + # Get the trajectory index + trajectory_index = self.get_trajectory_index(trajectory_id) + # Get the maximum length of the trajectory + max_length = self.trajectory_lengths[trajectory_index] + # Get the end times corresponding to the closest indices + step_indices = np.maximum(step_indices, 0) + step_indices = np.minimum(step_indices, max_length - 1) + # Get the annotations + task_indices: list[int] = [] + assert key.startswith("annotation."), f"Language key must start with 'annotation.', got {key}" + subkey = key.replace("annotation.", "") + annotation_meta = self.lerobot_modality_meta.annotation + assert annotation_meta is not None, f"Annotation metadata is None for {subkey}" + assert subkey in annotation_meta, ( + f"Annotation key {subkey} not found in metadata, available annotation keys: {annotation_meta.keys()}" + ) + subkey_meta = annotation_meta[subkey] + original_key = subkey_meta.original_key + if original_key is None: + original_key = key + for i in range(len(step_indices)): + task_indices.append(self.curr_traj_data[original_key][step_indices[i]].item()) + return self.tasks.loc[task_indices]["task"].tolist() + + def get_data_by_modality( + self, + trajectory_id: int, + modality: str, + key: str, + base_index: int, + ): + """Get the data corresponding to the modality for a trajectory by a base index. + This method will call the corresponding helper method based on the modality. + See the helper methods for more details. + NOTE: For the language modality, the data is padded with empty strings if no matching data is found. + + Args: + dataset (BaseSingleDataset): The dataset to retrieve the data from. + trajectory_id (int): The ID of the trajectory. + modality (str): The modality of the data. + key (str): The key of the data. + base_index (int): The base index of the trajectory. + """ + if modality == "video": + return self.get_video(trajectory_id, key, base_index) + elif modality == "state" or modality == "action": + return self.get_state_or_action(trajectory_id, modality, key, base_index) + elif modality == "language": + return self.get_language(trajectory_id, key, base_index) + else: + raise ValueError(f"Invalid modality: {modality}") + + +class CachedLeRobotSingleDataset(LeRobotSingleDataset): + def __init__(self, img_resize: tuple[int, int] | None = None, *args, **kwargs): + """ + This class caches the video frames for each trajectory and key. + It is recommended to use this class if the video frames need to be accessed multiple times. + + Args: + resize_img (tuple[int, int], optional): The size to resize the video frames to reduce memory usage. + """ + # Convert img_resize to tuple if it is not already + if img_resize is not None and not isinstance(img_resize, tuple): + img_resize = tuple(img_resize) + assert len(img_resize) == 2, f"Expected tuple of length 2, got {img_resize}" + self.img_resize = img_resize + + # Initialize img_resize attribute first to ensure it exists + super().__init__(*args, **kwargs) + cached_frames: dict[str, np.ndarray] = {} + + for key in self.modality_keys["video"]: + all_frames = [] + key = key.replace("video.", "") + for trajectory_id, trajectory_length in tqdm( + zip(self.trajectory_ids, self.trajectory_lengths), + total=len(self.trajectory_ids), + desc=f"Caching {key} frames", + ): + video_path = self.get_video_path(trajectory_id, key) + frames = get_all_frames( + video_path.as_posix(), + video_backend=self.video_backend, + video_backend_kwargs=self.video_backend_kwargs, + resize_size=img_resize, + ) + assert frames.ndim == 4, f"Expected 4D array, got {frames.shape} array" + assert frames.shape[3] == 3, f"Expected 3 channels, got {frames.shape[3]} channels" + # assert ( + # frames.shape[0] == trajectory_length + # ), f"Expected {trajectory_length} frames, got {frames.shape[0]} frames" + all_frames.append(frames) + cached_frames[key] = np.concatenate(all_frames, axis=0) + print(f"{key}: {cached_frames[key].shape}") + self.cached_frames = cached_frames + self.start_indices = np.cumsum(self.trajectory_lengths) - self.trajectory_lengths + + def get_video(self, trajectory_id: int, key: str, base_index: int) -> np.ndarray: + step_indices = self.delta_indices[key] + base_index + # Get the trajectory index + trajectory_index = self.get_trajectory_index(trajectory_id) + # Ensure the indices are within the valid range + # This is equivalent to padding the video with extra frames at the beginning and end + step_indices = np.maximum(step_indices, 0) + step_indices = np.minimum(step_indices, self.trajectory_lengths[trajectory_index] - 1) + assert key.startswith("video."), f"Video key must start with 'video.', got {key}" + # Get the sub-key + key = key.replace("video.", "") + # Calculate the absolute indices + absolute_indices = self.start_indices[trajectory_index] + step_indices + return self.cached_frames[key][absolute_indices] + + def get_step_data(self, trajectory_id: int, base_index: int) -> dict: + """Get the RAW data for a single step. No transforms are applied. + + Args: + trajectory_id (str): The ID of the trajectory. + base_index (int): The base index of the step. + + Returns: + dict: The data for the step. + """ + data = {} + self.curr_traj_data = self.get_trajectory_data(trajectory_id) + # Get the data for all modalities + for modality in self.modality_keys: + # Get the data corresponding to each key in the modality + for key in self.modality_keys[modality]: + data[key] = self.get_data_by_modality(trajectory_id, modality, key, base_index) + return data + + def set_transforms_metadata(self, metadata: DatasetMetadata): + """Set the metadata for the transforms. This is useful for transforms that need to know the metadata, such as the normalization values.""" + if self.img_resize is not None: + all_video_keys = [key for key in self.modality_keys["video"]] + for key in metadata.modalities.video: + if key in all_video_keys: + metadata.modalities.video[key].resolution = self.img_resize + super().set_transforms_metadata(metadata) + + +class WrappedLeRobotSingleDataset(LeRobotSingleDataset): + def __init__(self, *args, data_split="full", **kwargs): + super().__init__(*args, **kwargs) + + if data_split == "full": + pass + elif data_split == "train": + self._all_steps = self._all_steps[: -len(self) // 20] + elif data_split == "test": + self._all_steps = self._all_steps[-len(self) // 20 :] + + print(f"Dataset is split into {data_split} data, with {len(self._all_steps)} steps.") + + def _get_trajectories(self) -> tuple[np.ndarray, np.ndarray]: + """Get the trajectories in the dataset.""" + # Get trajectory lengths, IDs, and whitelist from dataset metadata + episode_path = self.dataset_path / LE_ROBOT_EPISODE_FILENAME + with open(episode_path, "r") as f: + episode_metadata = [json.loads(line) for line in f] + trajectory_ids = [] + trajectory_lengths = [] + for episode in episode_metadata: + trajectory_ids.append(episode["episode_index"]) + trajectory_lengths.append(episode["length"]) + return np.array(trajectory_ids), np.array(trajectory_lengths) + + def __getitem__(self, index: int) -> dict: + """Get the data for a single step in a trajectory. + + Args: + index (int): The index of the step to get. + + Returns: + dict: The data for the step. + """ + try: + # pdb.set_trace() + trajectory_id, base_index = self.all_steps[index] + original_outputs = self.transforms(self.get_step_data(trajectory_id, base_index)) + + # delta_actions = original_outputs["action"][1:] - original_outputs["action"][:-1] + # delta_actions = original_outputs["action"][1:] - original_outputs["action"][[0]] + # delta_actions /= torch.linspace(1, len(delta_actions), steps=len(delta_actions))[:, None] + + def printvideo(videos, filename): + t_videos = rearrange(videos, "c f h w -> f h w c") + t_videos = t_videos.detach().to(dtype=torch.uint8).cpu().contiguous().numpy() + writer = imageio.get_writer(filename, fps=5) + for frame in t_videos: + writer.append_data(frame) + + frames = torch.from_numpy(original_outputs["video"]) + # frames = torch.from_numpy(original_outputs["video"]) + frames = torch.clamp(frames * 255.0, 0, 255).to(torch.uint8) + frames = frames.squeeze(1).transpose(0, 1) + # printvideo(frames, "example.mp4") + + text = "" + if "annotation.human.coarse_action" in original_outputs: + text = original_outputs["annotation.human.coarse_action"][0].split(":")[-1].strip() + + video_path = { + key: self.get_video_path(trajectory_id, key.replace("video.", "")) + for key in self.modality_keys["video"] + } + data = { + "__key__": original_outputs["state"], + "action": original_outputs["action"], + # "action": original_outputs["action"][:-1], + # "action": delta_actions, + # "action": torch.zeros_like(delta_actions), + "video": frames, + # "video_path": video_path, + "ai_caption": "", + "text": text, + "t5_text_embeddings": torch.zeros(512, 1024, dtype=torch.bfloat16).cuda(), + "t5_text_mask": torch.ones(512, dtype=torch.int64).cuda(), + "fps": 4, + "image_size": 256 * torch.ones(4).cuda(), + "num_frames": 13, + "padding_mask": torch.zeros(1, 256, 256).cuda(), + } + return data + except Exception as e: + print(f"Error occurred while getting item {index}: {e}") + print("Retrying with a random index...") + return self.__getitem__(randint(0, len(self) - 1)) + + +class LeRobotDataset(torch.utils.data.Dataset): + def __init__( + self, + num_frames=81, + time_division_factor=4, + time_division_remainder=1, + max_pixels=1920 * 1080, + data_file_keys=("video",), + image_file_extension=("jpg", "jpeg", "png", "webp"), + video_file_extension=("mp4", "avi", "mov", "wmv", "mkv", "flv", "webm"), + repeat=1, + args=None, + dataset_path=None, + data_split="train", + embodiment=None, + downscaled_res=False, + ): + if args is not None: + # height = args.height + # width = args.width + max_pixels = args.max_pixels + num_frames = args.num_frames + data_file_keys = args.data_file_keys.split(",") + repeat = args.dataset_repeat + dataset_path = args.dataset_path + embodiment = args.embodiment + downscaled_res = args.downscaled_res + + self.num_frames = num_frames + self.time_division_factor = time_division_factor + self.time_division_remainder = time_division_remainder + self.max_pixels = max_pixels + # self.height = height + # self.width = width + # self.height_division_factor = height_division_factor + # self.width_division_factor = width_division_factor + self.data_file_keys = data_file_keys + self.image_file_extension = image_file_extension + self.video_file_extension = video_file_extension + self.repeat = repeat + + # from gr00t_dreams.data.dataset import WrappedLeRobotSingleDataset + # from gr00t_dreams.groot_configs import construct_modality_config_and_transforms + from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.groot_configs import ( + construct_modality_config_and_transforms, + ) + + self.dataset_path = [] + for p in dataset_path.split(","): + if (Path(p) / "data").exists() and (Path(p) / "meta").exists() and (Path(p) / "videos").exists(): + self.dataset_path.append(p) + else: + # This is not a LeRobot dataset, assume it's a directory of LeRobot datasets + for sub_p in Path(p).iterdir(): + if sub_p.is_dir(): + self.dataset_path.append(str(sub_p)) + + self.lerobot_datasets = [] + for p in self.dataset_path: + config, train_transform, test_transform = construct_modality_config_and_transforms( + num_frames=(num_frames + 1), embodiment=embodiment, downscaled_res=downscaled_res + ) # Add an additional prefix frame as baseline to compute delta actions + self.lerobot_datasets.append( + WrappedLeRobotSingleDataset( + dataset_path=p, + modality_configs=config, + transforms=train_transform if data_split == "train" or data_split == "full" else test_transform, + embodiment_tag="gr1_unified" if "gr1" in embodiment else embodiment, + data_split=data_split, + ) + ) + print(f"Loaded lerobot {data_split} dataset from {self.dataset_path} with {len(self)} samples.") + + # if height is not None and width is not None: + # print("Height and width are fixed. Setting `dynamic_resolution` to False.") + # self.dynamic_resolution = False + # elif height is None and width is None: + # print("Height and width are none. Setting `dynamic_resolution` to True.") + # self.dynamic_resolution = True + + def __getitem__(self, data_id): + data_id %= len(self) + for dataset in self.lerobot_datasets: + if data_id < len(dataset): + break + data_id -= len(dataset) + lerobot_data = dataset[data_id] + + prompt = lerobot_data["text"] + + video = lerobot_data["video"] + video_frames = [] + for i in range(1, video.shape[1]): # Skip first frame (used only as action baseline) + frame = video[:, i, :, :] + frame = Image.fromarray(frame.permute(1, 2, 0).numpy()) + video_frames.append(frame) + if len(video_frames) != self.num_frames: + print( + f"Warning: Expected {self.num_frames} frames, but got {len(video_frames)} frames. Randomly sampling an item instead." + ) + return self.__getitem__(random.randint(0, len(self) - 1)) # noqa: F821 + video_frames = np.stack([np.array(frame, dtype=np.uint8) for frame in video_frames]) + + # Cumulative baselined delta actions (old version) + # NOTE: Need to tweak this after (num_frames + 1) change + # delta_actions = lerobot_data["action"][1:] - lerobot_data["action"][[0]] + # Chunked cumulative baselined delta actions (for chunked action architecture) + actions = lerobot_data["action"] + delta_actions = [] + for t in range(1, len(actions) - 1, self.time_division_factor): + delta_actions.append(actions[t : t + self.time_division_factor] - actions[t - 1]) + delta_actions = torch.cat(delta_actions, dim=0) + + data = { + "prompt": prompt, + "video": torch.from_numpy(video_frames).permute(3, 0, 1, 2), + # "action": torch.from_numpy(delta_actions), + "action": (delta_actions), + "ai_caption": "", + "text": prompt, + "t5_text_embeddings": torch.zeros(512, 1024, dtype=torch.bfloat16).cuda(), + "t5_text_mask": torch.ones(512, dtype=torch.int64).cuda(), + "fps": 4, + "image_size": 256 * torch.ones(4).cuda(), + "num_frames": 13, + "padding_mask": torch.zeros(1, 256, 256).cuda(), + "__key__": lerobot_data["__key__"], + } + # data = { + # "prompt": prompt, + # "video": video_frames, + # "action": delta_actions, + # } + return data + + def __len__(self): + return sum([len(d) for d in self.lerobot_datasets]) * self.repeat diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/embodiment_tags.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/embodiment_tags.py new file mode 100644 index 0000000000000000000000000000000000000000..e31586fb9ad1c7127b0f895e71e0114ee594d618 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/embodiment_tags.py @@ -0,0 +1,50 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from enum import Enum + + +class EmbodimentTag(Enum): + GR1 = "gr1" + """ + The GR1 dataset. + """ + + GR1_unified = "gr1_unified" + """ + The GR1 unified dataset. + """ + + FRANKA = "franka" + """ + The FRANKA dataset. + """ + + SO100 = "so100" + """ + The SO100 dataset. + """ + + ROBOCASA = "robocasa_panda_omron" + """ + The ROBOCASA dataset. + """ + + NEW_EMBODIMENT = "new_embodiment" + """ + Any new embodiment for finetuning. + """ + + AGIBOT = "agibot" diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/schema.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/schema.py new file mode 100644 index 0000000000000000000000000000000000000000..297c3374fe0b0498cdf5f45d8b9ccbba22509b63 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/schema.py @@ -0,0 +1,228 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from enum import Enum +from typing import Optional + +import numpy as np +from numpy.typing import NDArray +from pydantic import BaseModel, Field, field_serializer, field_validator + +from .embodiment_tags import EmbodimentTag + +# Common schema + + +class RotationType(Enum): + """Type of rotation representation""" + + AXIS_ANGLE = "axis_angle" + QUATERNION = "quaternion" + ROTATION_6D = "rotation_6d" + MATRIX = "matrix" + EULER_ANGLES_RPY = "euler_angles_rpy" + EULER_ANGLES_RYP = "euler_angles_ryp" + EULER_ANGLES_PRY = "euler_angles_pry" + EULER_ANGLES_PYR = "euler_angles_pyr" + EULER_ANGLES_YRP = "euler_angles_yrp" + EULER_ANGLES_YPR = "euler_angles_ypr" + + +# LeRobot schema + + +class LeRobotModalityField(BaseModel): + """Metadata for a LeRobot modality field.""" + + original_key: Optional[str] = Field( + default=None, + description="The original key of the modality in the LeRobot dataset", + ) + + +class LeRobotStateActionMetadata(LeRobotModalityField): + """Metadata for a LeRobot modality.""" + + start: int = Field( + ..., + description="The start index of the modality in the concatenated state/action vector", + ) + end: int = Field( + ..., + description="The end index of the modality in the concatenated state/action vector", + ) + rotation_type: Optional[RotationType] = Field(default=None, description="The type of rotation for the modality") + absolute: bool = Field(default=True, description="Whether the modality is absolute") + dtype: str = Field( + default="float64", + description="The data type of the modality. Defaults to float64.", + ) + range: Optional[tuple[float, float]] = Field( + default=None, + description="The range of the modality, if applicable. Defaults to None.", + ) + original_key: Optional[str] = Field( + default=None, + description="The original key of the modality in the LeRobot dataset.", + ) + + +class LeRobotStateMetadata(LeRobotStateActionMetadata): + """Metadata for a LeRobot state modality.""" + + original_key: Optional[str] = Field( + default="observation.state", # LeRobot convention for states + description="The original key of the state modality in the LeRobot dataset", + ) + + +class LeRobotActionMetadata(LeRobotStateActionMetadata): + """Metadata for a LeRobot action modality.""" + + original_key: Optional[str] = Field( + default="action", # LeRobot convention for actions + description="The original key of the action modality in the LeRobot dataset", + ) + + +class LeRobotModalityMetadata(BaseModel): + """Metadata for a LeRobot modality.""" + + state: dict[str, LeRobotStateMetadata] = Field( + ..., + description="The metadata for the state modality. The keys are the names of each split of the state vector.", + ) + action: dict[str, LeRobotActionMetadata] = Field( + ..., + description="The metadata for the action modality. The keys are the names of each split of the action vector.", + ) + video: dict[str, LeRobotModalityField] = Field( + ..., + description="The metadata for the video modality. The keys are the new names of each video modality.", + ) + annotation: Optional[dict[str, LeRobotModalityField]] = Field( + default=None, + description="The metadata for the annotation modality. The keys are the new names of each annotation modality.", + ) + + def get_key_meta(self, key: str) -> LeRobotModalityField: + """Get the metadata for a key in the LeRobot modality metadata. + + Args: + key (str): The key to get the metadata for. + + Returns: + LeRobotModalityField: The metadata for the key. + + Example: + lerobot_modality_meta = LeRobotModalityMetadata.model_validate(U.load_json(modality_meta_path)) + lerobot_modality_meta.get_key_meta("state.joint_shoulder_y") + lerobot_modality_meta.get_key_meta("video.main_camera") + lerobot_modality_meta.get_key_meta("annotation.human.action.task_description") + """ + split_key = key.split(".") + modality = split_key[0] + subkey = ".".join(split_key[1:]) + if modality == "state": + if subkey not in self.state: + raise ValueError( + f"Key: {key}, state key {subkey} not found in metadata, available state keys: {self.state.keys()}" + ) + return self.state[subkey] + elif modality == "action": + if subkey not in self.action: + raise ValueError( + f"Key: {key}, action key {subkey} not found in metadata, available action keys: {self.action.keys()}" + ) + return self.action[subkey] + elif modality == "video": + if subkey not in self.video: + raise ValueError( + f"Key: {key}, video key {subkey} not found in metadata, available video keys: {self.video.keys()}" + ) + return self.video[subkey] + elif modality == "annotation": + assert self.annotation is not None, "Trying to get annotation metadata for a dataset with no annotations" + if subkey not in self.annotation: + raise ValueError( + f"Key: {key}, annotation key {subkey} not found in metadata, available annotation keys: {self.annotation.keys()}" + ) + return self.annotation[subkey] + else: + raise ValueError(f"Key: {key}, unexpected modality: {modality}") + + +# Dataset schema (parsed from LeRobot schema and simplified) + + +class DatasetStatisticalValues(BaseModel): + model_config = {"arbitrary_types_allowed": True} + + max: NDArray = Field(..., description="Maximum values") + min: NDArray = Field(..., description="Minimum values") + mean: NDArray = Field(..., description="Mean values") + std: NDArray = Field(..., description="Standard deviation") + q01: NDArray = Field(..., description="1st percentile values") + q99: NDArray = Field(..., description="99th percentile values") + + @field_validator("*", mode="before") + @classmethod + def convert_list_to_ndarray(cls, v): + """Convert lists to numpy arrays when loading from JSON.""" + if isinstance(v, list): + return np.array(v) + return v + + @field_serializer("*", when_used="json") + def serialize_ndarray(self, v: NDArray) -> list[float]: + return v.tolist() # type: ignore + + +class DatasetStatistics(BaseModel): + state: dict[str, DatasetStatisticalValues] = Field(..., description="Statistics of the state") + action: dict[str, DatasetStatisticalValues] = Field(..., description="Statistics of the action") + + +class VideoMetadata(BaseModel): + """Metadata of the video modality""" + + resolution: tuple[int, int] = Field(..., description="Resolution of the video") + channels: int = Field(..., description="Number of channels in the video", gt=0) + fps: float = Field(..., description="Frames per second", gt=0) + + +class StateActionMetadata(BaseModel): + absolute: bool = Field(..., description="Whether the state or action is absolute") + rotation_type: Optional[RotationType] = Field(None, description="Type of rotation, if any") + shape: tuple[int, ...] = Field(..., description="Shape of the state or action") + continuous: bool = Field(..., description="Whether the state or action is continuous") + + +class DatasetModalities(BaseModel): + video: dict[str, VideoMetadata] = Field(..., description="Metadata of the video") + state: dict[str, StateActionMetadata] = Field(..., description="Metadata of the state") + action: dict[str, StateActionMetadata] = Field(..., description="Metadata of the action") + + +class DatasetMetadata(BaseModel): + """Metadata of the trainable dataset + + Changes: + - Update to use the new RawCommitHashMetadataMetadata_V1_2 + """ + + statistics: DatasetStatistics = Field(..., description="Statistics of the dataset") + modalities: DatasetModalities = Field(..., description="Metadata of the modalities") + embodiment_tag: EmbodimentTag = Field(..., description="Embodiment tag of the dataset") diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..9921f74b18fbab4679a530f6a7f32394e2662bfa --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/__init__.py @@ -0,0 +1,37 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# from .base import ComposedModalityTransform, InvertibleModalityTransform, ModalityTransform +# from .concat import ConcatTransform +# from .state_action import ( +# StateActionDropout, +# StateActionPerturbation, +# StateActionSinCosTransform, +# StateActionToTensor, +# StateActionTransform, +# ) +# from .video import ( +# VideoColorJitter, +# VideoCrop, +# VideoGrayscale, +# VideoHorizontalFlip, +# VideoRandomGrayscale, +# VideoRandomPosterize, +# VideoRandomRotation, +# VideoResize, +# VideoToNumpy, +# VideoToTensor, +# VideoTransform, +# ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/base.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/base.py new file mode 100644 index 0000000000000000000000000000000000000000..66a299532b191665edea73dc0d31021073b64c51 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/base.py @@ -0,0 +1,128 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from abc import ABC, abstractmethod +from typing import Any + +from pydantic import BaseModel, ConfigDict, Field, PrivateAttr + +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.schema import DatasetMetadata + + +class ModalityTransform(BaseModel, ABC): + """ + Abstract class for transforming data modalities, e.g. video frame augmentation or action normalization. + """ + + apply_to: list[str] = Field(..., description="The keys to apply the transform to.") + training: bool = Field(default=True, description="Whether to apply the transform in training mode.") + _dataset_metadata: DatasetMetadata | None = PrivateAttr(default=None) + + model_config = ConfigDict(arbitrary_types_allowed=True) + + @property + def dataset_metadata(self) -> DatasetMetadata: + assert self._dataset_metadata is not None, ( + "Dataset metadata is not set. Please call set_metadata() before calling apply()." + ) + return self._dataset_metadata + + @dataset_metadata.setter + def dataset_metadata(self, value: DatasetMetadata): + self._dataset_metadata = value + + def set_metadata(self, dataset_metadata: DatasetMetadata): + """ + Set the dataset metadata. This is useful for transforms that need to know the dataset metadata, e.g. to normalize actions. + Subclasses can override this method if they need to do something more complex. + """ + self.dataset_metadata = dataset_metadata + + def __call__(self, data: dict[str, Any]) -> dict[str, Any]: + """Apply the transformation to the data corresponding to target_keys and return the processed data. + + Args: + data (dict[str, Any]): The data to transform. + example: data = { + "video.image_side_0": np.ndarray, + "action.eef_position": np.ndarray, + ... + } + + Returns: + dict[str, Any]: The transformed data. + example: transformed_data = { + "video.image_side_0": np.ndarray, + "action.eef_position": torch.Tensor, # Normalized and converted to tensor + ... + } + """ + return self.apply(data) + + @abstractmethod + def apply(self, data: dict[str, Any]) -> dict[str, Any]: + """Apply the transformation to the data corresponding to keys matching the `apply_to` regular expression and return the processed data.""" + + def train(self): + self.training = True + + def eval(self): + self.training = False + + +class InvertibleModalityTransform(ModalityTransform): + @abstractmethod + def unapply(self, data: dict[str, Any]) -> dict[str, Any]: + """Reverse the transformation to the data corresponding to keys matching the `apply_to` regular expression and return the processed data.""" + + +class ComposedModalityTransform(ModalityTransform): + """Compose multiple modality transforms.""" + + transforms: list[ModalityTransform] = Field(..., description="The transforms to compose.") + apply_to: list[str] = Field(default_factory=list, description="Will be ignored for composed transforms.") + training: bool = Field(default=True, description="Whether to apply the transform in training mode.") + + model_config = ConfigDict(arbitrary_types_allowed=True, from_attributes=True) + + def set_metadata(self, dataset_metadata: DatasetMetadata): + for transform in self.transforms: + transform.set_metadata(dataset_metadata) + + def apply(self, data: dict[str, Any]) -> dict[str, Any]: + for i, transform in enumerate(self.transforms): + try: + data = transform(data) + except Exception as e: + raise ValueError(f"Error applying transform {i} to data: {e}") from e + return data + + def unapply(self, data: dict[str, Any]) -> dict[str, Any]: + for i, transform in enumerate(reversed(self.transforms)): + if isinstance(transform, InvertibleModalityTransform): + try: + data = transform.unapply(data) + except Exception as e: + step = len(self.transforms) - i - 1 + raise ValueError(f"Error unapplying transform {step} to data: {e}") from e + return data + + def train(self): + for transform in self.transforms: + transform.train() + + def eval(self): + for transform in self.transforms: + transform.eval() diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/concat.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/concat.py new file mode 100644 index 0000000000000000000000000000000000000000..8db83b8b013e95dd3aea0de80f91ca4c9d51788d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/concat.py @@ -0,0 +1,205 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Optional + +import numpy as np +import torch +from pydantic import Field + +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.schema import DatasetMetadata, StateActionMetadata +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.transform.base import InvertibleModalityTransform + + +class ConcatTransform(InvertibleModalityTransform): + """ + Concatenate the keys according to specified order. + """ + + # -- We inherit from ModalityTransform, so we keep apply_to as well -- + apply_to: list[str] = Field(default_factory=list, description="Not used in this transform, kept for compatibility.") + + video_concat_order: list[str] = Field( + ..., + description="Concatenation order for each video modality. Format: ['video.ego_view_pad_res224_freq20', ...]", + ) + + state_concat_order: Optional[list[str]] = Field( + default=None, + description="Concatenation order for each state modality. Format: ['state.position', 'state.velocity', ...].", + ) + + action_concat_order: Optional[list[str]] = Field( + default=None, + description="Concatenation order for each action modality. " + "Format: ['action.position', 'action.velocity', ...].", + ) + + action_dims: dict[str, int] = Field( + default_factory=dict, + description="The dimensions of the action keys.", + ) + state_dims: dict[str, int] = Field( + default_factory=dict, + description="The dimensions of the state keys.", + ) + + def model_dump(self, *args, **kwargs): + if kwargs.get("mode", "python") == "json": + include = { + "apply_to", + "video_concat_order", + "state_concat_order", + "action_concat_order", + } + else: + include = kwargs.pop("include", None) + + return super().model_dump(*args, include=include, **kwargs) + + def apply(self, data: dict) -> dict: + grouped_keys = {} + for key in data.keys(): + try: + modality, _ = key.split(".") + except: # noqa: E722 + ### Handle language annotation special case + if "annotation" in key: + modality = "language" + else: + modality = "others" + if modality not in grouped_keys: + grouped_keys[modality] = [] + grouped_keys[modality].append(key) + + if "video" in grouped_keys: + # Check if keys in video_concat_order, state_concat_order, action_concat_order are + # ineed contained in the data. If not, then the keys are misspecified + video_keys = grouped_keys["video"] + assert self.video_concat_order is not None, f"{self.video_concat_order=}, {video_keys=}" + assert all(item in video_keys for item in self.video_concat_order), ( + f"keys in video_concat_order are misspecified, \n{video_keys=}, \n{self.video_concat_order=}" + ) + + # Process each video view + unsqueezed_videos = [] + for video_key in self.video_concat_order: + video_data = data.pop(video_key) + unsqueezed_video = np.expand_dims(video_data, axis=-4) # [..., H, W, C] -> [..., 1, H, W, C] + unsqueezed_videos.append(unsqueezed_video) + # Concatenate along the new axis + unsqueezed_video = np.concatenate(unsqueezed_videos, axis=-4) # [..., V, H, W, C] + + # Video + data["video"] = unsqueezed_video + + # "state" + if "state" in grouped_keys: + state_keys = grouped_keys["state"] + assert self.state_concat_order is not None, f"{self.state_concat_order=}" + assert all(item in state_keys for item in self.state_concat_order), ( + f"keys in state_concat_order are misspecified, \n{state_keys=}, \n{self.state_concat_order=}" + ) + # Check the state dims + for key in self.state_concat_order: + target_shapes = [self.state_dims[key]] + if self.is_rotation_key(key): + target_shapes.append(6) # Allow for rotation_6d + # if key in ["state.right_arm", "state.right_hand"]: + target_shapes.append(self.state_dims[key] * 2) # Allow for sin-cos transform + assert data[key].shape[-1] in target_shapes, ( + f"State dim mismatch for {key=}, {data[key].shape[-1]=}, {target_shapes=}" + ) + # Concatenate the state keys + # We'll have StateActionToTensor before this transform, so here we use torch.cat + data["state"] = torch.cat([data.pop(key) for key in self.state_concat_order], dim=-1) # [T, D_state] + + if "action" in grouped_keys: + action_keys = grouped_keys["action"] + assert self.action_concat_order is not None, f"{self.action_concat_order=}" + # Check if all keys in concat_order are present + assert set(self.action_concat_order) == set(action_keys), ( + f"{set(self.action_concat_order)=}, {set(action_keys)=}" + ) + # Record the action dims + for key in self.action_concat_order: + target_shapes = [self.action_dims[key]] + if self.is_rotation_key(key): + target_shapes.append(3) # Allow for axis angle + assert self.action_dims[key] == data[key].shape[-1], ( + f"Action dim mismatch for {key=}, {self.action_dims[key]=}, {data[key].shape[-1]=}" + ) + # Concatenate the action keys + # We'll have StateActionToTensor before this transform, so here we use torch.cat + data["action"] = torch.cat([data.pop(key) for key in self.action_concat_order], dim=-1) # [T, D_action] + + return data + + def unapply(self, data: dict) -> dict: + start_dim = 0 + assert "action" in data, f"{data.keys()=}" + # For those dataset without actions (LAPA), we'll never run unapply + assert self.action_concat_order is not None, f"{self.action_concat_order=}" + action_tensor = data.pop("action") + for key in self.action_concat_order: + if key not in self.action_dims: + raise ValueError(f"Action dim {key} not found in action_dims.") + end_dim = start_dim + self.action_dims[key] + data[key] = action_tensor[..., start_dim:end_dim] + start_dim = end_dim + if "state" in data: + assert self.state_concat_order is not None, f"{self.state_concat_order=}" + start_dim = 0 + state_tensor = data.pop("state") + for key in self.state_concat_order: + end_dim = start_dim + self.state_dims[key] + data[key] = state_tensor[..., start_dim:end_dim] + start_dim = end_dim + return data + + def __call__(self, data: dict) -> dict: + return self.apply(data) + + def get_modality_metadata(self, key: str) -> StateActionMetadata: + modality, subkey = key.split(".") + assert self.dataset_metadata is not None, "Metadata not set" + modality_config = getattr(self.dataset_metadata.modalities, modality) + assert subkey in modality_config, f"{subkey=} not found in {modality_config=}" + assert isinstance(modality_config[subkey], StateActionMetadata), ( + f"Expected {StateActionMetadata} for {subkey=}, got {type(modality_config[subkey])=}" + ) + return modality_config[subkey] + + def get_state_action_dims(self, key: str) -> int: + """Get the dimension of a state or action key from the dataset metadata.""" + modality_config = self.get_modality_metadata(key) + shape = modality_config.shape + assert len(shape) == 1, f"{shape=}" + return shape[0] + + def is_rotation_key(self, key: str) -> bool: + modality_config = self.get_modality_metadata(key) + return modality_config.rotation_type is not None + + def set_metadata(self, dataset_metadata: DatasetMetadata): + """Set the metadata and compute the dimensions of the state and action keys.""" + super().set_metadata(dataset_metadata) + # Pre-compute the dimensions of the state and action keys + if self.action_concat_order is not None: + for key in self.action_concat_order: + self.action_dims[key] = self.get_state_action_dims(key) + if self.state_concat_order is not None: + for key in self.state_concat_order: + self.state_dims[key] = self.get_state_action_dims(key) diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/state_action.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/state_action.py new file mode 100644 index 0000000000000000000000000000000000000000..5774900152b83cc6991398c99944c38842b0e20e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/state_action.py @@ -0,0 +1,579 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import functools +import random +from typing import Any, ClassVar + +import numpy as np + +# import pytorch3d.transforms as pt +import torch +from pydantic import Field, PrivateAttr, field_validator, model_validator + +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.schema import ( + DatasetMetadata, + RotationType, + StateActionMetadata, +) +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.transform.base import ( + InvertibleModalityTransform, + ModalityTransform, +) + + +class RotationTransform: + """Adapted from https://github.com/real-stanford/diffusion_policy/blob/548a52bbb105518058e27bf34dcf90bf6f73681a/diffusion_policy/model/common/rotation_transformer.py""" + + valid_reps = ["axis_angle", "euler_angles", "quaternion", "rotation_6d", "matrix"] + + def __init__(self, from_rep="axis_angle", to_rep="rotation_6d"): + """ + Valid representations + + Always use matrix as intermediate representation. + """ + if from_rep.startswith("euler_angles"): + from_convention = from_rep.split("_")[-1] + from_rep = "euler_angles" + from_convention = from_convention.replace("r", "X").replace("p", "Y").replace("y", "Z") + else: + from_convention = None + if to_rep.startswith("euler_angles"): + to_convention = to_rep.split("_")[-1] + to_rep = "euler_angles" + to_convention = to_convention.replace("r", "X").replace("p", "Y").replace("y", "Z") + else: + to_convention = None + assert from_rep != to_rep, f"from_rep and to_rep cannot be the same: {from_rep}" + assert from_rep in self.valid_reps, f"Invalid from_rep: {from_rep}" + assert to_rep in self.valid_reps, f"Invalid to_rep: {to_rep}" + + forward_funcs = list() + inverse_funcs = list() + + if from_rep != "matrix": + funcs = [getattr(pt, f"{from_rep}_to_matrix"), getattr(pt, f"matrix_to_{from_rep}")] # noqa: F821 + if from_convention is not None: + funcs = [functools.partial(func, convention=from_convention) for func in funcs] + forward_funcs.append(funcs[0]) + inverse_funcs.append(funcs[1]) + + if to_rep != "matrix": + funcs = [getattr(pt, f"matrix_to_{to_rep}"), getattr(pt, f"{to_rep}_to_matrix")] # noqa: F821 + if to_convention is not None: + funcs = [functools.partial(func, convention=to_convention) for func in funcs] + forward_funcs.append(funcs[0]) + inverse_funcs.append(funcs[1]) + + inverse_funcs = inverse_funcs[::-1] + + self.forward_funcs = forward_funcs + self.inverse_funcs = inverse_funcs + + @staticmethod + def _apply_funcs(x: torch.Tensor, funcs: list) -> torch.Tensor: + assert isinstance(x, torch.Tensor) + for func in funcs: + x = func(x) + return x + + def forward(self, x: torch.Tensor) -> torch.Tensor: + assert isinstance(x, torch.Tensor), f"Unexpected input type: {type(x)}. Expected type: {torch.Tensor}" + return self._apply_funcs(x, self.forward_funcs) + + def inverse(self, x: torch.Tensor) -> torch.Tensor: + assert isinstance(x, torch.Tensor), f"Unexpected input type: {type(x)}. Expected type: {torch.Tensor}" + return self._apply_funcs(x, self.inverse_funcs) + + +class Normalizer: + valid_modes = ["q99", "mean_std", "min_max", "binary"] + + def __init__(self, mode: str, statistics: dict): + self.mode = mode + self.statistics = statistics + for key, value in self.statistics.items(): + self.statistics[key] = torch.tensor(value) + + def forward(self, x: torch.Tensor) -> torch.Tensor: + assert isinstance(x, torch.Tensor), f"Unexpected input type: {type(x)}. Expected type: {torch.Tensor}" + + # Normalize the tensor + if self.mode == "q99": + # Range of q99 is [-1, 1] + q01 = self.statistics["q01"].to(x.dtype) + q99 = self.statistics["q99"].to(x.dtype) + + # In the case of q01 == q99, the normalization will be undefined + # So we set the normalized values to the original values + mask = q01 != q99 + normalized = torch.zeros_like(x) + + # Normalize the values where q01 != q99 + # Formula: 2 * (x - q01) / (q99 - q01) - 1 + normalized[..., mask] = (x[..., mask] - q01[..., mask]) / (q99[..., mask] - q01[..., mask]) + normalized[..., mask] = 2 * normalized[..., mask] - 1 + + # Set the normalized values to the original values where q01 == q99 + normalized[..., ~mask] = x[..., ~mask].to(x.dtype) + + # Clip the normalized values to be between -1 and 1 + normalized = torch.clamp(normalized, -1, 1) + + elif self.mode == "mean_std": + # Range of mean_std is not fixed, but can be positive or negative + mean = self.statistics["mean"].to(x.dtype) + std = self.statistics["std"].to(x.dtype) + + # In the case of std == 0, the normalization will be undefined + # So we set the normalized values to the original values + mask = std != 0 + normalized = torch.zeros_like(x) + + # Normalize the values where std != 0 + # Formula: (x - mean) / std + normalized[..., mask] = (x[..., mask] - mean[..., mask]) / std[..., mask] + + # Set the normalized values to the original values where std == 0 + normalized[..., ~mask] = x[..., ~mask].to(x.dtype) + + elif self.mode == "min_max": + # Range of min_max is [-1, 1] + min = self.statistics["min"].to(x.dtype) + max = self.statistics["max"].to(x.dtype) + + # In the case of min == max, the normalization will be undefined + # So we set the normalized values to the original values + mask = min != max + normalized = torch.zeros_like(x) + + # Normalize the values where min != max + # Formula: 2 * (x - min) / (max - min) - 1 + normalized[..., mask] = (x[..., mask] - min[..., mask]) / (max[..., mask] - min[..., mask]) + normalized[..., mask] = 2 * normalized[..., mask] - 1 + + # Set the normalized values to the original values where min == max + # normalized[..., ~mask] = x[..., ~mask].to(x.dtype) + # Set the normalized values to 0 where min == max + normalized[..., ~mask] = 0 + + elif self.mode == "scale": + # Range of scale is [0, 1] + min = self.statistics["min"].to(x.dtype) + max = self.statistics["max"].to(x.dtype) + abs_max = torch.max(torch.abs(min), torch.abs(max)) + mask = abs_max != 0 + normalized = torch.zeros_like(x) + normalized[..., mask] = x[..., mask] / abs_max[..., mask] + normalized[..., ~mask] = 0 + + elif self.mode == "binary": + # Range of binary is [0, 1] + normalized = (x > 0.5).to(x.dtype) + else: + raise ValueError(f"Invalid normalization mode: {self.mode}") + + return normalized + + def inverse(self, x: torch.Tensor) -> torch.Tensor: + assert isinstance(x, torch.Tensor), f"Unexpected input type: {type(x)}. Expected type: {torch.Tensor}" + if self.mode == "q99": + q01 = self.statistics["q01"].to(x.dtype) + q99 = self.statistics["q99"].to(x.dtype) + return (x + 1) / 2 * (q99 - q01) + q01 + elif self.mode == "mean_std": + mean = self.statistics["mean"].to(x.dtype) + std = self.statistics["std"].to(x.dtype) + return x * std + mean + elif self.mode == "min_max": + min = self.statistics["min"].to(x.dtype) + max = self.statistics["max"].to(x.dtype) + return (x + 1) / 2 * (max - min) + min + elif self.mode == "binary": + return (x > 0.5).to(x.dtype) + else: + raise ValueError(f"Invalid normalization mode: {self.mode}") + + +class StateActionToTensor(InvertibleModalityTransform): + """ + Transforms states and actions to tensors. + """ + + input_dtypes: dict[str, np.dtype] = Field(default_factory=dict, description="The input dtypes for each state key.") + output_dtypes: dict[str, torch.dtype] = Field( + default_factory=dict, description="The output dtypes for each state key." + ) + + def model_dump(self, *args, **kwargs): + if kwargs.get("mode", "python") == "json": + include = {"apply_to"} + else: + include = kwargs.pop("include", None) + + return super().model_dump(*args, include=include, **kwargs) + + @field_validator("input_dtypes", "output_dtypes", mode="before") + def validate_dtypes(cls, v): + for key, dtype in v.items(): + if isinstance(dtype, str): + if dtype.startswith("torch."): + dtype_split = dtype.split(".")[-1] + v[key] = getattr(torch, dtype_split) + elif dtype.startswith("np.") or dtype.startswith("numpy."): + dtype_split = dtype.split(".")[-1] + v[key] = np.dtype(dtype_split) + else: + raise ValueError(f"Invalid dtype: {dtype}") + return v + + def apply(self, data: dict[str, Any]) -> dict[str, Any]: + for key in self.apply_to: + if key not in data: + continue + value = data[key] + assert isinstance(value, np.ndarray), f"Unexpected input type: {type(value)}. Expected type: {np.ndarray}" + data[key] = torch.from_numpy(value) + if key in self.output_dtypes: + data[key] = data[key].to(self.output_dtypes[key]) + return data + + def unapply(self, data: dict[str, Any]) -> dict[str, Any]: + for key in self.apply_to: + if key not in data: + continue + value = data[key] + assert isinstance(value, torch.Tensor), ( + f"Unexpected input type: {type(value)}. Expected type: {torch.Tensor}" + ) + data[key] = value.numpy() + if key in self.input_dtypes: + data[key] = data[key].astype(self.input_dtypes[key]) + return data + + +class StateActionTransform(InvertibleModalityTransform): + """ + Class for state or action transform. + + Args: + apply_to (list[str]): The keys in the modality to load and transform. + normalization_modes (dict[str, str]): The normalization modes for each state key. + If a state key in apply_to is not present in the dictionary, it will not be normalized. + target_rotations (dict[str, str]): The target representations for each state key. + If a state key in apply_to is not present in the dictionary, it will not be rotated. + """ + + # Configurable attributes + apply_to: list[str] = Field(..., description="The keys in the modality to load and transform.") + normalization_modes: dict[str, str] = Field( + default_factory=dict, description="The normalization modes for each state key." + ) + target_rotations: dict[str, str] = Field( + default_factory=dict, description="The target representations for each state key." + ) + normalization_statistics: dict[str, dict] = Field( + default_factory=dict, description="The statistics for each state key." + ) + modality_metadata: dict[str, StateActionMetadata] = Field( + default_factory=dict, description="The modality metadata for each state key." + ) + + # Model variables + _rotation_transformers: dict[str, RotationTransform] = PrivateAttr(default_factory=dict) + _normalizers: dict[str, Normalizer] = PrivateAttr(default_factory=dict) + _input_dtypes: dict[str, np.dtype | torch.dtype] = PrivateAttr(default_factory=dict) + + # Model constants + _DEFAULT_MIN_MAX_STATISTICS: ClassVar[dict] = { + "rotation_6d": { + "min": [-1, -1, -1, -1, -1, -1], + "max": [1, 1, 1, 1, 1, 1], + }, + "euler_angles": { + "min": [-np.pi, -np.pi, -np.pi], + "max": [np.pi, np.pi, np.pi], + }, + "quaternion": { + "min": [-1, -1, -1, -1], + "max": [1, 1, 1, 1], + }, + "axis_angle": { + "min": [-np.pi, -np.pi, -np.pi], + "max": [np.pi, np.pi, np.pi], + }, + } + + def model_dump(self, *args, **kwargs): + if kwargs.get("mode", "python") == "json": + include = {"apply_to", "normalization_modes", "target_rotations"} + else: + include = kwargs.pop("include", None) + + return super().model_dump(*args, include=include, **kwargs) + + @field_validator("modality_metadata", mode="before") + def validate_modality_metadata(cls, v): + for modality_key, config in v.items(): + if isinstance(config, dict): + config = StateActionMetadata.model_validate(config) + else: + assert isinstance(config, StateActionMetadata), f"Invalid source rotation config: {config}" + v[modality_key] = config + return v + + @model_validator(mode="after") + def validate_normalization_statistics(self): + for modality_key, normalization_statistics in self.normalization_statistics.items(): + if modality_key in self.normalization_modes: + normalization_mode = self.normalization_modes[modality_key] + if normalization_mode == "min_max": + assert "min" in normalization_statistics and "max" in normalization_statistics, ( + f"Min and max statistics are required for min_max normalization, but got {normalization_statistics}" + ) + assert len(normalization_statistics["min"]) == len(normalization_statistics["max"]), ( + f"Min and max statistics must have the same length, but got {normalization_statistics['min']} and {normalization_statistics['max']}" + ) + elif normalization_mode == "mean_std": + assert "mean" in normalization_statistics and "std" in normalization_statistics, ( + f"Mean and std statistics are required for mean_std normalization, but got {normalization_statistics}" + ) + assert len(normalization_statistics["mean"]) == len(normalization_statistics["std"]), ( + f"Mean and std statistics must have the same length, but got {normalization_statistics['mean']} and {normalization_statistics['std']}" + ) + elif normalization_mode == "q99": + assert "q01" in normalization_statistics and "q99" in normalization_statistics, ( + f"q01 and q99 statistics are required for q99 normalization, but got {normalization_statistics}" + ) + assert len(normalization_statistics["q01"]) == len(normalization_statistics["q99"]), ( + f"q01 and q99 statistics must have the same length, but got {normalization_statistics['q01']} and {normalization_statistics['q99']}" + ) + elif normalization_mode == "binary": + assert len(normalization_statistics) == 1, ( + f"Binary normalization should only have one value, but got {normalization_statistics}" + ) + assert normalization_statistics[0] in [ + 0, + 1, + ], f"Binary normalization should only have 0 or 1, but got {normalization_statistics[0]}" + else: + raise ValueError(f"Invalid normalization mode: {normalization_mode}") + return self + + def set_metadata(self, dataset_metadata: DatasetMetadata): + dataset_statistics = dataset_metadata.statistics + modality_metadata = dataset_metadata.modalities + + # Check that all state keys specified in apply_to have their modality_metadata + for key in self.apply_to: + split_key = key.split(".") + assert len(split_key) == 2, "State keys should have two parts: 'modality.key'" + if key not in self.modality_metadata: + modality, state_key = split_key + assert hasattr(modality_metadata, modality), f"{modality} config not found" + assert state_key in getattr(modality_metadata, modality), f"{state_key} config not found" + self.modality_metadata[key] = getattr(modality_metadata, modality)[state_key] + + # Check that all state keys specified in normalization_modes have their statistics in state_statistics + for key in self.normalization_modes: + split_key = key.split(".") + assert len(split_key) == 2, "State keys should have two parts: 'modality.key'" + modality, state_key = split_key + assert hasattr(dataset_statistics, modality), f"{modality} statistics not found" + assert state_key in getattr(dataset_statistics, modality), f"{state_key} statistics not found" + assert len(getattr(modality_metadata, modality)[state_key].shape) == 1, ( + f"{getattr(modality_metadata, modality)[state_key].shape=}" + ) + self.normalization_statistics[key] = getattr(dataset_statistics, modality)[state_key].model_dump() + + # Initialize the rotation transformers + for key in self.target_rotations: + # Get the original representation of the state + from_rep = self.modality_metadata[key].rotation_type + assert from_rep is not None, f"Source rotation type not found for {key}" + + # Get the target representation of the state, will raise an error if the target representation is not valid + to_rep = RotationType(self.target_rotations[key]) + + # If the original representation is not the same as the target representation, initialize the rotation transformer + if from_rep != to_rep: + self._rotation_transformers[key] = RotationTransform(from_rep=from_rep.value, to_rep=to_rep.value) + + # Initialize the normalizers + for key in self.normalization_modes: + modality, state_key = key.split(".") + # If the state has a nontrivial rotation, we need to handle it more carefully + # For absolute rotations, we need to convert them to the target representation and normalize them using min_max mode, + # since we can infer the bounds by the representation + # For relative rotations, we cannot normalize them as we don't know the bounds + if key in self._rotation_transformers: + # Case 1: Absolute rotation + if self.modality_metadata[key].absolute: + # Check that the normalization mode is valid + assert self.normalization_modes[key] == "min_max", ( + "Absolute rotations that are converted to other formats must be normalized using `min_max` mode" + ) + rotation_type = RotationType(self.target_rotations[key]).value + # If the target representation is euler angles, we need to parse the convention + if rotation_type.startswith("euler_angles"): + rotation_type = "euler_angles" + # Get the statistics for the target representation + statistics = self._DEFAULT_MIN_MAX_STATISTICS[rotation_type] + # Case 2: Relative rotation + else: + raise ValueError( + f"Cannot normalize relative rotations: {key} that's converted to {self.target_rotations[key]}" + ) + # If the state is not continuous, we should not use normalization modes other than binary + elif not self.modality_metadata[key].continuous and self.normalization_modes[key] != "binary": + raise ValueError(f"{key} is not continuous, so it should be normalized using `binary` mode") + # Initialize the normalizer + else: + statistics = self.normalization_statistics[key] + self._normalizers[key] = Normalizer(mode=self.normalization_modes[key], statistics=statistics) + + def apply(self, data: dict[str, Any]) -> dict[str, Any]: + for key in self.apply_to: + if key not in data: + # We allow some keys to be missing in the data, and only process the keys that are present + continue + if key not in self._input_dtypes: + input_dtype = data[key].dtype + assert isinstance(input_dtype, torch.dtype), ( + f"Unexpected input dtype: {input_dtype}. Expected type: {torch.dtype}" + ) + self._input_dtypes[key] = input_dtype + else: + assert data[key].dtype == self._input_dtypes[key], ( + f"All states corresponding to the same key must be of the same dtype, input dtype: {data[key].dtype}, expected dtype: {self._input_dtypes[key]}" + ) + # Rotate the state + state = data[key] + if key in self._rotation_transformers: + state = self._rotation_transformers[key].forward(state) + # Normalize the state + if key in self._normalizers: + state = self._normalizers[key].forward(state) + data[key] = state + return data + + def unapply(self, data: dict[str, Any]) -> dict[str, Any]: + for key in self.apply_to: + if key not in data: + continue + state = data[key] + assert isinstance(state, torch.Tensor), ( + f"Unexpected state type: {type(state)}. Expected type: {torch.Tensor}" + ) + # Unnormalize the state + if key in self._normalizers: + state = self._normalizers[key].inverse(state) + # Change the state back to its original representation + if key in self._rotation_transformers: + state = self._rotation_transformers[key].inverse(state) + assert isinstance(state, torch.Tensor), ( + f"State should be tensor after unapplying transformations, but got {type(state)}" + ) + # Only convert back to the original dtype if it's known, i.e. `apply` was called before + # If not, we don't know the original dtype, so we don't convert + if key in self._input_dtypes: + original_dtype = self._input_dtypes[key] + if isinstance(original_dtype, np.dtype): + state = state.numpy().astype(original_dtype) + elif isinstance(original_dtype, torch.dtype): + state = state.to(original_dtype) + else: + raise ValueError(f"Invalid input dtype: {original_dtype}") + data[key] = state + return data + + +class StateActionPerturbation(ModalityTransform): + """ + Class for state or action perturbation. + + Args: + apply_to (list[str]): The keys in the modality to load and transform. + std (float): Standard deviation of the noise to be added to the state or action. + """ + + # Configurable attributes + std: float = Field(..., description="Standard deviation of the noise to be added to the state or action.") + + def apply(self, data: dict[str, Any]) -> dict[str, Any]: + if not self.training: + # Don't perturb the data in eval mode + return data + if self.std < 0: + # If the std is negative, we don't add any noise + return data + for key in self.apply_to: + state = data[key] + assert isinstance(state, torch.Tensor) + transformed_data_min = torch.min(state) + transformed_data_max = torch.max(state) + noise = torch.randn_like(state) * self.std + state += noise + # Clip to the original range + state = torch.clamp(state, transformed_data_min, transformed_data_max) + data[key] = state + return data + + +class StateActionDropout(ModalityTransform): + """ + Class for state or action dropout. + + Args: + apply_to (list[str]): The keys in the modality to load and transform. + dropout_prob (float): Probability of dropping out a state or action. + """ + + # Configurable attributes + dropout_prob: float = Field(..., description="Probability of dropping out a state or action.") + + def apply(self, data: dict[str, Any]) -> dict[str, Any]: + if not self.training: + # Don't drop out the data in eval mode + return data + if self.dropout_prob < 0: + # If the dropout probability is negative, we don't drop out any states + return data + if self.dropout_prob > 1e-9 and random.random() < self.dropout_prob: + for key in self.apply_to: + state = data[key] + assert isinstance(state, torch.Tensor) + state = torch.zeros_like(state) + data[key] = state + return data + + +class StateActionSinCosTransform(ModalityTransform): + """ + Class for state or action sin-cos transform. + + Args: + apply_to (list[str]): The keys in the modality to load and transform. + """ + + def apply(self, data: dict[str, Any]) -> dict[str, Any]: + for key in self.apply_to: + state = data[key] + assert isinstance(state, torch.Tensor) + sin_state = torch.sin(state) + cos_state = torch.cos(state) + data[key] = torch.cat([sin_state, cos_state], dim=-1) + return data diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/video.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/video.py new file mode 100644 index 0000000000000000000000000000000000000000..80abc8d3e5e91fcc155215cfb1d865739087e61e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/data/transform/video.py @@ -0,0 +1,559 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any, Callable, ClassVar, Literal + +import albumentations as A +import cv2 +import numpy as np +import torch +import torchvision.transforms.v2 as T +from einops import rearrange +from pydantic import Field, PrivateAttr, field_validator + +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.schema import DatasetMetadata +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.transform.base import ModalityTransform + + +class VideoTransform(ModalityTransform): + # Configurable attributes + backend: str = Field(default="torchvision", description="The backend to use for the transformations") + + # Model variables + _train_transform: Callable | None = PrivateAttr(default=None) + _eval_transform: Callable | None = PrivateAttr(default=None) + _original_resolutions: dict[str, tuple[int, int]] = PrivateAttr(default_factory=dict) + + # Model constants + _INTERPOLATION_MAP: ClassVar[dict[str, dict[str, Any]]] = PrivateAttr( + { + "nearest": { + "albumentations": cv2.INTER_NEAREST, + "torchvision": T.InterpolationMode.NEAREST, + }, + "linear": { + "albumentations": cv2.INTER_LINEAR, + "torchvision": T.InterpolationMode.BILINEAR, + }, + "cubic": { + "albumentations": cv2.INTER_CUBIC, + "torchvision": T.InterpolationMode.BICUBIC, + }, + "area": { + "albumentations": cv2.INTER_AREA, + "torchvision": None, # Torchvision does not support this interpolation mode + }, + "lanczos4": { + "albumentations": cv2.INTER_LANCZOS4, # Lanczos with a 4x4 filter + "torchvision": T.InterpolationMode.LANCZOS, + # Torchvision does not specify filter size, might be different from 4x4 + }, + "linear_exact": { + "albumentations": cv2.INTER_LINEAR_EXACT, + "torchvision": None, # Torchvision does not support this interpolation mode + }, + "nearest_exact": { + "albumentations": cv2.INTER_NEAREST_EXACT, + "torchvision": T.InterpolationMode.NEAREST_EXACT, + }, + "max": { + "albumentations": cv2.INTER_MAX, + "torchvision": None, + }, + } + ) + + @property + def train_transform(self) -> Callable: + assert self._train_transform is not None, ( + "Transform is not set. Please call set_metadata() before calling apply()." + ) + return self._train_transform + + @train_transform.setter + def train_transform(self, value: Callable): + self._train_transform = value + + @property + def eval_transform(self) -> Callable | None: + return self._eval_transform + + @eval_transform.setter + def eval_transform(self, value: Callable | None): + self._eval_transform = value + + @property + def original_resolutions(self) -> dict[str, tuple[int, int]]: + assert self._original_resolutions is not None, ( + "Original resolutions are not set. Please call set_metadata() before calling apply()." + ) + return self._original_resolutions + + @original_resolutions.setter + def original_resolutions(self, value: dict[str, tuple[int, int]]): + self._original_resolutions = value + + def check_input(self, data: dict[str, Any]): + if self.backend == "torchvision": + for key in self.apply_to: + assert isinstance(data[key], torch.Tensor), f"Video {key} is not a torch tensor" + assert data[key].ndim in [ + 4, + 5, + ], f"Expected video {key} to have 4 or 5 dimensions (T, C, H, W or T, B, C, H, W), got {data[key].ndim}" + elif self.backend == "albumentations": + for key in self.apply_to: + assert isinstance(data[key], np.ndarray), f"Video {key} is not a numpy array" + assert data[key].ndim in [ + 4, + 5, + ], f"Expected video {key} to have 4 or 5 dimensions (T, C, H, W or T, B, C, H, W), got {data[key].ndim}" + else: + raise ValueError(f"Backend {self.backend} not supported") + + def set_metadata(self, dataset_metadata: DatasetMetadata): + super().set_metadata(dataset_metadata) + self.original_resolutions = {} + for key in self.apply_to: + split_keys = key.split(".") + assert len(split_keys) == 2, f"Invalid key: {key}. Expected format: modality.key" + sub_key = split_keys[1] + if sub_key in dataset_metadata.modalities.video: + self.original_resolutions[key] = dataset_metadata.modalities.video[sub_key].resolution + else: + raise ValueError( + f"Video key {sub_key} not found in dataset metadata. Available keys: {dataset_metadata.modalities.video.keys()}" + ) + train_transform = self.get_transform(mode="train") + eval_transform = self.get_transform(mode="eval") + if self.backend == "albumentations": + self.train_transform = A.ReplayCompose(transforms=[train_transform]) # type: ignore + if eval_transform is not None: + self.eval_transform = A.ReplayCompose(transforms=[eval_transform]) # type: ignore + else: + assert train_transform is not None, "Train transform must be set" + self.train_transform = train_transform + self.eval_transform = eval_transform + + def apply(self, data: dict[str, Any]) -> dict[str, Any]: + if self.training: + transform = self.train_transform + else: + transform = self.eval_transform + if transform is None: + return data + assert transform is not None, "Transform is not set. Please call set_metadata() before calling apply()." + try: + self.check_input(data) + except AssertionError as e: + raise ValueError(f"Input data does not match the expected format for {self.__class__.__name__}: {e}") from e + + # Concatenate views + views = [data[key] for key in self.apply_to] + num_views = len(views) + is_batched = views[0].ndim == 5 + bs = views[0].shape[0] if is_batched else 1 + if isinstance(views[0], torch.Tensor): + views = torch.cat(views, 0) + elif isinstance(views[0], np.ndarray): + views = np.concatenate(views, 0) + else: + raise ValueError(f"Unsupported view type: {type(views[0])}") + if is_batched: + views = rearrange(views, "(v b) t c h w -> (v b t) c h w", v=num_views, b=bs) + # Apply the transform + if self.backend == "torchvision": + views = transform(views) + elif self.backend == "albumentations": + assert isinstance(transform, A.ReplayCompose), "Transform must be a ReplayCompose" + first_frame = views[0] + transformed = transform(image=first_frame) + replay_data = transformed["replay"] + transformed_first_frame = transformed["image"] + + if len(views) > 1: + # Apply the same transformations to the rest of the frames + transformed_frames = [transform.replay(replay_data, image=frame)["image"] for frame in views[1:]] + # Add the first frame back + transformed_frames = [transformed_first_frame] + transformed_frames + else: + # If there is only one frame, just make a list with one frame + transformed_frames = [transformed_first_frame] + + # Delete the replay data to save memory + del replay_data + views = np.stack(transformed_frames, 0) + + else: + raise ValueError(f"Backend {self.backend} not supported") + # Split views + if is_batched: + views = rearrange(views, "(v b t) c h w -> v b t c h w", v=num_views, b=bs) + else: + views = rearrange(views, "(v t) c h w -> v t c h w", v=num_views) + for key, view in zip(self.apply_to, views): + data[key] = view + return data + + @classmethod + def _validate_interpolation(cls, interpolation: str): + if interpolation not in cls._INTERPOLATION_MAP: + raise ValueError(f"Interpolation mode {interpolation} not supported") + + def _get_interpolation(self, interpolation: str, backend: str = "torchvision"): + """ + Get the interpolation mode for the given backend. + + Args: + interpolation (str): The interpolation mode. + backend (str): The backend to use. + + Returns: + Any: The interpolation mode for the given backend. + """ + return self._INTERPOLATION_MAP[interpolation][backend] + + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable | None: + raise NotImplementedError( + "set_transform is not implemented for VideoTransform. Please implement this function to set the transforms." + ) + + +class VideoCrop(VideoTransform): + height: int | None = Field(default=None, description="The height of the input image") + width: int | None = Field(default=None, description="The width of the input image") + scale: float = Field( + ..., + description="The scale of the crop. The crop size is (width * scale, height * scale)", + ) + + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable: + """Get the transform for the given mode. + + Args: + mode (Literal["train", "eval"]): The mode to get the transform for. + + Returns: + Callable: If mode is "train", return a random crop transform. If mode is "eval", return a center crop transform. + """ + # 1. Check the input resolution + assert len(set(self.original_resolutions.values())) == 1, ( + f"All video keys must have the same resolution, got: {self.original_resolutions}" + ) + if self.height is None: + assert self.width is None, "Height and width must be either both provided or both None" + self.width, self.height = self.original_resolutions[self.apply_to[0]] + else: + assert self.width is not None, "Height and width must be either both provided or both None" + # 2. Create the transform + size = (int(self.height * self.scale), int(self.width * self.scale)) + if self.backend == "torchvision": + if mode == "train": + return T.RandomCrop(size) + elif mode == "eval": + return T.CenterCrop(size) + else: + raise ValueError(f"Crop mode {mode} not supported") + elif self.backend == "albumentations": + if mode == "train": + return A.RandomCrop(height=size[0], width=size[1], p=1) + elif mode == "eval": + return A.CenterCrop(height=size[0], width=size[1], p=1) + else: + raise ValueError(f"Crop mode {mode} not supported") + else: + raise ValueError(f"Backend {self.backend} not supported") + + def check_input(self, data: dict[str, Any]): + super().check_input(data) + # Check the input resolution + for key in self.apply_to: + if self.backend == "torchvision": + height, width = data[key].shape[-2:] + elif self.backend == "albumentations": + height, width = data[key].shape[-3:-1] + else: + raise ValueError(f"Backend {self.backend} not supported") + assert height == self.height and width == self.width, ( + f"Video {key} has invalid shape {height, width}, expected {self.height, self.width}" + ) + + +class VideoResize(VideoTransform): + height: int = Field(..., description="The height of the resize") + width: int = Field(..., description="The width of the resize") + interpolation: str = Field(default="linear", description="The interpolation mode") + antialias: bool = Field(default=True, description="Whether to apply antialiasing") + + @field_validator("interpolation") + def validate_interpolation(cls, v): + cls._validate_interpolation(v) + return v + + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable: + """Get the resize transform. Same transform for both train and eval. + + Args: + mode (Literal["train", "eval"]): The mode to get the transform for. + + Returns: + Callable: The resize transform. + """ + interpolation = self._get_interpolation(self.interpolation, self.backend) + if interpolation is None: + raise ValueError(f"Interpolation mode {self.interpolation} not supported for torchvision") + if self.backend == "torchvision": + size = (self.height, self.width) + return T.Resize(size, interpolation=interpolation, antialias=self.antialias) + elif self.backend == "albumentations": + return A.Resize( + height=self.height, + width=self.width, + interpolation=interpolation, + p=1, + ) + else: + raise ValueError(f"Backend {self.backend} not supported") + + +class VideoRandomRotation(VideoTransform): + degrees: float | tuple[float, float] = Field(..., description="The degrees of the random rotation") + interpolation: str = Field("linear", description="The interpolation mode") + + @field_validator("interpolation") + def validate_interpolation(cls, v): + cls._validate_interpolation(v) + return v + + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable | None: + """Get the random rotation transform, only used in train mode. + + Args: + mode (Literal["train", "eval"]): The mode to get the transform for. + + Returns: + Callable | None: The random rotation transform. None for eval mode. + """ + if mode == "eval": + return None + interpolation = self._get_interpolation(self.interpolation, self.backend) + if interpolation is None: + raise ValueError(f"Interpolation mode {self.interpolation} not supported for torchvision") + if self.backend == "torchvision": + return T.RandomRotation(self.degrees, interpolation=interpolation) # type: ignore + elif self.backend == "albumentations": + return A.Rotate(limit=self.degrees, interpolation=interpolation, p=1) + else: + raise ValueError(f"Backend {self.backend} not supported") + + +class VideoHorizontalFlip(VideoTransform): + p: float = Field(..., description="The probability of the horizontal flip") + + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable | None: + """Get the horizontal flip transform, only used in train mode. + + Args: + mode (Literal["train", "eval"]): The mode to get the transform for. + + Returns: + Callable | None: If mode is "train", return a horizontal flip transform. If mode is "eval", return None. + """ + if mode == "eval": + return None + if self.backend == "torchvision": + return T.RandomHorizontalFlip(self.p) + elif self.backend == "albumentations": + return A.HorizontalFlip(p=self.p) + else: + raise ValueError(f"Backend {self.backend} not supported") + + +class VideoGrayscale(VideoTransform): + p: float = Field(..., description="The probability of the grayscale transformation") + + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable | None: + """Get the grayscale transform, only used in train mode. + + Args: + mode (Literal["train", "eval"]): The mode to get the transform for. + + Returns: + Callable | None: If mode is "train", return a grayscale transform. If mode is "eval", return None. + """ + if mode == "eval": + return None + if self.backend == "torchvision": + return T.RandomGrayscale(self.p) + elif self.backend == "albumentations": + return A.ToGray(p=self.p) + else: + raise ValueError(f"Backend {self.backend} not supported") + + +class VideoColorJitter(VideoTransform): + brightness: float | tuple[float, float] = Field(..., description="The brightness of the color jitter") + contrast: float | tuple[float, float] = Field(..., description="The contrast of the color jitter") + saturation: float | tuple[float, float] = Field(..., description="The saturation of the color jitter") + hue: float | tuple[float, float] = Field(..., description="The hue of the color jitter") + + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable | None: + """Get the color jitter transform, only used in train mode. + + Args: + mode (Literal["train", "eval"]): The mode to get the transform for. + + Returns: + Callable | None: If mode is "train", return a color jitter transform. If mode is "eval", return None. + """ + if mode == "eval": + return None + if self.backend == "torchvision": + return T.ColorJitter( + brightness=self.brightness, + contrast=self.contrast, + saturation=self.saturation, + hue=self.hue, + ) + elif self.backend == "albumentations": + return A.ColorJitter( + brightness=self.brightness, + contrast=self.contrast, + saturation=self.saturation, + hue=self.hue, + p=1, + ) + else: + raise ValueError(f"Backend {self.backend} not supported") + + +class VideoRandomGrayscale(VideoTransform): + p: float = Field(..., description="The probability of the grayscale transformation") + + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable | None: + """Get the grayscale transform, only used in train mode. + + Args: + mode (Literal["train", "eval"]): The mode to get the transform for. + + Returns: + Callable | None: If mode is "train", return a grayscale transform. If mode is "eval", return None. + """ + if mode == "eval": + return None + if self.backend == "torchvision": + return T.RandomGrayscale(self.p) + elif self.backend == "albumentations": + return A.ToGray(p=self.p) + else: + raise ValueError(f"Backend {self.backend} not supported") + + +class VideoRandomPosterize(VideoTransform): + bits: int = Field(..., description="The number of bits to posterize the image") + p: float = Field(..., description="The probability of the posterize transformation") + + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable | None: + """Get the posterize transform, only used in train mode. + + Args: + mode (Literal["train", "eval"]): The mode to get the transform for. + + Returns: + Callable | None: If mode is "train", return a posterize transform. If mode is "eval", return None. + """ + if mode == "eval": + return None + if self.backend == "torchvision": + return T.RandomPosterize(bits=self.bits, p=self.p) + elif self.backend == "albumentations": + return A.Posterize(num_bits=self.bits, p=self.p) + else: + raise ValueError(f"Backend {self.backend} not supported") + + +class VideoToTensor(VideoTransform): + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable: + """Get the to tensor transform. Same transform for both train and eval. + + Args: + mode (Literal["train", "eval"]): The mode to get the transform for. + + Returns: + Callable: The to tensor transform. + """ + if self.backend == "torchvision": + return self.__class__.to_tensor + else: + raise ValueError(f"Backend {self.backend} not supported") + + def check_input(self, data: dict): + """Check if the input data has the correct shape. + Expected video shape: [T, H, W, C], dtype np.uint8 + """ + for key in self.apply_to: + assert key in data, f"Key {key} not found in data. Available keys: {data.keys()}" + assert data[key].ndim in [ + 4, + 5, + ], f"Video {key} must have 4 or 5 dimensions, got {data[key].ndim}" + assert data[key].dtype == np.uint8, f"Video {key} must have dtype uint8, got {data[key].dtype}" + input_resolution = data[key].shape[-3:-1][::-1] + if key in self.original_resolutions: + expected_resolution = self.original_resolutions[key] + else: + expected_resolution = input_resolution + # assert ( + # input_resolution == expected_resolution + # ), f"Video {key} has invalid resolution {input_resolution}, expected {expected_resolution}. Full shape: {data[key].shape}" + + @staticmethod + def to_tensor(frames: np.ndarray) -> torch.Tensor: + """Convert numpy array to tensor efficiently. + + Args: + frames: numpy array of shape [T, H, W, C] in uint8 format + Returns: + tensor of shape [T, C, H, W] in range [0, 1] + """ + frames_tensor = torch.from_numpy(frames).to(torch.float32) / 255.0 + return frames_tensor.permute(0, 3, 1, 2) # [T, C, H, W] + + +class VideoToNumpy(VideoTransform): + def get_transform(self, mode: Literal["train", "eval"] = "train") -> Callable: + """Get the to numpy transform. Same transform for both train and eval. + + Args: + mode (Literal["train", "eval"]): The mode to get the transform for. + + Returns: + Callable: The to numpy transform. + """ + if self.backend == "torchvision": + return self.__class__.to_numpy + else: + raise ValueError(f"Backend {self.backend} not supported") + + @staticmethod + def to_numpy(frames: torch.Tensor) -> np.ndarray: + """Convert tensor back to numpy array efficiently. + + Args: + frames: tensor of shape [T, C, H, W] in range [0, 1] + Returns: + numpy array of shape [T, H, W, C] in uint8 format + """ + return (frames.permute(0, 2, 3, 1) * 255).to(torch.uint8).cpu().numpy() diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/groot_configs.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/groot_configs.py new file mode 100644 index 0000000000000000000000000000000000000000..2b80fdb493c02be13e10a1b87a78fb94521ee5dd --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/groot_configs.py @@ -0,0 +1,184 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.dataset import ModalityConfig + +# from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.transform import ( +# VideoCrop, +# VideoResize, +# VideoToTensor, +# ) +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.transform.base import ComposedModalityTransform +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.transform.concat import ConcatTransform +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.transform.state_action import ( + StateActionToTensor, + StateActionTransform, +) +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.transform.video import ( + VideoCrop, + VideoResize, + VideoToTensor, +) + + +def construct_modality_config_and_transforms(num_frames, embodiment, downscaled_res=False): + if embodiment == "gr1": + timestep_interval = 2 + delta_indices = list(range(0, num_frames * timestep_interval, timestep_interval)) + video_key = "video.ego_view_freq20" if not downscaled_res else "video.ego_view_bg_crop_pad_res256_freq20" + config = { + "video": ModalityConfig( + delta_indices=delta_indices, + modality_keys=[video_key], + ), + "state": ModalityConfig( + delta_indices=[0], + modality_keys=[ + "state.left_arm", + "state.right_arm", + "state.left_hand", + "state.right_hand", + "state.waist", + ], + ), + "action": ModalityConfig( + delta_indices=delta_indices, + modality_keys=[ + "action.left_arm", + "action.right_arm", + "action.left_hand", + "action.right_hand", + "action.waist", + ], + ), + } + elif embodiment == "gr1_video_only": + timestep_interval = 1 + delta_indices = list(range(0, num_frames * timestep_interval, timestep_interval)) + config = { + "video": ModalityConfig( + delta_indices=delta_indices, + modality_keys=["video.ego_view_bg_crop_pad_res256_freq20"], + ), + "state": ModalityConfig( + delta_indices=[0], + modality_keys=[ + "state.left_arm", + "state.right_arm", + "state.left_hand", + "state.right_hand", + "state.waist", + ], + ), + "action": ModalityConfig( + delta_indices=delta_indices, + modality_keys=[ + "action.left_arm", + "action.right_arm", + "action.left_hand", + "action.right_hand", + "action.waist", + ], + ), + "language": ModalityConfig(delta_indices=[0], modality_keys=["annotation.human.coarse_action"]), + } + elif embodiment == "agibot": + timestep_interval = 4 + delta_indices = list(range(0, num_frames * timestep_interval, timestep_interval)) + video_key = "video.top_head" if not downscaled_res else "video.top_head_pad_res256_freq10" + config = { + "video": ModalityConfig( + delta_indices=delta_indices, + modality_keys=[video_key], + ), + "state": ModalityConfig( + delta_indices=[0], + modality_keys=[ + "state.left_arm_joint_position", + "state.right_arm_joint_position", + "state.left_effector_position", + "state.right_effector_position", + "state.head_position", + "state.waist_position", + ], + ), + "action": ModalityConfig( + delta_indices=delta_indices, + modality_keys=[ + "action.left_arm_joint_position", + "action.right_arm_joint_position", + "action.left_effector_position", + "action.right_effector_position", + "action.head_position", + "action.waist_position", + "action.robot_velocity", + ], + ), + } + + video_modality, state_modality, action_modality = config["video"], config["state"], config["action"] + if embodiment == "gr1" or embodiment == "gr1_video_only": + width = 832 if not downscaled_res else 256 + height = 480 if not downscaled_res else 256 + elif embodiment == "agibot": + width = 640 if not downscaled_res else 256 + height = 480 if not downscaled_res else 256 + + train_transform = ComposedModalityTransform( + transforms=[ + VideoToTensor(apply_to=video_modality.modality_keys), + VideoCrop(apply_to=video_modality.modality_keys, scale=0.95), + VideoResize(apply_to=video_modality.modality_keys, height=height, width=width, interpolation="linear"), + # VideoColorJitter(apply_to=video_modality.modality_keys, brightness=0.3, contrast=0.4, saturation=0.5, hue=0.08), + StateActionToTensor(apply_to=state_modality.modality_keys), + StateActionTransform( + apply_to=state_modality.modality_keys, + normalization_modes={key: "min_max" for key in state_modality.modality_keys}, + ), + StateActionToTensor(apply_to=action_modality.modality_keys), + StateActionTransform( + apply_to=action_modality.modality_keys, + normalization_modes={key: "min_max" for key in action_modality.modality_keys}, + ), + ConcatTransform( + video_concat_order=video_modality.modality_keys, + state_concat_order=state_modality.modality_keys, + action_concat_order=action_modality.modality_keys, + ), + ] + ) + test_transform = ComposedModalityTransform( + transforms=[ + VideoToTensor(apply_to=video_modality.modality_keys), + VideoResize(apply_to=video_modality.modality_keys, height=height, width=width, interpolation="linear"), + StateActionToTensor(apply_to=state_modality.modality_keys), + StateActionTransform( + apply_to=state_modality.modality_keys, + normalization_modes={key: "min_max" for key in state_modality.modality_keys}, + ), + StateActionToTensor(apply_to=action_modality.modality_keys), + StateActionTransform( + apply_to=action_modality.modality_keys, + normalization_modes={key: "min_max" for key in action_modality.modality_keys}, + ), + ConcatTransform( + video_concat_order=video_modality.modality_keys, + state_concat_order=state_modality.modality_keys, + action_concat_order=action_modality.modality_keys, + ), + ] + ) + + return config, train_transform, test_transform diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/py.typed b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/py.typed new file mode 100644 index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/eval.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/eval.py new file mode 100644 index 0000000000000000000000000000000000000000..3a2b7d709bae2cd7f6207849e8ab3265a46392f6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/eval.py @@ -0,0 +1,113 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import matplotlib.pyplot as plt +import numpy as np +from gr00t_dreams.data.dataset import LeRobotSingleDataset +from gr00t_dreams.model.policy import BasePolicy + +# numpy print precision settings 3, dont use exponential notation +np.set_printoptions(precision=3, suppress=True) + + +def download_from_hg(repo_id: str, repo_type: str) -> str: + """ + Download the model/dataset from the hugging face hub. + return the path to the downloaded + """ + from huggingface_hub import snapshot_download + + repo_path = snapshot_download(repo_id, repo_type=repo_type) + return repo_path + + +def calc_mse_for_single_trajectory( + policy: BasePolicy, + dataset: LeRobotSingleDataset, + traj_id: int, + modality_keys: list, + steps=300, + action_horizon=16, + plot=False, +): + state_joints_across_time = [] + gt_action_joints_across_time = [] + pred_action_joints_across_time = [] + + for step_count in range(steps): + data_point = dataset.get_step_data(traj_id, step_count) + + # NOTE this is to get all modality keys concatenated + # concat_state = data_point[f"state.{modality_keys[0]}"][0] + # concat_gt_action = data_point[f"action.{modality_keys[0]}"][0] + concat_state = np.concatenate([data_point[f"state.{key}"][0] for key in modality_keys], axis=0) + concat_gt_action = np.concatenate([data_point[f"action.{key}"][0] for key in modality_keys], axis=0) + + state_joints_across_time.append(concat_state) + gt_action_joints_across_time.append(concat_gt_action) + + if step_count % action_horizon == 0: + print("inferencing at step: ", step_count) + action_chunk = policy.get_action(data_point) + for j in range(action_horizon): + # NOTE: concat_pred_action = action[f"action.{modality_keys[0]}"][j] + # the np.atleast_1d is to ensure the action is a 1D array, handle where single value is returned + concat_pred_action = np.concatenate( + [np.atleast_1d(action_chunk[f"action.{key}"][j]) for key in modality_keys], + axis=0, + ) + pred_action_joints_across_time.append(concat_pred_action) + + # plot the joints + state_joints_across_time = np.array(state_joints_across_time) + gt_action_joints_across_time = np.array(gt_action_joints_across_time) + pred_action_joints_across_time = np.array(pred_action_joints_across_time)[:steps] + assert state_joints_across_time.shape == gt_action_joints_across_time.shape == pred_action_joints_across_time.shape + + # calc MSE across time + mse = np.mean((gt_action_joints_across_time - pred_action_joints_across_time) ** 2) + print("Unnormalized Action MSE across single traj:", mse) + + num_of_joints = state_joints_across_time.shape[1] + + if plot: + fig, axes = plt.subplots(nrows=num_of_joints, ncols=1, figsize=(8, 4 * num_of_joints)) + + # Add a global title showing the modality keys + fig.suptitle( + f"Trajectory {traj_id} - Modalities: {', '.join(modality_keys)}", + fontsize=16, + color="blue", + ) + + for i, ax in enumerate(axes): + ax.plot(state_joints_across_time[:, i], label="state joints") + ax.plot(gt_action_joints_across_time[:, i], label="gt action joints") + ax.plot(pred_action_joints_across_time[:, i], label="pred action joints") + + # put a dot every ACTION_HORIZON + for j in range(0, steps, action_horizon): + if j == 0: + ax.plot(j, gt_action_joints_across_time[j, i], "ro", label="inference point") + else: + ax.plot(j, gt_action_joints_across_time[j, i], "ro") + + ax.set_title(f"Joint {i}") + ax.legend() + + plt.tight_layout() + plt.show() + + return mse diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/experiment.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/experiment.py new file mode 100644 index 0000000000000000000000000000000000000000..191cc0fc08412b8c7144859bbe187b131ab8f5a9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/experiment.py @@ -0,0 +1,62 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import shutil +from pathlib import Path + +import torch +from transformers import Trainer, TrainerCallback + + +def safe_save_model_for_hf_trainer(trainer: Trainer, output_dir: str): + """Collects the state dict and dump to disk.""" + if trainer.deepspeed: + torch.cuda.synchronize() + trainer.save_model(output_dir, _internal_call=True) + return + + state_dict = trainer.model.state_dict() + if trainer.args.should_save: + cpu_state_dict = {key: value.cpu() for key, value in state_dict.items()} + del state_dict + trainer._save(output_dir, state_dict=cpu_state_dict) # noqa + + +class CheckpointFormatCallback(TrainerCallback): + """This callback format checkpoint to make them standalone. For now, it copies all config + files to /checkpoint-{step}/experiment_cfg/: + - conf.yaml + - initial_actions.npz + - metadata.json + """ + + def __init__(self, run_name: str, exp_cfg_dir: Path | None = None): + """ + Args: + run_name: Name of the experiment run + exp_cfg_dir: Path to the directory containing all experiment metadata + """ + self.exp_cfg_dir = exp_cfg_dir + + def on_save(self, args, state, control, **kwargs): + """Called after the trainer saves a checkpoint.""" + if state.is_world_process_zero: + checkpoint_dir = Path(args.output_dir) / f"checkpoint-{state.global_step}" + + # Copy experiment config directory if provided + if self.exp_cfg_dir is not None: + exp_cfg_dst = checkpoint_dir / self.exp_cfg_dir.name + if self.exp_cfg_dir.exists(): + shutil.copytree(self.exp_cfg_dir, exp_cfg_dst, dirs_exist_ok=True) diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/misc.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/misc.py new file mode 100644 index 0000000000000000000000000000000000000000..0e246bb987095d3051140aba4919d8da934bc5ea --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/misc.py @@ -0,0 +1,80 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Functions that work on nested structures of torch.Tensor or numpy array +""" + +from collections.abc import Sequence + +import numpy as np +import torch +import tree + + +def any_describe_str(x, shape_only=False): + """ + Describe type, shape, device, data type (of np array/tensor) + Very useful for debugging + """ + t = type(x) + tname = type(x).__name__ + if isinstance(x, np.ndarray): + shape = list(x.shape) + if x.size == 1: + if shape_only: + return f"np scalar: {x.item()} {shape}" + else: + return f"np scalar: {x.item()} {shape} {x.dtype}" + else: + if shape_only: + return f"np: {shape}" + else: + return f"np: {shape} {x.dtype}" + elif torch.is_tensor(x): + shape = list(x.size()) + if x.numel() == 1: + if shape_only: + return f"torch scalar: {x.item()} {shape}" + else: + return f"torch scalar: {x.item()} {shape} {x.dtype} {x.device}" + else: + if shape_only: + return f"torch: {shape}" + else: + return f"torch: {shape} {x.dtype} {x.device}" + elif isinstance(x, str): + return x + elif isinstance(x, Sequence): + return f"{tname}[{len(x)}]" + elif x is None: + return "None" + elif np.issubdtype(t, np.number) or np.issubdtype(t, np.bool_): + return f"{tname}: {x}" + else: + return f"{tname}" + + +def any_describe(x, msg="", *, shape_only=False): + # from omlet.utils import yaml_dumps + from pprint import pprint + + if isinstance(x, str) and msg != "": + x, msg = msg, x + + if msg: + msg += ": " + print(msg, end="") + pprint(tree.map_structure(lambda i: any_describe_str(i, shape_only=shape_only), x)) diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/peft.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/peft.py new file mode 100644 index 0000000000000000000000000000000000000000..7d851e52fcea2b9edf44b55aa7914250c500e90e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/peft.py @@ -0,0 +1,56 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import torch +from peft import LoraConfig, get_peft_model + + +def _wrap_forward(model): + def _forward(inputs): + backbone_inputs, action_inputs = model.prepare_input(inputs) + backbone_outputs = model.backbone(backbone_inputs) + action_head_outputs = model.action_head(backbone_outputs, action_inputs) + model.validate_data(action_head_outputs, backbone_outputs, is_training=True) + return action_head_outputs + + model.forward = _forward + return model + + +def get_lora_model(model, rank=32, lora_alpha=16, lora_dropout=0.1): + target_modules = [] + + # Inspect model structure to find the correct paths + for name, module in model.named_modules(): + # Look for linear layers in attention mechanisms + if isinstance(module, torch.nn.Linear): + if any(x in name for x in ["q_proj", "v_proj", "to_q", "to_v", "k_proj", "to_k"]): + target_modules.append(name) + + lora_config = LoraConfig( + r=rank, + lora_alpha=lora_alpha, + target_modules=target_modules, + lora_dropout=lora_dropout, + bias="none", + task_type="CAUSAL_LM", + ) + + model = get_peft_model(model, lora_config) + model.print_trainable_parameters() + + model = _wrap_forward(model) + + return model diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/video.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/video.py new file mode 100644 index 0000000000000000000000000000000000000000..58a777f34ea837e55eae54b794fff077f99d4019 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/gr00t_dreams/utils/video.py @@ -0,0 +1,198 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import av +import cv2 +import decord # noqa: F401 +import numpy as np + +import torch # noqa: F401 # isort: skip +import torchvision # noqa: F401 # isort: skip + + +def get_frames_by_indices( + video_path: str, + indices: list[int] | np.ndarray, + video_backend: str = "decord", + video_backend_kwargs: dict = {}, +) -> np.ndarray: + if video_backend == "decord": + vr = decord.VideoReader(video_path, **video_backend_kwargs) + frames = vr.get_batch(indices) + return frames.asnumpy() + elif video_backend == "opencv": + frames = [] + cap = cv2.VideoCapture(video_path, **video_backend_kwargs) + for idx in indices: + cap.set(cv2.CAP_PROP_POS_FRAMES, idx) + ret, frame = cap.read() + if not ret: + raise ValueError(f"Unable to read frame at index {idx}") + frames.append(frame) + cap.release() + frames = np.array(frames) + return frames + else: + raise NotImplementedError + + +def get_frames_by_timestamps( + video_path: str, + timestamps: list[float] | np.ndarray, + video_backend: str = "decord", + video_backend_kwargs: dict = {}, +) -> np.ndarray: + """Get frames from a video at specified timestamps. + Args: + video_path (str): Path to the video file. + timestamps (list[int] | np.ndarray): Timestamps to retrieve frames for, in seconds. + video_backend (str, optional): Video backend to use. Defaults to "decord". + Returns: + np.ndarray: Frames at the specified timestamps. + """ + if video_backend == "decord": + vr = decord.VideoReader(video_path, **video_backend_kwargs) + num_frames = len(vr) + # Retrieve the timestamps for each frame in the video + frame_ts: np.ndarray = vr.get_frame_timestamp(range(num_frames)) + # Map each requested timestamp to the closest frame index + # Only take the first element of the frame_ts array which corresponds to start_seconds + indices = np.abs(frame_ts[:, :1] - timestamps).argmin(axis=0) + frames = vr.get_batch(indices) + return frames.asnumpy() + elif video_backend == "opencv": + # Open the video file + cap = cv2.VideoCapture(video_path, **video_backend_kwargs) + if not cap.isOpened(): + raise ValueError(f"Unable to open video file: {video_path}") + # Retrieve the total number of frames + num_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT)) + # Calculate timestamps for each frame + fps = cap.get(cv2.CAP_PROP_FPS) + frame_ts = np.arange(num_frames) / fps + frame_ts = frame_ts[:, np.newaxis] # Reshape to (num_frames, 1) for broadcasting + # Map each requested timestamp to the closest frame index + indices = np.abs(frame_ts - timestamps).argmin(axis=0) + frames = [] + for idx in indices: + cap.set(cv2.CAP_PROP_POS_FRAMES, idx) + ret, frame = cap.read() + if not ret: + raise ValueError(f"Unable to read frame at index {idx}") + frames.append(frame) + cap.release() + frames = np.array(frames) + return frames + elif video_backend == "torchvision_av": + # set backend + torchvision.set_video_backend("pyav") + # set a video stream reader + reader = torchvision.io.VideoReader(video_path, "video") + # set the first and last requested timestamps + # Note: previous timestamps are usually loaded, since we need to access the previous key frame + first_ts = timestamps[0] + last_ts = timestamps[-1] + # access closest key frame of the first requested frame + # Note: closest key frame timestamp is usally smaller than `first_ts` (e.g. key frame can be the first frame of the video) + # for details on what `seek` is doing see: https://pyav.basswood-io.com/docs/stable/api/container.html?highlight=inputcontainer#av.container.InputContainer.seek + reader.seek(first_ts, keyframes_only=True) + # load all frames until last requested frame + loaded_frames = [] + loaded_ts = [] + for frame in reader: + current_ts = frame["pts"] + loaded_frames.append(frame["data"]) + loaded_ts.append(current_ts) + if current_ts >= last_ts: + break + if len(loaded_frames) >= len(timestamps): + break + reader.container.close() + reader = None + frames = np.array(loaded_frames) + return frames.transpose(0, 2, 3, 1) + else: + raise NotImplementedError + + +def get_all_frames( + video_path: str, + video_backend: str = "decord", + video_backend_kwargs: dict = {}, + resize_size: tuple[int, int] | None = None, +) -> np.ndarray: + """Get all frames from a video. + Args: + video_path (str): Path to the video file. + video_backend (str, optional): Video backend to use. Defaults to "decord". + video_backend_kwargs (dict, optional): Keyword arguments for the video backend. + resize_size (tuple[int, int], optional): Resize size for the frames. Defaults to None. + """ + if video_backend == "decord": + vr = decord.VideoReader(video_path, **video_backend_kwargs) + frames = vr.get_batch(range(len(vr))).asnumpy() + elif video_backend == "pyav": + container = av.open(video_path) + frames = [] + for frame in container.decode(video=0): + frame = frame.to_ndarray(format="rgb24") + frames.append(frame) + frames = np.array(frames) + elif video_backend == "torchvision_av": + # set backend and reader + torchvision.set_video_backend("pyav") + reader = torchvision.io.VideoReader(video_path, "video") + frames = [] + for frame in reader: + frames.append(frame["data"]) + frames = np.array(frames) + frames = frames.transpose(0, 2, 3, 1) + else: + raise NotImplementedError(f"Video backend {video_backend} not implemented") + # resize frames if specified + if resize_size is not None: + frames = [cv2.resize(frame, resize_size) for frame in frames] + frames = np.array(frames) + return frames + + +def get_all_frames_and_timestamps( + video_path: str, + video_backend: str = "decord", + video_backend_kwargs: dict = {}, +) -> tuple[np.ndarray, np.ndarray]: + """Get all frames from a video. + + Returns: + tuple[np.ndarray, np.ndarray]: Frames and timestamps. + """ + if video_backend == "decord": + vr = decord.VideoReader(video_path, **video_backend_kwargs) + frames = vr.get_batch(range(len(vr))).asnumpy() + return frames, vr.get_frame_timestamp(range(len(vr)))[:, 0] + + elif video_backend == "pyav": + container = av.open(video_path) + stream = container.streams.video[0] + assert stream.time_base is not None + frames = [] + timestamps = [] + for frame in container.decode(video=0): + frames.append(frame.to_ndarray(format="rgb24")) + timestamps.append(frame.pts * stream.time_base) + container.close() + return np.stack(frames), np.array(timestamps) + + else: + raise NotImplementedError diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/webdataset_mv_s3.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/webdataset_mv_s3.py new file mode 100644 index 0000000000000000000000000000000000000000..f2c8f191c184b48862ce48844749d40e0f674d0e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/webdataset_mv_s3.py @@ -0,0 +1,370 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""WebDataset loader for multi-view action-conditional robot data from S3. + +This module extends the base WebDataset loader to support multi-view camera setups. +It builds on webdataset_s3.py the same way dataset_mv_local.py builds on dataset_local.py. + +Supports multi-view wdinfo format where videos are organized by camera subdirectories: + videos/base_0/00000000.tar + videos/base_1/00000000.tar + videos/wrist/00000000.tar + annotations/00000000.tar + +The wdinfo.json should include: +{ + "multi_view": true, + "camera_ids": ["base_0", "base_1", "wrist"], + "data_keys": ["videos", "annotations"], + ... +} + +Run this command to interactively debug: +PYTHONPATH=. python cosmos_policy/_src/predict2/action/datasets/webdataset_mv_s3.py +""" + +import json +import random +import time +from typing import Callable + +import torch + +from cosmos_policy._src.imaginaire.datasets.webdataset.config.schema import DatasetInfo, TarSample +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.object_store import ObjectStore +from cosmos_policy._src.predict2.action.datasets.webdataset_s3 import ( + ActionConditionedWebDatasetS3, + ActionDataAugmentor, +) + + +class ActionDataAugmentorMultiView(ActionDataAugmentor): + """Multi-view augmentor that concatenates frames from multiple cameras.""" + + def _get_obs( + self, sample: dict, frame_ids: list[int], cam_id: list | None = None, pre_encode: bool = False + ) -> tuple[torch.Tensor, list[str]]: + """Get observation frames from multiple camera views. + + Args: + sample: WebDataset sample containing video data + frame_ids: List of frame indices to extract + pre_encode: Whether to use pre-encoded videos + + Returns: + Tuple of (concatenated video tensor [T, C, H, W*num_views], camera IDs used) + """ + del cam_id # Unused - multi-view always uses self.cam_ids configuration + # cam_ids format: [["base_0", "base_1"], "wrist_0"] + # First element: list to randomly sample from + # Second element: fixed camera + # Note: Check for list-like objects (including OmegaConf ListConfig), not just Python list + first_cam = self.cam_ids[0] + is_list_like = not isinstance(first_cam, str) and hasattr(first_cam, "__iter__") + temp_cam_id_0 = random.choice(list(first_cam)) if is_list_like else first_cam + temp_cam_id_1 = self.cam_ids[1] + + frames_0 = self._get_frames(sample, frame_ids, cam_id=temp_cam_id_0, pre_encode=pre_encode) + frames_1 = self._get_frames(sample, frame_ids, cam_id=temp_cam_id_1, pre_encode=pre_encode) + # Concatenate along width dimension (dim=3 for [T, C, H, W]) + frames = torch.cat([frames_0, frames_1], dim=3) + return frames, [temp_cam_id_0, temp_cam_id_1] + + +class MultiViewVideoOrganizer: + """Organizes multi-view video data from WebDataset samples. + + When loading multi-view data, the base WebDataset creates keys like + "videos/base_0", "videos/base_1", etc. This organizer restructures + the sample to have: + sample["videos"] = { + "base_0": , + "base_1": , + ... + } + """ + + def __init__(self, camera_ids: list[str]): + """Initialize the multi-view video organizer. + + Args: + camera_ids: List of camera IDs to organize + """ + self.camera_ids = self._flatten_camera_ids(camera_ids) + self.is_generator = True + + def _flatten_camera_ids(self, cam_ids: list) -> list[str]: + """Flatten nested camera ID lists (e.g., [["base_0", "base_1"], "wrist"]).""" + flat_ids = [] + for cam_id in cam_ids: + # Check for list-like types (including OmegaConf ListConfig), but not strings + if not isinstance(cam_id, str) and hasattr(cam_id, "__iter__"): + flat_ids.extend(cam_id) + else: + flat_ids.append(cam_id) + return flat_ids + + def __call__(self, data_stream): + """Reorganize video data in samples.""" + for sample in data_stream: + try: + # Look for video data with camera-specific keys + videos_dict = {} + missing_cams = [] + + for cam_id in self.camera_ids: + video_key = f"videos_{cam_id}" + video_data = sample.get(video_key) + if video_data is not None: + videos_dict[cam_id] = video_data + else: + missing_cams.append(cam_id) + + # If we found any camera videos, organize them + if videos_dict: + sample["videos"] = videos_dict + + yield sample + + except Exception as e: + log.warning(f"Error organizing multi-view sample: {e}") + yield sample + + +class ActionConditionedMultiViewWebDatasetS3(ActionConditionedWebDatasetS3): + """Multi-view WebDataset loader extending the base single-view loader. + + Supports multi-view wdinfo format where videos are organized by camera subdirectories. + """ + + def __init__(self, *args, **kwargs): + """Initialize the multi-view WebDataset S3 loader.""" + # Multi-view specific attributes (set during wdinfo parsing) + self.multi_view = False + self.camera_ids_from_wdinfo: list[str] = [] + + super().__init__(*args, **kwargs) + + def parse_dataset_info(self, dataset_info: list[DatasetInfo], use_multithread: bool = True) -> None: + """Parse metadata about the list of tar files with multi-view support. + + This overrides the base method to handle multi-view wdinfo format where + videos are stored in camera-specific subdirectories. + """ + log.info(f"[MultiView] Start parsing dataset info with {len(dataset_info)} entries") + tic = time.time() + + for dset_num, dset_info in enumerate(dataset_info): + if len(dset_info.wdinfo) == 0: + log.warning(f"No wdinfo found for dataset {dset_num}, skipping...") + continue + + use_object_store = dset_info.object_store_config.enabled + self.use_object_store = use_object_store + dset_id = f"dset: {dset_num}" + + if use_object_store: + object_store_reader = ObjectStore(config_object_storage=dset_info.object_store_config) + easy_io_backend_dset = object_store_reader.easy_io_backend + bucket_dset = dset_info.object_store_config.bucket + else: + object_store_reader = None + easy_io_backend_dset = None + bucket_dset = None + + tar_samples = [] + total_key_count = 0 + chunk_sizes = [] + + for wdinfo_path in dset_info.wdinfo: + log.info(f"[MultiView] Processing wdinfo: {wdinfo_path}") + + if use_object_store: + if not object_store_reader.object_exists(wdinfo_path): + raise FileNotFoundError(f"{wdinfo_path} not found") + cur_dset_info = object_store_reader.load_object(key=wdinfo_path, type="json") + else: + with open(wdinfo_path) as fp: + cur_dset_info = json.load(fp) + + data_root = cur_dset_info["root"] + # Strip s3://bucket/ prefix from root if present + if data_root.startswith("s3://"): + parts = data_root[5:].split("/", 1) + data_root = parts[1] if len(parts) > 1 else "" + + tar_files_list = cur_dset_info["data_list"] + is_multi_view = cur_dset_info.get("multi_view", False) + camera_ids = cur_dset_info.get("camera_ids", []) + data_keys = cur_dset_info.get("data_keys", self.data_keys) + + if is_multi_view: + self.multi_view = True + self.camera_ids_from_wdinfo = camera_ids + log.info(f"[MultiView] Detected multi-view dataset with cameras: {camera_ids}") + log.info(f"[MultiView] Data keys: {data_keys}") + + # For multi-view, we need to create separate "virtual" keys for each camera + # The base WebDataset will construct paths as: root/key/tar_file + # We need: root/videos/camera_id/tar_file for videos + # root/annotations/tar_file for other data + + # Create modified keys list for multi-view + multi_view_keys = [] + for key in data_keys: + if key == "videos": + # Add a key for each camera: "videos/base_0", "videos/base_1", etc. + for cam_id in camera_ids: + multi_view_keys.append(f"videos/{cam_id}") + else: + multi_view_keys.append(key) + + log.info(f"[MultiView] Expanded keys for loading: {multi_view_keys}") + + local_tar_samples = [ + TarSample( + path=tar_file, + root=data_root, + keys=(dset_info.per_dataset_keys if dset_info.per_dataset_keys else multi_view_keys), + meta=dset_info, + dset_id=dset_id, + sample_keys_full_list=None, + ) + for tar_file in tar_files_list + ] + else: + # Standard single-view handling - delegate to parent + local_tar_samples = [ + TarSample( + path=tar_file, + root=data_root, + keys=(dset_info.per_dataset_keys if dset_info.per_dataset_keys else self.data_keys), + meta=dset_info, + dset_id=dset_id, + sample_keys_full_list=None, + ) + for tar_file in tar_files_list + ] + + tar_samples.extend(local_tar_samples) + total_key_count += cur_dset_info["total_key_count"] + chunk_sizes.append(cur_dset_info["chunk_size"]) + + # Store results + self.wdinfo.tar_files.extend(tar_samples) + self.wdinfo.total_key_count += total_key_count + if chunk_sizes: + self.wdinfo.chunk_size = chunk_sizes[0] + if easy_io_backend_dset: + self.easy_io_backend[dset_id] = easy_io_backend_dset + if bucket_dset: + self.bucket[dset_id] = bucket_dset + + toc = time.time() + log.info( + f"[MultiView] Parsed {len(dataset_info)} wdinfos " + f"(num_keys={self.wdinfo.total_key_count}, num_tars={len(self.wdinfo.tar_files)}) " + f"in {(toc - tic):.2f}s" + ) + + def build_data_augmentor(self, augmentor_cfg: dict) -> Callable: + """Build multi-view data augmentor with video organizer.""" + from functools import partial + + from cosmos_policy._src.imaginaire.datasets.webdataset.webdataset import Dataset as WebDatasetBase + from cosmos_policy._src.imaginaire.lazy_config import instantiate + + augmentations: list = [] + + # Add multi-view video organizer if using multi-view wdinfo + if self.multi_view: + multi_view_organizer = MultiViewVideoOrganizer(camera_ids=self.cam_ids) + augmentations.append(multi_view_organizer) + + action_augmentor = ActionDataAugmentorMultiView( + fps_downsample_ratio=self.fps_downsample_ratio, + num_action_per_chunk=self.num_action_per_chunk, + accumulate_action=self.accumulate_action, + video_size=self.video_size, + normalize=self.normalize, + load_action=self.load_action, + load_t5_embeddings=self.load_t5_embeddings, + state_key=self.state_key, + gripper_key=self.gripper_key, + gripper_rescale_factor=self.gripper_rescale_factor, + cam_ids=self.cam_ids, + ) + augmentations.append(action_augmentor) + + for aug in augmentor_cfg.keys(): + augmentations.append(instantiate(augmentor_cfg[aug])) + + return partial(WebDatasetBase.augmentor_fn, augmentations=augmentations) + + +if __name__ == "__main__": + from cosmos_policy._src.imaginaire.config import ObjectStoreConfig + from cosmos_policy._src.imaginaire.datasets.webdataset.config.schema import DatasetConfig, DatasetInfo + from cosmos_policy._src.imaginaire.datasets.webdataset.distributors import ShardlistBasic + from cosmos_policy._src.imaginaire.utils import log + + dataset_info = DatasetInfo( + wdinfo=[ + "user/sync_gcp/pi_ablation_20251010/wdinfo/short_high_gripper_movement_segmented_episodes_30h_webdataset/wdinfo.json" + ], + object_store_config=ObjectStoreConfig( + enabled=True, bucket="debug", credentials="credentials/s3_robotics.secret" + ), + per_dataset_keys=[], + ) + + config = DatasetConfig( + dataset_info=[dataset_info], + keys=["videos", "annotations"], + streaming_download=True, + buffer_size=100, + augmentation={}, + distributor=ShardlistBasic(), + decoders=["rgb"], + remove_extension_from_keys=True, + ) + + dataset = ActionConditionedMultiViewWebDatasetS3( + config=config, + fps_downsample_ratio=2, + num_action_per_chunk=12, + cam_ids=[["base_0", "base_1"], "wrist"], + accumulate_action=False, + video_size=[480, 640], + load_action=True, + load_t5_embeddings=False, + state_key="ee_pose", + gripper_key="gripper_chunk", + gripper_rescale_factor=10.0, + ) + + webdataset = dataset.build_dataset() + log.info(f"Created multi-view WebDataset with {dataset.wdinfo.total_key_count} total keys") + + for i, sample in enumerate(webdataset): + log.info(f"Sample {i}: keys={list(sample.keys())}") + if "video" in sample: + log.info(f" video shape: {sample['video'].shape}") + if "action" in sample: + log.info(f" action shape: {sample['action'].shape}") + if i >= 2: + break diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/datasets/webdataset_s3.py b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/webdataset_s3.py new file mode 100644 index 0000000000000000000000000000000000000000..9e70d7513ebf43be413b34c4bbc93ae27176f815 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/datasets/webdataset_s3.py @@ -0,0 +1,761 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""WebDataset loader for action-conditional robot data from S3. + +This module provides a WebDataset-based loader for reading robot action data +stored in tar files on S3. It reuses the WebDataset infrastructure from +cosmos_policy._src.imaginaire and maintains compatibility with the output format of dataset_s3.py. + +Supports both single-view and multi-view datasets: +- Single-view: videos stored in videos/{tar_file} +- Multi-view: videos stored in videos/{camera_id}/{tar_file} + +For multi-view, the wdinfo.json should include: +{ + "multi_view": true, + "camera_ids": ["base_0", "base_1", "wrist"], + ... +} +""" + +import io +import json +import time +from typing import Callable + +import numpy as np +import torch +from decord import VideoReader, cpu +from omegaconf import DictConfig +from torchvision import transforms as T +from webdataset.handlers import reraise_exception + +from cosmos_policy._src.imaginaire.datasets.webdataset.config.schema import ( + AugmentorConfig, + DatasetConfig, + DatasetInfo, + TarSample, +) +from cosmos_policy._src.imaginaire.datasets.webdataset.webdataset import Dataset as WebDatasetBase +from cosmos_policy._src.imaginaire.lazy_config import instantiate +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.dataset_utils import Resize_Preprocess, ToTensorVideo, euler2rotm, rotm2euler +from cosmos_policy._src.imaginaire.utils.object_store import ObjectStore + + +class ActionDataAugmentor: + """Augmentor to transform WebDataset samples into action dataset format. + + This augmentor processes WebDataset tar samples containing robot trajectories + and formats them to match the output of dataset_s3.py. + """ + + def __init__( + self, + fps_downsample_ratio: int = 1, + num_action_per_chunk: int = 1, + accumulate_action: bool = False, + video_size: list[int] | None = None, + normalize: bool = False, + load_action: bool = True, + load_t5_embeddings: bool = False, + state_key: str = "state", + gripper_key: str = "gripper_chunk", + gripper_rescale_factor: float = 1.0, + cam_ids: list | None = None, + c_act_scaler: np.ndarray | None = None, + ): + """Initialize the action data augmentor. + + Args: + fps_downsample_ratio: Interval between sampled frames + num_action_per_chunk: Number of actions per sequence (NOT frames - we need num_action_per_chunk + 1 frames) + accumulate_action: Whether to accumulate actions relative to first frame + video_size: Target [H, W] for video frames + normalize: Whether to normalize video frames + load_action: Whether to load actions + load_t5_embeddings: Whether to load T5 embeddings + state_key: Key to access robot states + gripper_key: Key to access gripper states + gripper_rescale_factor: Scaling factor for gripper actions + cam_ids: List of camera IDs to sample from + c_act_scaler: Action scaling factors + """ + self.fps_downsample_ratio = fps_downsample_ratio + self.num_action_per_chunk = num_action_per_chunk + # sequence_length is the number of frames, which is num_action_per_chunk + 1 + # This matches the behavior of dataset_local.py + self.sequence_length = 1 + num_action_per_chunk + self.accumulate_action = accumulate_action + self.video_size = video_size + self.normalize = normalize + self.load_action = load_action + self.load_t5_embeddings = load_t5_embeddings + self.state_key = state_key + self.gripper_key = gripper_key + self.gripper_rescale_factor = gripper_rescale_factor + self.cam_ids = cam_ids or ["base_0"] + self.is_generator = True # Mark as generator for WebDataset + + # Default action scaler if not provided + if c_act_scaler is None: + self.c_act_scaler = np.array([20.0, 20.0, 20.0, 20.0, 20.0, 20.0, gripper_rescale_factor]) + else: + self.c_act_scaler = c_act_scaler + + # Initialize video transforms (matching dataset_local.py) + self.to_tensor_video = ToTensorVideo() + video_size_tuple = tuple(self.video_size) if self.video_size else (256, 256) + self.preprocess = T.Compose( + [ + ToTensorVideo(), + Resize_Preprocess(video_size_tuple), + T.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True), + ] + ) + self.not_norm_preprocess = T.Compose([ToTensorVideo(), Resize_Preprocess(video_size_tuple)]) + + def _get_robot_states(self, data: dict, frame_ids: list[int]) -> tuple[np.ndarray, np.ndarray]: + """Extract robot arm and gripper states for specified frames.""" + # Get states from the data dict - try multiple possible keys + state_keys_to_try = [ + self.state_key, + ] + states = None + for sk in state_keys_to_try: + if sk in data and data[sk] is not None: + states = data[sk] + break + + gripper_keys_to_try = [self.gripper_key] + gripper_states = None + for gk in gripper_keys_to_try: + if gk in data and data[gk] is not None: + gripper_states = data[gk] + break + + if states is None: + available_keys = list(data.keys()) + raise KeyError( + f"Could not find state data with key '{self.state_key}'. " + f"Tried keys: {state_keys_to_try}. Available keys in data: {available_keys}" + ) + + # Extract states for the requested frames + arm_states = [] + gripper_values = [] + + for frame_id in frame_ids: + if isinstance(states, (list, np.ndarray)): + state = states[frame_id] + else: + state = states + + # Handle different state formats + if isinstance(state, dict): + # Dictionary format with position and rotation + position = state.get("position", state.get("xyz", state.get("pos"))) + rotation = state.get("rotation", state.get("euler", state.get("rot"))) + arm_state = np.concatenate([position, rotation]) + else: + # Direct array format [x, y, z, rx, ry, rz] + arm_state = np.array(state[:6]) + + arm_states.append(arm_state) + + # Get gripper state + if gripper_states is not None: + if isinstance(gripper_states, (list, np.ndarray)): + gripper = gripper_states[frame_id] + else: + gripper = gripper_states + gripper_values.append(float(gripper)) + else: + # Default gripper state if not provided + gripper_values.append(0.0) + + return np.array(arm_states), np.array(gripper_values) + + def _get_actions(self, arm_states: np.ndarray, gripper_states: np.ndarray, accumulate_action: bool) -> torch.Tensor: + """Compute relative actions between consecutive frames. + + This matches the implementation in dataset_local.py exactly. + """ + l, _ = arm_states.shape + action = np.zeros((l - 1, 7), dtype=np.float32) + + if accumulate_action: + # Accumulate actions relative to base frame, with reset every 4 frames + # This matches dataset_local.py behavior + base_xyz = arm_states[0, 0:3] + base_rpy = arm_states[0, 3:6] + base_rotm = euler2rotm(base_rpy) + for k in range(1, l): + curr_xyz = arm_states[k, 0:3] + curr_rpy = arm_states[k, 3:6] + curr_gripper = gripper_states[k] + curr_rotm = euler2rotm(curr_rpy) + rel_xyz = np.dot(base_rotm.T, curr_xyz - base_xyz) + rel_rotm = base_rotm.T @ curr_rotm + rel_rpy = rotm2euler(rel_rotm) + action[k - 1, 0:3] = rel_xyz + action[k - 1, 3:6] = rel_rpy + action[k - 1, 6] = curr_gripper + if k % 4 == 0: + base_xyz = arm_states[k, 0:3] + base_rpy = arm_states[k, 3:6] + base_rotm = euler2rotm(base_rpy) + else: + # Compute relative actions between consecutive frames + for k in range(1, l): + prev_xyz = arm_states[k - 1, 0:3] + prev_rpy = arm_states[k - 1, 3:6] + prev_rotm = euler2rotm(prev_rpy) + curr_xyz = arm_states[k, 0:3] + curr_rpy = arm_states[k, 3:6] + curr_gripper = gripper_states[k] + curr_rotm = euler2rotm(curr_rpy) + rel_xyz = np.dot(prev_rotm.T, curr_xyz - prev_xyz) + rel_rotm = prev_rotm.T @ curr_rotm + rel_rpy = rotm2euler(rel_rotm) + action[k - 1, 0:3] = rel_xyz + action[k - 1, 3:6] = rel_rpy + action[k - 1, 6] = curr_gripper + + return torch.from_numpy(action) + + def _get_frames(self, sample: dict, frame_ids: list[int], cam_id: str, pre_encode: bool) -> torch.Tensor: + """Get video frames for a specific camera. + + Args: + sample: WebDataset sample containing video data. For multi-view, + videos are organized as sample["videos"][cam_id] = video_bytes. + For single-view, sample["video"] or sample["videos"]["video"] = video_bytes. + frame_ids: List of frame indices to extract + cam_id: Camera ID to look for + pre_encode: Whether to use pre-encoded videos + + Returns: + Video tensor of shape [T, C, H, W] + """ + if pre_encode: + raise NotImplementedError("Pre-encoded videos are not supported for this dataset.") + + # Try to get video data using the expected key format: videos_ + video_key = f"videos_{cam_id}" + video_data = sample.get(video_key) + + # If not found, try alternative structures + if video_data is None and "videos" in sample and isinstance(sample["videos"], dict): + if cam_id in sample["videos"]: + video_data = sample["videos"][cam_id] + elif "video" in sample["videos"]: + # Single video in videos dict + video_data = sample["videos"]["video"] + + # If still not found, raise a helpful error with available keys + if video_data is None: + available_keys = [k for k in sample.keys() if not k.startswith("__")] + raise KeyError( + f"Video data not found for camera '{cam_id}'. Tried key '{video_key}'. Available keys: {available_keys}" + ) + + # Handle nested dict structure - extract bytes from dict + if video_data is not None and isinstance(video_data, dict): + # Try common keys that might contain video bytes + for video_key in ["video", "video_path", "mp4", "data"]: + if video_key in video_data and isinstance(video_data[video_key], bytes): + video_data = video_data[video_key] + break + else: + # If no known key found, look for any bytes value + for v in video_data.values(): + if isinstance(v, bytes): + video_data = v + break + + if video_data is not None and isinstance(video_data, bytes): + frames = self._load_video(video_data, frame_ids) + frames = frames.astype(np.uint8) + frames = torch.from_numpy(frames).permute(0, 3, 1, 2) # (l, c, h, w) + elif video_data is not None: + extra_info = "" + if isinstance(video_data, dict): + extra_info = f", dict keys: {list(video_data.keys())}" + log.warning(f"Unexpected video data type: {type(video_data)}, expected bytes{extra_info}") + H, W = self.video_size or [256, 256] + frames = torch.zeros(len(frame_ids), 3, H, W, dtype=torch.uint8) + else: + H, W = self.video_size or [256, 256] + frames = torch.zeros(len(frame_ids), 3, H, W, dtype=torch.uint8) + + if self.normalize: + frames = self.preprocess(frames) + else: + frames = self.not_norm_preprocess(frames) + frames = torch.clamp(frames * 255.0, 0, 255).to(torch.uint8) + return frames + + def _get_obs(self, sample: dict, frame_ids: list[int], cam_id: str | None, pre_encode: bool): + """Get observation frames from the sample. + + Args: + sample: WebDataset sample containing video data + frame_ids: List of frame indices to extract + cam_id: Camera ID to use, or None to randomly select + pre_encode: Whether to use pre-encoded videos + + Returns: + Tuple of (video tensor [T, C, H, W], camera ID used) + """ + if cam_id is None: + selected_cam_id: str = np.random.choice(self.cam_ids) if len(self.cam_ids) > 1 else self.cam_ids[0] + else: + selected_cam_id = cam_id + frames = self._get_frames(sample, frame_ids, cam_id=selected_cam_id, pre_encode=pre_encode) + return frames, selected_cam_id + + def __call__(self, data_stream): + """Process WebDataset samples and yield formatted data.""" + sample_count = 0 + for sample in data_stream: + try: + # Debug: Log first few samples to understand what keys are present + if sample_count < 3: + all_keys = list(sample.keys()) + sample_count += 1 + + # Extract key information + key = sample.get("__key__", "unknown") + url = sample.get("__url__") # Preserve URL for downstream processing + + annotation = {} + ann_data = sample["annotations"] + # Handle both raw JSON strings and already-parsed dicts + if isinstance(ann_data, str): + annotation = json.loads(ann_data) + elif isinstance(ann_data, bytes): + annotation = json.loads(ann_data.decode("utf-8")) + elif isinstance(ann_data, dict): + annotation = ann_data + + # Determine frame indices to use + if "frame_ids" in sample: + frame_ids = sample["frame_ids"] + else: + # Create frame indices based on sequence length + # sequence_length = num_action_per_chunk + 1 (matching dataset_local.py behavior) + total_frames = len(annotation.get(self.state_key, [])) + frame_ids = list( + range( + 0, + min(total_frames, self.sequence_length * self.fps_downsample_ratio), + self.fps_downsample_ratio, + ) + ) + + # Get video frames using _get_obs (can be overridden for multi-view) + video_tensor, _ = self._get_obs(sample, frame_ids, cam_id="base_0", pre_encode=False) + # Permute from [T, C, H, W] to [C, T, H, W] + video_tensor = video_tensor.permute(1, 0, 2, 3) + + # Get robot states and compute actions + output_data = {"__key__": key} + if url is not None: + output_data["__url__"] = url + + if self.load_action: + # Merge annotation data with sample for state extraction + merged_data = {**sample, **annotation} + arm_states, gripper_states = self._get_robot_states(merged_data, frame_ids) + actions = self._get_actions(arm_states, gripper_states, self.accumulate_action) + actions *= self.c_act_scaler + output_data["action"] = actions.float() + + # Add video data + output_data["video"] = video_tensor.to(dtype=torch.uint8) + + # Add annotation file reference + output_data["annotation_file"] = f"tar:{key}" + + # Add T5 embeddings and metadata + if self.load_t5_embeddings and "t5_embeddings.npy" in sample: + t5_embeddings = np.load(sample["t5_embeddings.npy"]) + output_data["t5_text_embeddings"] = torch.from_numpy(t5_embeddings) + else: + output_data["t5_text_embeddings"] = torch.zeros(512, 1024, dtype=torch.bfloat16) + output_data["ai_caption"] = annotation.get("caption", "") + + output_data["t5_text_mask"] = torch.ones(512, dtype=torch.int64) + output_data["fps"] = 4 # Default FPS + output_data["image_size"] = 256 * torch.ones(4) + output_data["num_frames"] = len(frame_ids) + output_data["padding_mask"] = torch.zeros(1, 256, 256) + + yield output_data + + except Exception as e: + log.warning(f"Error processing sample {sample.get('__key__', 'unknown')}: {e}") + continue + + def _load_video(self, video_data: bytes, frame_ids: list[int]) -> np.ndarray: + """Process raw video data and extract frames. + + Args: + video_data: Raw video bytes + frame_ids: List of frame indices to extract + + Returns: + Video frames as numpy array of shape [T, H, W, C] + """ + vr = VideoReader(io.BytesIO(video_data), ctx=cpu(0), num_threads=2) + assert (np.array(frame_ids) < len(vr)).all() + assert (np.array(frame_ids) >= 0).all() + vr.seek(0) + frame_data = vr.get_batch(frame_ids).asnumpy() + return frame_data + + +class ActionConditionedWebDatasetS3(WebDatasetBase): + """WebDataset loader for action-conditional robot data from S3. + + This class extends the base WebDataset to load robot action data stored + in WebDataset tar format on S3, maintaining compatibility with the + output format of ActionConditionedDatasetS3. + + Supports both single-view and multi-view folder structures: + - Videos are expected at: {root}/videos/{camera_id}/{tar_file} + - Other data (annotations, etc.) at: {root}/{key}/{tar_file} + + For single-view, pass cam_ids=["base_0"] (or whichever single camera). + For multi-view, pass cam_ids=["base_0", "base_1", "wrist"]. + """ + + def __init__( + self, + config: DatasetConfig | DictConfig, + fps_downsample_ratio: int = 1, + num_action_per_chunk: int = 1, + cam_ids: list | None = None, + accumulate_action: bool = False, + video_size: list[int] | None = None, + normalize: bool = False, + load_action: bool = True, + load_t5_embeddings: bool = False, + state_key: str = "state", + gripper_key: str = "gripper_chunk", + gripper_rescale_factor: float = 1.0, + handler: Callable | None = None, + **kwargs, + ): + """Initialize the WebDataset S3 loader. + + Args: + config: WebDataset configuration with S3 paths + fps_downsample_ratio: Interval between sampled frames + num_action_per_chunk: Number of actions per sequence (NOT frames - we need num_action_per_chunk + 1 frames) + cam_ids: List of camera IDs to sample from (e.g., ["base_0"] for single-view) + accumulate_action: Whether to accumulate actions + video_size: Target [H, W] for video frames + normalize: Whether to normalize video frames + load_action: Whether to load actions + load_t5_embeddings: Whether to load T5 embeddings + state_key: Key to access robot states + gripper_key: Key to access gripper states + gripper_rescale_factor: Scaling factor for gripper + handler: Error handler + **kwargs: Additional arguments passed to base class + """ + # Store cam_ids BEFORE calling super().__init__() since parse_dataset_info needs it + self.cam_ids = cam_ids or ["base_0"] + + # Initialize base WebDataset with S3 support + if handler is None: + handler = reraise_exception + super().__init__(config=config, handler=handler) + + # Store action dataset specific parameters + self.fps_downsample_ratio = fps_downsample_ratio + self.num_action_per_chunk = num_action_per_chunk + self.accumulate_action = accumulate_action + self.video_size = video_size + self.normalize = normalize + self.load_action = load_action + self.load_t5_embeddings = load_t5_embeddings + self.state_key = state_key + self.gripper_key = gripper_key + self.gripper_rescale_factor = gripper_rescale_factor + + def parse_dataset_info(self, dataset_info: list[DatasetInfo], use_multithread: bool = True) -> None: + """Parse metadata about the list of tar files with camera-aware path expansion. + + This overrides the base method to expand video keys with camera IDs. + For example, if keys=["videos", "annotations"] and cam_ids=["base_0"], + the expanded keys become ["videos/base_0", "annotations"]. + + This allows loading from folder structures like: + videos/base_0/00000000.tar + annotations/00000000.tar + """ + log.info(f"[ActionDataset] Start parsing dataset info with {len(dataset_info)} entries") + log.info(f"[ActionDataset] Camera IDs: {self.cam_ids}") + tic = time.time() + + for dset_num, dset_info in enumerate(dataset_info): + if len(dset_info.wdinfo) == 0: + log.warning(f"No wdinfo found for dataset {dset_num}, skipping...") + continue + + use_object_store = dset_info.object_store_config.enabled + self.use_object_store = use_object_store + dset_id = f"dset: {dset_num}" + + if use_object_store: + object_store_reader = ObjectStore(config_object_storage=dset_info.object_store_config) + easy_io_backend_dset = object_store_reader.easy_io_backend + bucket_dset = dset_info.object_store_config.bucket + else: + object_store_reader = None + easy_io_backend_dset = None + bucket_dset = None + + tar_samples = [] + total_key_count = 0 + chunk_sizes = [] + + for wdinfo_path in dset_info.wdinfo: + log.info(f"[ActionDataset] Processing wdinfo: {wdinfo_path}") + + if use_object_store: + if not object_store_reader.object_exists(wdinfo_path): + raise FileNotFoundError(f"{wdinfo_path} not found") + cur_dset_info = object_store_reader.load_object(key=wdinfo_path, type="json") + else: + with open(wdinfo_path) as fp: + cur_dset_info = json.load(fp) + + # Debug: Log wdinfo content for troubleshooting + log.info( + f"[ActionDataset] wdinfo content: root={cur_dset_info.get('root')}, " + f"data_keys={cur_dset_info.get('data_keys')}, " + f"num_tars={len(cur_dset_info.get('data_list', []))}" + ) + + data_root = cur_dset_info["root"] + # Strip s3://bucket/ prefix from root if present + if data_root.startswith("s3://"): + parts = data_root[5:].split("/", 1) + data_root = parts[1] if len(parts) > 1 else "" + + tar_files_list = cur_dset_info["data_list"] + + # Get data keys from wdinfo or fall back to config keys + data_keys = cur_dset_info.get("data_keys", self.data_keys) + + # Expand video keys with camera IDs + # e.g., ["videos", "annotations"] -> ["videos/base_0", "annotations"] + expanded_keys = [] + for key in data_keys: + if key == "videos": + # Add a key for each camera: "videos/base_0", "videos/base_1", etc. + for cam_id in self.cam_ids: + expanded_keys.append(f"videos/{cam_id}") + else: + expanded_keys.append(key) + + log.info(f"[ActionDataset] Original keys: {data_keys}") + log.info(f"[ActionDataset] Expanded keys: {expanded_keys}") + + # Debug: Show example tar paths that will be constructed + if tar_files_list: + example_tar = tar_files_list[0] + log.info(f"[ActionDataset] Example tar paths for '{example_tar}':") + for key in expanded_keys: + full_path = f"{data_root}/{key}/{example_tar}" + log.info(f"[ActionDataset] -> {full_path}") + + local_tar_samples = [ + TarSample( + path=tar_file, + root=data_root, + keys=(dset_info.per_dataset_keys if dset_info.per_dataset_keys else expanded_keys), + meta=dset_info, + dset_id=dset_id, + sample_keys_full_list=None, + ) + for tar_file in tar_files_list + ] + + tar_samples.extend(local_tar_samples) + total_key_count += cur_dset_info["total_key_count"] + chunk_sizes.append(cur_dset_info["chunk_size"]) + + # Store results + self.wdinfo.tar_files.extend(tar_samples) + self.wdinfo.total_key_count += total_key_count + if chunk_sizes: + self.wdinfo.chunk_size = chunk_sizes[0] + if easy_io_backend_dset: + self.easy_io_backend[dset_id] = easy_io_backend_dset + if bucket_dset: + self.bucket[dset_id] = bucket_dset + + toc = time.time() + log.info( + f"[ActionDataset] Parsed {len(dataset_info)} wdinfos " + f"(num_keys={self.wdinfo.total_key_count}, num_tars={len(self.wdinfo.tar_files)}) " + f"in {(toc - tic):.2f}s" + ) + + def build_data_augmentor(self, augmentor_cfg: dict[str, AugmentorConfig]) -> Callable: + """Build data augmentors including the action data processor. + + This overrides the base method to add our custom ActionDataAugmentor + at the beginning of the augmentation pipeline. + """ + # Create action data augmentor + action_augmentor = ActionDataAugmentor( + fps_downsample_ratio=self.fps_downsample_ratio, + num_action_per_chunk=self.num_action_per_chunk, + accumulate_action=self.accumulate_action, + video_size=self.video_size, + normalize=self.normalize, + load_action=self.load_action, + load_t5_embeddings=self.load_t5_embeddings, + state_key=self.state_key, + gripper_key=self.gripper_key, + gripper_rescale_factor=self.gripper_rescale_factor, + cam_ids=self.cam_ids, + ) + + # Build other augmentors from config + augmentations = [action_augmentor] + for aug in augmentor_cfg.keys(): + augmentations.append(instantiate(augmentor_cfg[aug])) + + # Return the augmentor function + from functools import partial + + return partial(WebDatasetBase.augmentor_fn, augmentations=augmentations) + + +# Example usage and testing +if __name__ == "__main__": + """Test the WebDataset S3 loader. + + Run with: + PYTHONPATH=. python cosmos_policy/_src/predict2/action/datasets/webdataset_s3.py + """ + + from cosmos_policy._src.imaginaire.config import ObjectStoreConfig + from cosmos_policy._src.imaginaire.datasets.webdataset.config.schema import DatasetConfig, DatasetInfo + from cosmos_policy._src.imaginaire.datasets.webdataset.distributors import ShardlistBasic + + # Create dataset info objects + dataset_info = DatasetInfo( + wdinfo=[ + "user/sync_gcp/pi_ablation_20251010/wdinfo/short_high_gripper_movement_segmented_episodes_30h_webdataset/wdinfo.json" + ], + object_store_config=ObjectStoreConfig( + enabled=True, bucket="debug", credentials="credentials/s3_robotics.secret" + ), + per_dataset_keys=[], # Use empty list instead of None + ) + + # Create the dataset configuration + config = DatasetConfig( + dataset_info=[dataset_info], + keys=["video", "json", "states"], + streaming_download=True, + buffer_size=100, + augmentation={}, + distributor=ShardlistBasic(), + decoders=["rgb", "json"], + remove_extension_from_keys=True, + ) + + # Create dataset + dataset = ActionConditionedWebDatasetS3( + config=config, + fps_downsample_ratio=1, + num_action_per_chunk=16, + cam_ids=["base_0"], + accumulate_action=False, + video_size=[256, 256], + load_action=True, + load_t5_embeddings=False, + state_key="ee_pose", + ) + + # Build the actual dataset + webdataset = dataset.build_dataset() + + log.info(f"Created WebDataset with {dataset.wdinfo.total_key_count} total keys") + + # Example of expected tar file structure: + # Each tar file should contain samples with the following files per sample: + # - {sample_id}.json: annotation file with robot states, episode metadata + # - {sample_id}.base_0.mp4: video from camera "base_0" + # - {sample_id}.t5_embeddings.npy: (optional) T5 text embeddings + # + # The json file should contain: + # { + # "episode_id": "episode_001", + # "ee_pose": [...], // or "state" - list of robot states per frame + # "gripper_state": [...], // gripper states per frame + # "caption": "robot picking up object" // optional text description + # } + # + # === SINGLE-VIEW wdinfo.json === + # The wdinfo.json file should follow the standard WebDataset format: + # { + # "root": "s3://bucket/path/to/tars/", + # "data_list": ["shard_000.tar", "shard_001.tar", ...], + # "total_key_count": 10000, + # "chunk_size": 1000 + # } + # + # === MULTI-VIEW === + # For multi-view datasets, see webdataset_mv_s3.py which provides: + # - ActionConditionedMultiViewWebDatasetS3: extends this class for multi-view + # - ActionDataAugmentorMultiView: augmentor that concatenates camera views + # + # Multi-view wdinfo.json format: + # { + # "root": "s3://bucket/path/to/tars/", + # "data_list": ["00000000.tar", "00000001.tar", ...], + # "data_keys": ["videos", "annotations", "actions"], + # "multi_view": true, + # "camera_ids": ["base_0", "base_1", "wrist"], + # "total_key_count": 10000, + # "chunk_size": 100 + # } + # + # Multi-view directory structure: + # output_dir/ + # ├── videos/ + # │ ├── base_0/ + # │ │ ├── 00000000.tar + # │ │ └── 00000001.tar + # │ ├── base_1/ + # │ │ └── ... + # │ └── wrist/ + # │ └── ... + # ├── annotations/ + # │ ├── 00000000.tar + # │ └── 00000001.tar + # └── actions/ + # └── ... diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference.py b/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference.py new file mode 100644 index 0000000000000000000000000000000000000000..c94aadb51c0c2dcdf0d2d29c27c7eaa598d287ef --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference.py @@ -0,0 +1,343 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" + + +# ---------------------------------- benchmark ---------------------------------- +CUDA_VISIBLE_DEVICES=0 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference.py \ +--experiment=cosmos_predict2p5_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_256x320 \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_256x320/checkpoints/iter_000016000 \ + --input_video_root /project/cosmos/user/bridge/ \ + --input_json_sub_folder annotation/test_100 \ + --save_root results/cosmos_predict2p5_2B_reason_embeddings_action_conditioned_rectified_flow_bridge_13frame_256x320-val-16k \ + --resolution 256,320 --guidance 0 --chunk_size 12 --camera_id 0 --save_fps 4 +""" + +import argparse +import json +import os +from glob import glob + +import mediapy +import numpy as np +import torch +from loguru import logger + +from cosmos_policy._src.imaginaire.utils import distributed +from cosmos_policy._src.predict2.action.datasets.dataset_utils import euler2rotm, rotm2euler, rotm2quat +from cosmos_policy._src.predict2.action.inference.inference_pipeline import ( + _DEFAULT_NEGATIVE_PROMPT, + ActionVideo2WorldInference, +) + +_IMAGE_EXTENSIONS = [".png", ".jpg", ".jpeg", "webp"] +_VIDEO_EXTENSIONS = [".mp4"] + +_ACTION_SCALER = 20.0 + + +def parse_arguments() -> argparse.Namespace: + """Parses command-line arguments for the Video2World inference script.""" + parser = argparse.ArgumentParser(description="Image2World/Video2World inference script") + parser.add_argument("--experiment", type=str, required=True, help="Experiment config") + parser.add_argument("--chunk_size", type=int, default=12, help="Chunk size for action conditioning") + parser.add_argument("--guidance", type=int, default=7, help="Guidance value") + parser.add_argument("--seed", type=int, default=1, help="Guidance value") + parser.add_argument( + "--ckpt_path", + type=str, + default="", + help="Path to the checkpoint. If not provided, will use the one specify in the config", + ) + parser.add_argument("--s3_cred", type=str, default="credentials/s3_checkpoint.secret") + parser.add_argument( + "--resolution", + type=str, + default="none", + help="Resolution of the video (H,W). Be default it will use model trained resolution. 9:16", + ) + parser.add_argument("--input_video_root", type=str, default="bridge/annotation/test_100", help="Action root") + parser.add_argument("--input_json_sub_folder", type=str, default="bridge/annotation/test_100", help="Action root") + parser.add_argument("--save_root", type=str, default="results/image2world", help="Save root") + + # for pi dataset + parser.add_argument("--camera_id", type=str, default="base", help="Camera id") + parser.add_argument("--start", type=int, default=0) + parser.add_argument("--end", type=int, default=100) + parser.add_argument("--fps_downsample_ratio", type=int, default=1) + parser.add_argument("--gripper_scale", type=float, default=1.0) + parser.add_argument("--gripper_key", type=str, default="continuous_gripper_state", help="Gripper key") + parser.add_argument("--state_key", type=str, default="state", help="State key") + + parser.add_argument("--reverse", action="store_true", help="Reverse the video") + parser.add_argument("--single_chunk", action="store_true", help="Single chunk") + parser.add_argument("--start_frame_idx", type=int, default=0, help="Start frame index") + parser.add_argument("--save_fps", type=int, default=20, help="Save fps") + + parser.add_argument( + "--negative_prompt", + type=str, + default=_DEFAULT_NEGATIVE_PROMPT, + help="Custom negative prompt for classifier-free guidance. If not specified, uses default embeddings from S3.", + ) + parser.add_argument( + "--num_latent_conditional_frames", + type=int, + default=1, + help="Number of latent conditional frames (0, 1 or 2). For images, both values work by duplicating frames. For videos, uses the first N frames.", + ) + # Context parallel arguments + parser.add_argument( + "--context_parallel_size", + type=int, + default=1, + help="Context parallel size (number of GPUs to split context over). Set to 8 for 8 GPUs", + ) + return parser.parse_args() + + +def _get_robot_states(label, state_key="state", gripper_key="continuous_gripper_state"): + """ + Extracts the robot arm and gripper states from the label dictionary for the specified frame indices. + + Args: + label (dict): Dictionary containing robot state information, with keys "state" and "continuous_gripper_state". + frame_ids (list or np.ndarray): List or array of frame indices to extract. + + Returns: + tuple: + - np.ndarray: Array of arm states for the selected frames, shape (len(frame_ids), state_dim). + - np.ndarray: Array of gripper states for the selected frames, shape (len(frame_ids),). + """ + + all_states = np.array(label[state_key]) + all_cont_gripper_states = np.array(label[gripper_key]) + + return all_states, all_cont_gripper_states + + +def _get_actions(arm_states, gripper_states, sequence_length, use_quat=False): + """ + Compute the relative actions between consecutive robot states. + + Args: + arm_states (np.ndarray): Array of arm states with shape (sequence_length, 6), where each state contains + [x, y, z, roll, pitch, yaw] or similar. + gripper_states (np.ndarray): Array of gripper states with shape (sequence_length,). + sequence_length (int): Number of states in the sequence. + use_quat (bool): If True, represent rotation as quaternion; otherwise, use Euler angles. + + Returns: + np.ndarray: Array of actions with shape (sequence_length - 1, 7), where each action contains + [relative_xyz (3), relative_rotation (3), gripper_state (1)]. + """ + + if use_quat: + action = np.zeros((sequence_length - 1, 8)) + else: + action = np.zeros((sequence_length - 1, 7)) + + for k in range(1, sequence_length): + prev_xyz = arm_states[k - 1, 0:3] + prev_rpy = arm_states[k - 1, 3:6] + prev_rotm = euler2rotm(prev_rpy) + curr_xyz = arm_states[k, 0:3] + curr_rpy = arm_states[k, 3:6] + curr_gripper = gripper_states[k] + curr_rotm = euler2rotm(curr_rpy) + rel_xyz = np.dot(prev_rotm.T, curr_xyz - prev_xyz) + rel_rotm = prev_rotm.T @ curr_rotm + + if use_quat: + rel_rot = rotm2quat(rel_rotm) + action[k - 1, 0:3] = rel_xyz + action[k - 1, 3:7] = rel_rot + action[k - 1, 7] = curr_gripper + else: + rel_rot = rotm2euler(rel_rotm) + action[k - 1, 0:3] = rel_xyz + action[k - 1, 3:6] = rel_rot + action[k - 1, 6] = curr_gripper + return action # (l - 1, act_dim) + + +def get_action_sequence_from_states( + data, + fps_downsample_ratio=1, + use_quat=False, + state_key="state", + gripper_scale=1.0, + gripper_key="continuous_gripper_state", +): + """ + Get the action sequence from the states. + """ + + arm_states, cont_gripper_states = _get_robot_states(data, state_key, gripper_key) + actions = _get_actions( + arm_states[::fps_downsample_ratio], + cont_gripper_states[::fps_downsample_ratio], + len(data[state_key][::fps_downsample_ratio]), + use_quat=use_quat, + ) + actions *= np.array( + [_ACTION_SCALER, _ACTION_SCALER, _ACTION_SCALER, _ACTION_SCALER, _ACTION_SCALER, _ACTION_SCALER, gripper_scale] + ) + + return actions + + +def get_video_id(img_path: str): + """Extract video ID from image path by removing directory and extension.""" + return img_path.split("/")[-1].split(".")[0] + + +def main(): + torch.enable_grad(False) # Disable gradient calculations for inference + args = parse_arguments() + + # Validate num_latent_conditional_frames at the very beginning + if args.num_latent_conditional_frames not in [0, 1, 2]: + raise ValueError( + f"num_latent_conditional_frames must be 0, 1 or 2, but got {args.num_latent_conditional_frames}" + ) + + # Determine supported extensions based on num_latent_conditional_frames + if args.num_latent_conditional_frames > 1: + supported_extensions = _VIDEO_EXTENSIONS + # Check if input folder contains any videos + has_videos = False + for file_name in os.listdir(args.input_root): + file_ext = os.path.splitext(file_name)[1].lower() + if file_ext in _VIDEO_EXTENSIONS: + has_videos = True + break + + if not has_videos: + raise ValueError( + f"num_latent_conditional_frames={args.num_latent_conditional_frames} > 1 requires video inputs, " + f"but no videos found in {args.input_root}. Found extensions: " + f"{set(os.path.splitext(f)[1].lower() for f in os.listdir(args.input_root) if os.path.splitext(f)[1])}" + ) + + logger.info(f"Using video-only mode with {args.num_latent_conditional_frames} conditional frames") + elif args.num_latent_conditional_frames == 1: + supported_extensions = _IMAGE_EXTENSIONS + _VIDEO_EXTENSIONS + logger.info(f"Using image+video mode with {args.num_latent_conditional_frames} conditional frame") + + # Initialize the inference handler with context parallel support + video2world_cli = ActionVideo2WorldInference( + args.experiment, args.ckpt_path, args.s3_cred, context_parallel_size=args.context_parallel_size + ) + + mem_bytes = torch.cuda.memory_allocated(device=torch.device("cuda" if torch.cuda.is_available() else "cpu")) + logger.info(f"GPU memory usage after model dcp.load: {mem_bytes / (1024**3):.2f} GB") + + # get input video and annotation path + input_video_path = os.path.join(args.input_video_root) + input_json_path = os.path.join(args.input_video_root, args.input_json_sub_folder) + input_json_list = glob(os.path.join(input_json_path, "*.json")) + + # Only process files on rank 0 if using distributed processing + rank0 = True + if args.context_parallel_size > 1: + rank0 = distributed.get_rank() == 0 + + # Ensure save directory exists + os.makedirs(args.save_root, exist_ok=True) + + # Process each file in the input directory + for annotation_path in input_json_list[args.start : args.end]: + with open(annotation_path, "r") as f: + json_data = json.load(f) + + # Convert camera_id to integer if it's a string and can be converted to an integer + camera_id = ( + int(args.camera_id) if isinstance(args.camera_id, str) and args.camera_id.isdigit() else args.camera_id + ) + + if isinstance(json_data["videos"][camera_id], dict): + video_path = os.path.join(input_video_path, json_data["videos"][camera_id]["video_path"]) + else: + video_path = os.path.join(input_video_path, json_data["videos"][camera_id]) + + actions = get_action_sequence_from_states( + json_data, + fps_downsample_ratio=args.fps_downsample_ratio, + state_key=args.state_key, + gripper_scale=args.gripper_scale, + gripper_key=args.gripper_key, + ) + + actions = actions[: len(actions)] + video_array = mediapy.read_video(video_path) + img_array = video_array[args.start_frame_idx] + + # Resize img_array with arg.resolution if specified + if args.resolution != "none": + try: + h, w = map(int, args.resolution.split(",")) + img_array = mediapy.resize_image(img_array, (h, w)) + except Exception as e: + logger.warning(f"Failed to resize image to {args.resolution}: {e}") + + img_name = annotation_path.split("/")[-1].split(".")[0] + + frames = [img_array] + chunk_video = [] + + video_name = f"{args.save_root}/{img_name.replace('.jpg', '.mp4')}" + chunk_video_name = f"{args.save_root}/{img_name + '_chunk.mp4'}" + logger.info(f"Saving video to {video_name}") + if os.path.exists(chunk_video_name): + logger.info(f"Video already exists: {chunk_video_name}") + continue + + for i in range(args.start_frame_idx, len(actions), args.chunk_size): + next_img_array, video_clamped = video2world_cli.step_inference( + img_array=img_array, + action=actions[i : i + args.chunk_size], + guidance=args.guidance, + seed=i, + ) + frames.append(next_img_array) + img_array = next_img_array + chunk_video.append(video_clamped) + + if args.single_chunk: + break + + chunk_list = [chunk_video[0]] + [chunk_video[i][: args.chunk_size] for i in range(1, len(chunk_video))] + chunk_video = np.concatenate(chunk_list, axis=0) + if args.single_chunk: + chunk_video_name = f"{args.save_root}/{img_name + '_single_chunk.mp4'}" + else: + chunk_video_name = f"{args.save_root}/{img_name + '_chunk.mp4'}" + + if rank0: + mediapy.write_video(chunk_video_name, chunk_video, fps=args.save_fps) + logger.info(f"Saved video to {chunk_video_name}") + + # Synchronize all processes before cleanup + if args.context_parallel_size > 1: + torch.distributed.barrier() + + # Clean up distributed resources + video2world_cli.cleanup() + + +if __name__ == "__main__": + main() diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference_gr00t.py b/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference_gr00t.py new file mode 100644 index 0000000000000000000000000000000000000000..015dd599af245d6abd6d2a6c973a49da957812be --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference_gr00t.py @@ -0,0 +1,364 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" + + +# ---------------------------------- benchmark ---------------------------------- + + +CUDA_VISIBLE_DEVICES=0 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame/checkpoints/iter_000014000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame-14k\ + --resolution 480,832 --guidance 0 --chunk_size 12 + +CUDA_VISIBLE_DEVICES=7 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame/checkpoints/iter_000020000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame-20k\ + --resolution 480,832 --guidance 0 --chunk_size 12 --start 80 --end 100 + +CUDA_VISIBLE_DEVICES=1 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame/checkpoints/iter_000028000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame-28k\ + --resolution 480,832 --guidance 0 --chunk_size 12 --start 0 --end 100 + +CUDA_VISIBLE_DEVICES=7 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes/checkpoints/iter_000004000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes-4k\ + --resolution 480,832 --guidance 0 --chunk_size 12 --start 80 --end 100 + +CUDA_VISIBLE_DEVICES=0 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_49frame_full \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_49frame_full/checkpoints/iter_000004000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_49frame_full-6k\ + --resolution 480,832 --guidance 0 --chunk_size 48 --start 80 --end 100 + +CUDA_VISIBLE_DEVICES=0 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_conditioned_posttrained_rl_merged_action_gr00t_gr1_customized_13frame_full \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_conditioned_posttrained_rl_merged_action_gr00t_gr1_customized_13frame_full/checkpoints/iter_000004000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_conditioned_posttrained_rl_merged_action_gr00t_gr1_customized_13frame_full-4k\ + --resolution 480,832 --guidance 0 --chunk_size 12 --start 0 --end 100 + +CUDA_VISIBLE_DEVICES=1 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_conditioned_posttrained_rl_merged_action_gr00t_gr1_customized_13frame_full \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_conditioned_posttrained_rl_merged_action_gr00t_gr1_customized_13frame_full/checkpoints/iter_000008000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_conditioned_posttrained_rl_merged_action_gr00t_gr1_customized_13frame_full-8k\ + --resolution 480,832 --guidance 0 --chunk_size 12 --start 0 --end 100 + +CUDA_VISIBLE_DEVICES=6 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_49frame_full \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_49frame_full/checkpoints/iter_000010000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_49frame_full-10k\ + --resolution 480,832 --guidance 0 --chunk_size 48 --start 90 --end 100 + +CUDA_VISIBLE_DEVICES=6 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_73frame_full \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_73frame_full/checkpoints/iter_000006000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_73frame_full-6k\ + --resolution 480,832 --guidance 0 --chunk_size 72 --start 70 --end 80 + +CUDA_VISIBLE_DEVICES=5 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes/checkpoints/iter_000008000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes-8k\ + --resolution 480,832 --guidance 0 --chunk_size 12 --start 75 --end 100 + + +CUDA_VISIBLE_DEVICES=6 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes/checkpoints/iter_000010000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes-10k\ + --resolution 480,832 --guidance 0 --chunk_size 12 --start 90 --end 100 + +CUDA_VISIBLE_DEVICES=7 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes/checkpoints/iter_000014000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes-14k\ + --resolution 480,832 --guidance 0 --chunk_size 12 --start 90 --end 100 + + +CUDA_VISIBLE_DEVICES=7 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes/checkpoints/iter_000016000 \ + --input_video_root results/gr00t_gr1/gt \ + --save_root results/gr00t_gr1/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes-16k\ + --resolution 480,832 --guidance 0 --chunk_size 12 --start 0 --end 100 +""" + +import argparse +import os +from glob import glob + +import mediapy +import numpy as np +import torch +from loguru import logger + +from cosmos_policy._src.imaginaire.utils import distributed +from cosmos_policy._src.predict2.action.inference.inference_pipeline import ( + _DEFAULT_NEGATIVE_PROMPT, + ActionVideo2WorldInference, +) + +_IMAGE_EXTENSIONS = [".png", ".jpg", ".jpeg", "webp"] +_VIDEO_EXTENSIONS = [".mp4"] + +_ACTION_SCALER = 20.0 + + +def parse_arguments() -> argparse.Namespace: + """Parses command-line arguments for the Video2World inference script.""" + parser = argparse.ArgumentParser(description="Image2World/Video2World inference script") + parser.add_argument("--experiment", type=str, required=True, help="Experiment config") + parser.add_argument("--chunk_size", type=int, default=12, help="Chunk size for action conditioning") + parser.add_argument( + "--num_chunks", type=int, default=12, help="Number of chunks to generate (-1 for all available chunks)" + ) + parser.add_argument("--guidance", type=int, default=7, help="Guidance value") + parser.add_argument("--seed", type=int, default=1, help="Guidance value") + parser.add_argument( + "--ckpt_path", + type=str, + default="", + help="Path to the checkpoint. If not provided, will use the one specify in the config", + ) + parser.add_argument("--s3_cred", type=str, default="credentials/s3_checkpoint.secret") + parser.add_argument( + "--resolution", + type=str, + default="none", + help="Resolution of the video (H,W). Be default it will use model trained resolution. 9:16", + ) + parser.add_argument("--input_video_root", type=str, default="bridge/annotation/test_100", help="Action root") + parser.add_argument("--save_root", type=str, default="results/image2world", help="Save root") + + # for pi dataset + parser.add_argument("--camera_id", type=str, default="base", help="Camera id") + parser.add_argument("--start", type=int, default=0) + parser.add_argument("--end", type=int, default=100) + parser.add_argument("--fps_downsample_ratio", type=int, default=1) + parser.add_argument("--gripper_scale", type=float, default=1.0) + parser.add_argument("--gripper_key", type=str, default="continuous_gripper_state", help="Gripper key") + parser.add_argument("--state_key", type=str, default="state", help="State key") + + parser.add_argument("--reverse", action="store_true", help="Reverse the video") + parser.add_argument("--single_chunk", action="store_true", help="Single chunk") + parser.add_argument("--start_frame_idx", type=int, default=0, help="Start frame index") + parser.add_argument("--save_fps", type=int, default=10, help="Save fps") + + parser.add_argument( + "--negative_prompt", + type=str, + default=_DEFAULT_NEGATIVE_PROMPT, + help="Custom negative prompt for classifier-free guidance. If not specified, uses default embeddings from S3.", + ) + parser.add_argument( + "--num_latent_conditional_frames", + type=int, + default=1, + help="Number of latent conditional frames (0, 1 or 2). For images, both values work by duplicating frames. For videos, uses the first N frames.", + ) + # Context parallel arguments + parser.add_argument( + "--context_parallel_size", + type=int, + default=1, + help="Context parallel size (number of GPUs to split context over). Set to 8 for 8 GPUs", + ) + return parser.parse_args() + + +def get_action_sequence_from_states( + data, + fps_downsample_ratio=1, + use_quat=False, + state_key="state", + gripper_scale=1.0, + gripper_key="continuous_gripper_state", +): + """ + Get the action sequence from the states. + """ + + actions = np.array(data["action"])[::fps_downsample_ratio][:-1] + return actions + + +def get_video_id(img_path: str): + """Extract video ID from image path by removing directory and extension.""" + return img_path.split("/")[-1].split(".")[0] + + +def main(): + torch.enable_grad(False) # Disable gradient calculations for inference + args = parse_arguments() + + # Validate num_latent_conditional_frames at the very beginning + if args.num_latent_conditional_frames not in [0, 1, 2]: + raise ValueError( + f"num_latent_conditional_frames must be 0, 1 or 2, but got {args.num_latent_conditional_frames}" + ) + + # Determine supported extensions based on num_latent_conditional_frames + if args.num_latent_conditional_frames > 1: + supported_extensions = _VIDEO_EXTENSIONS + # Check if input folder contains any videos + has_videos = False + for file_name in os.listdir(args.input_root): + file_ext = os.path.splitext(file_name)[1].lower() + if file_ext in _VIDEO_EXTENSIONS: + has_videos = True + break + + if not has_videos: + raise ValueError( + f"num_latent_conditional_frames={args.num_latent_conditional_frames} > 1 requires video inputs, " + f"but no videos found in {args.input_root}. Found extensions: " + f"{set(os.path.splitext(f)[1].lower() for f in os.listdir(args.input_root) if os.path.splitext(f)[1])}" + ) + + logger.info(f"Using video-only mode with {args.num_latent_conditional_frames} conditional frames") + elif args.num_latent_conditional_frames == 1: + supported_extensions = _IMAGE_EXTENSIONS + _VIDEO_EXTENSIONS + logger.info(f"Using image+video mode with {args.num_latent_conditional_frames} conditional frame") + + # Initialize the inference handler with context parallel support + video2world_cli = ActionVideo2WorldInference( + args.experiment, args.ckpt_path, args.s3_cred, context_parallel_size=args.context_parallel_size + ) + + mem_bytes = torch.cuda.memory_allocated(device=torch.device("cuda" if torch.cuda.is_available() else "cpu")) + logger.info(f"GPU memory usage after model dcp.load: {mem_bytes / (1024**3):.2f} GB") + + # get input video and annotation path + input_video_path = os.path.join(args.input_video_root) + + # Only process files on rank 0 if using distributed processing + rank0 = True + if args.context_parallel_size > 1: + rank0 = distributed.get_rank() == 0 + + # pdb.set_trace() + video_list = glob(os.path.join(input_video_path, "*.mp4")) + input_json_list = [video_path.replace(".mp4", "_actions.npy") for video_path in video_list] + + # Ensure save directory exists + os.makedirs(args.save_root, exist_ok=True) + + # Process each file in the input directory + for annotation_path, video_path in zip(input_json_list[args.start : args.end], video_list[args.start : args.end]): + actions = np.load(annotation_path) + + # Convert camera_id to integer if it's a string and can be converted to an integer + + actions = actions[: len(actions)] + video_array = mediapy.read_video(video_path) + + # Resize video_array with arg.resolution if specified + if args.resolution != "none": + try: + h, w = map(int, args.resolution.split(",")) + video_array = np.stack([mediapy.resize_image(frame, (h, w)) for frame in video_array], axis=0) + except Exception as e: + logger.warning(f"Failed to resize video to {args.resolution}: {e}") + + img_array = video_array[args.start_frame_idx] + # img_name = annotation_path.split("/")[-1].split(".")[0] + img_name = video_path.split("/")[-1].split(".")[0] + + frames = [img_array] + chunk_video = [] + video_array = video_array[:: args.fps_downsample_ratio] + + video_name = f"{args.save_root}/{img_name.replace('.jpg', '.mp4')}" + chunk_video_name = f"{args.save_root}/{img_name + '.mp4'}" + logger.info(f"Saving video to {video_name}") + if os.path.exists(chunk_video_name): + logger.info(f"Video already exists: {chunk_video_name}") + continue + + # Calculate the maximum number of chunks to generate + max_chunks = len(actions) // args.chunk_size + if args.num_chunks > 0: + max_chunks = min(max_chunks, args.num_chunks) + + logger.info(f"Generating {max_chunks} chunks (chunk_size={args.chunk_size}, total_actions={len(actions)})") + + chunk_count = 0 + for i in range(args.start_frame_idx, len(actions), args.chunk_size): + if actions[i : i + args.chunk_size].shape[0] != args.chunk_size: + break + + # Check if we've reached the desired number of chunks + if args.num_chunks > 0 and chunk_count >= args.num_chunks: + logger.info(f"Reached target number of chunks ({args.num_chunks}), stopping generation") + break + + logger.info(f"Generating chunk {chunk_count + 1}/{max_chunks}") + next_img_array, video_clamped = video2world_cli.step_inference( + img_array=img_array, + action=actions[i : i + args.chunk_size], + guidance=args.guidance, + seed=i, + ) + frames.append(next_img_array) + img_array = next_img_array + chunk_video.append(video_clamped) + chunk_count += 1 + + if args.single_chunk: + break + + chunk_list = [chunk_video[0]] + [chunk_video[i][: args.chunk_size] for i in range(1, len(chunk_video))] + chunk_video = np.concatenate(chunk_list, axis=0) + if args.single_chunk: + chunk_video_name = f"{args.save_root}/{img_name + '_single_chunk.mp4'}" + else: + # chunk_video_name = f"{args.save_root}/{img_name + '_chunk.mp4'}" + chunk_video_name = f"{args.save_root}/{img_name + '.mp4'}" + mediapy.write_video(chunk_video_name, chunk_video, fps=args.save_fps) + + # concat_video = np.concatenate([chunk_video, video_array[: chunk_video.shape[0]]], axis=2) + # concat_video_name = f"{args.save_root}/{img_name + '_concat.mp4'}" + # mediapy.write_video(concat_video_name, concat_video, fps=args.save_fps) + + logger.info(f"Saved video to {chunk_video_name}") + + # Synchronize all processes before cleanup + if args.context_parallel_size > 1: + torch.distributed.barrier() + + # Clean up distributed resources + video2world_cli.cleanup() + + +if __name__ == "__main__": + main() diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference_gr00t_warmup.py b/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference_gr00t_warmup.py new file mode 100644 index 0000000000000000000000000000000000000000..fe1f9626d11f39d227ef7d93e1ca211eccf9a2a7 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference_gr00t_warmup.py @@ -0,0 +1,194 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Please run the script with the following command: + +CUDA_VISIBLE_DEVICES=0 PYTHONPATH=. python cosmos_policy/_src/predict2/action/inference/inference_gr00t_warmup.py \ +--experiment=cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes_release \ + --ckpt_path s3://bucket/cosmos_predict2_action_conditioned/action_conditional/cosmos_predict2p5_2B_action_conditioned_gr00t_gr1_customized_13frame_full_16nodes/checkpoints/iter_000014000/model \ + --input_video_root /project/cosmos/user/gr00t_gr1 \ + --save_root datasets/gr1_warmup_regenerated_4step \ + --base_path_gr00t_gr1_local /project/cosmos/user/datasets/gr1_unified/gr1_unified.RU0226RemoveStaticFreq20 \ + --resolution 480,832 --guidance 0 --chunk_size 12 --start 0 --end 1000 --query_steps 0,9,18,27,34 +""" + +import argparse +import json +import os + +import mediapy +import numpy as np +import torch +import tqdm +from loguru import logger + +from cosmos_policy._src.predict2.action.datasets.gr00t_dreams.data.dataset import LeRobotDataset +from cosmos_policy._src.predict2.action.inference.inference_pipeline import ( + ActionVideo2WorldInference, +) + +_IMAGE_EXTENSIONS = [".png", ".jpg", ".jpeg", "webp"] +_VIDEO_EXTENSIONS = [".mp4"] + +_ACTION_SCALER = 20.0 + + +def parse_arguments() -> argparse.Namespace: + """Parses command-line arguments for the Video2World inference script.""" + parser = argparse.ArgumentParser(description="Image2World/Video2World inference script") + parser.add_argument("--experiment", type=str, required=True, help="Experiment config") + parser.add_argument("--chunk_size", type=int, default=12, help="Chunk size for action conditioning") + parser.add_argument("--guidance", type=int, default=7, help="Guidance value") + parser.add_argument("--seed", type=int, default=1, help="Guidance value") + parser.add_argument( + "--ckpt_path", + type=str, + default="", + help="Path to the checkpoint. If not provided, will use the one specify in the config", + ) + parser.add_argument("--s3_cred", type=str, default="credentials/s3_checkpoint.secret") + parser.add_argument( + "--resolution", + type=str, + default="none", + help="Resolution of the video (H,W). Be default it will use model trained resolution. 9:16", + ) + parser.add_argument("--input_video_root", type=str, default="bridge/annotation/test_100", help="Action root") + parser.add_argument("--save_root", type=str, default="results/image2world", help="Save root") + parser.add_argument( + "--base_path_gr00t_gr1_local", + type=str, + default="/project/cosmos/user/datasets/gr1_unified/gr1_unified.RU0226RemoveStaticFreq20", + help="Base path to the GR00T GR1 dataset", + ) + + parser.add_argument("--start", type=int, default=0, help="Start index for processing files") + parser.add_argument("--end", type=int, default=100, help="End index for processing files") + + parser.add_argument( + "--num_latent_conditional_frames", + type=int, + default=1, + help="Number of latent conditional frames (0, 1 or 2). For images, both values work by duplicating frames. For videos, uses the first N frames.", + ) + + parser.add_argument( + "--query_steps", + type=lambda x: [int(i) for i in x.split(",")], + default="0,9,18,27,34", + help="Query steps for the diffusion process", + ) + # Context parallel arguments + parser.add_argument( + "--context_parallel_size", + type=int, + default=1, + help="Context parallel size (number of GPUs to split context over). Set to 8 for 8 GPUs", + ) + return parser.parse_args() + + +def main(): + torch.enable_grad(False) # Disable gradient calculations for inference + args = parse_arguments() + + # Determine supported extensions based on num_latent_conditional_frames + if args.num_latent_conditional_frames > 1: + supported_extensions = _VIDEO_EXTENSIONS + # Check if input folder contains any videos + has_videos = False + for file_name in os.listdir(args.input_root): + file_ext = os.path.splitext(file_name)[1].lower() + if file_ext in _VIDEO_EXTENSIONS: + has_videos = True + break + + if not has_videos: + raise ValueError( + f"num_latent_conditional_frames={args.num_latent_conditional_frames} > 1 requires video inputs, " + f"but no videos found in {args.input_root}. Found extensions: " + f"{set(os.path.splitext(f)[1].lower() for f in os.listdir(args.input_root) if os.path.splitext(f)[1])}" + ) + + logger.info(f"Using video-only mode with {args.num_latent_conditional_frames} conditional frames") + elif args.num_latent_conditional_frames == 1: + supported_extensions = _IMAGE_EXTENSIONS + _VIDEO_EXTENSIONS + logger.info(f"Using image+video mode with {args.num_latent_conditional_frames} conditional frame") + + np.random.seed(0) + torch.manual_seed(0) + if torch.cuda.is_available(): + torch.cuda.manual_seed_all(0) + + dataset = LeRobotDataset( + num_frames=13, + time_division_factor=4, + time_division_remainder=1, + max_pixels=1920 * 1080, + data_file_keys=("video",), + image_file_extension=("jpg", "jpeg", "png", "webp"), + video_file_extension=("mp4", "avi", "mov", "wmv", "mkv", "flv", "webm"), + repeat=1, + args=None, + dataset_path=args.base_path_gr00t_gr1_local, + data_split="full", + embodiment="gr1", + downscaled_res=False, + ) + + # Initialize the inference handler with context parallel support + video2world_cli = ActionVideo2WorldInference( + args.experiment, args.ckpt_path, args.s3_cred, context_parallel_size=args.context_parallel_size + ) + + mem_bytes = torch.cuda.memory_allocated(device=torch.device("cuda" if torch.cuda.is_available() else "cpu")) + logger.info(f"GPU memory usage after model dcp.load: {mem_bytes / (1024**3):.2f} GB") + + os.makedirs(os.path.join(args.save_root, "latents"), exist_ok=True) + os.makedirs(os.path.join(args.save_root, "images"), exist_ok=True) + os.makedirs(os.path.join(args.save_root, "actions"), exist_ok=True) + os.makedirs(os.path.join(args.save_root, "videos"), exist_ok=True) + + print(len(dataset)) + for idx in tqdm.tqdm((range(args.start, args.end))): + data = dataset[idx] + img_np_array = data["video"][:, 0, :, :].permute(1, 2, 0).cpu().numpy() + video_np_array = data["video"].permute(1, 2, 3, 0).cpu().numpy() + action = data["action"].cpu().numpy() + + next_img_array, video_clamped, latents_to_save = video2world_cli.step_inference_with_latents( + img_array=img_np_array, + action=action, + guidance=args.guidance, + seed=0, + num_latent_conditional_frames=args.num_latent_conditional_frames, + query_steps=args.query_steps, + ) + + for k in latents_to_save: + latents_to_save[k] = latents_to_save[k].squeeze(0).cpu() + + torch.save(latents_to_save, os.path.join(args.save_root, "latents", f"{idx}.pt")) + mediapy.write_image(os.path.join(args.save_root, "images", f"{idx}.png"), img_np_array) + mediapy.write_video(os.path.join(args.save_root, "videos", f"{idx}.mp4"), video_np_array) + with open(os.path.join(args.save_root, "actions", f"{idx}.json"), "w") as f: + json.dump(action.tolist(), f, indent=4) + + exit() + + +if __name__ == "__main__": + main() diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference_pipeline.py b/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference_pipeline.py new file mode 100644 index 0000000000000000000000000000000000000000..07d7e5e4a0261b3901d4fcafb9fe3eb07b4363c9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/inference/inference_pipeline.py @@ -0,0 +1,360 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +import numpy as np +import torch +import torch.distributed as dist +import torchvision +from loguru import logger +from megatron.core import parallel_state + +from cosmos_policy._src.imaginaire.utils import distributed +from cosmos_policy._src.predict2.inference.get_t5_emb import get_text_embedding +from cosmos_policy._src.predict2.utils.model_loader import load_model_from_checkpoint + +_DEFAULT_NEGATIVE_PROMPT = "The video captures a series of frames showing ugly scenes, static with no motion, motion blur, over-saturation, shaky footage, low resolution, grainy texture, pixelated images, poorly lit areas, underexposed and overexposed scenes, poor color balance, washed out colors, choppy sequences, jerky movements, low frame rate, artifacting, color banding, unnatural transitions, outdated special effects, fake elements, unconvincing visuals, poorly edited content, jump cuts, visual noise, and flickering. Overall, the video is of poor quality." + + +class ActionVideo2WorldInference: + """ + Handles the Video2World inference process, including model loading, data preparation, + and video generation from an image/video and text prompt. Now supports context parallelism. + """ + + def __init__(self, experiment_name: str, ckpt_path: str, s3_credential_path: str, context_parallel_size: int = 1): + """ + Initializes the Video2WorldInference class. + + Loads the diffusion model and its configuration based on the provided + experiment name and checkpoint path. Sets up distributed processing if needed. + + Args: + experiment_name (str): Name of the experiment configuration. + ckpt_path (str): Path to the model checkpoint (local or S3). + s3_credential_path (str): Path to S3 credentials file (if loading from S3). + context_parallel_size (int): Number of GPUs for context parallelism. + """ + self.experiment_name = experiment_name + self.ckpt_path = ckpt_path + self.s3_credential_path = s3_credential_path + self.context_parallel_size = context_parallel_size + self.process_group = None + + # Initialize distributed processing if context parallel size > 1 + if self.context_parallel_size > 1: + self._init_distributed() + + # Load the model and config + model, config = load_model_from_checkpoint( + experiment_name=self.experiment_name, + s3_checkpoint_dir=self.ckpt_path, + config_file="cosmos_policy/_src/predict2/action/configs/action_conditioned/config.py", + load_ema_to_reg=True, + ) + + # Enable context parallel on the model if using context parallelism + if self.context_parallel_size > 1: + model.net.enable_context_parallel(self.process_group) + + self.model = model + self.config = config + self.batch_size = 1 + self.neg_t5_embeddings = None + + def _init_distributed(self): + """Initialize distributed processing for context parallelism.""" + + # Initialize distributed environment + distributed.init() + + # Initialize model parallel states + parallel_state.initialize_model_parallel( + context_parallel_size=self.context_parallel_size, + ) + + # Get the process group for context parallel + self.process_group = parallel_state.get_context_parallel_group() + + logger.info(f"Initialized context parallel with size {self.context_parallel_size}") + logger.info(f"Current rank: {distributed.get_rank()}, World size: {distributed.get_world_size()}") + + def _get_data_batch_input( + self, + video: torch.Tensor, + prompt: str, + num_conditional_frames: int = 1, + negative_prompt: str = _DEFAULT_NEGATIVE_PROMPT, + use_neg_prompt: bool = True, + ): + """ + Prepares the input data batch for the diffusion model. + + Constructs a dictionary containing the video tensor, text embeddings, + and other necessary metadata required by the model's forward pass. + Optionally includes negative text embeddings. + + Args: + video (torch.Tensor): The input video tensor (B, C, T, H, W). + prompt (str): The text prompt for conditioning. + num_conditional_frames (int): Number of conditional frames to use. + negative_prompt (str, optional): Custom negative prompt. + use_neg_prompt (bool, optional): Whether to include negative prompt embeddings. Defaults to True. + + Returns: + dict: A dictionary containing the prepared data batch, moved to the correct device and dtype. + """ + B, C, T, H, W = video.shape + + data_batch = { + "dataset_name": "video_data", + "video": video, + "fps": torch.randint(16, 32, (self.batch_size,)).float(), # Random FPS (might be used by model) + "padding_mask": torch.zeros(self.batch_size, 1, H, W), # Padding mask (assumed no padding here) + "num_conditional_frames": num_conditional_frames, # Specify number of conditional frames + } + + if use_neg_prompt: + assert negative_prompt is not None, "Negative prompt is required when use_neg_prompt is True" + + # Compute text embeddings + if self.model.text_encoder is not None: + data_batch["ai_caption"] = [prompt] + data_batch["t5_text_embeddings"] = self.model.text_encoder.compute_text_embeddings_online( + data_batch={"ai_caption": [prompt], "images": None}, + input_caption_key="ai_caption", + ) + if use_neg_prompt: + data_batch["neg_t5_text_embeddings"] = self.model.text_encoder.compute_text_embeddings_online( + data_batch={"ai_caption": [negative_prompt], "images": None}, + input_caption_key="ai_caption", + ) + else: + data_batch["t5_text_embeddings"] = get_text_embedding(prompt) + if use_neg_prompt: + data_batch["neg_t5_text_embeddings"] = get_text_embedding(negative_prompt) + + # Move tensors to GPU and convert to bfloat16 if they are floating point + for k, v in data_batch.items(): + if isinstance(v, torch.Tensor) and torch.is_floating_point(data_batch[k]): + data_batch[k] = v.cuda().to(dtype=torch.bfloat16) + + return data_batch + + def step_inference_with_latents( + self, + img_array: np.ndarray, + action: np.ndarray = None, + guidance: int = 3, + seed: int = 1, + num_latent_conditional_frames: int = 1, + query_steps: list[int] = None, + ): + """ + Runs a single inference step to generate the next video frame and the full video given an input image and action. + """ + + num_video_frames = action.shape[0] + 1 + + img_tensor = torchvision.transforms.functional.to_tensor(img_array).unsqueeze(0) + vid_input = torch.cat([img_tensor, torch.zeros_like(img_tensor).repeat(num_video_frames - 1, 1, 1, 1)], dim=0) + vid_input = (vid_input * 255.0).to(torch.uint8) # Convert to uint8 range if needed (might depend on model) + vid_input = vid_input.unsqueeze(0).permute(0, 2, 1, 3, 4) # Add batch dim B=1 and permute + + # Prepare the data batch with text embeddings + data_batch = self._get_data_batch_input( + vid_input, + prompt="", + num_conditional_frames=num_latent_conditional_frames, + negative_prompt="", + use_neg_prompt=False, + ) + + data_batch["action"] = torch.from_numpy(action).cuda().to(dtype=torch.bfloat16)[None, ...] + + mem_bytes = torch.cuda.memory_allocated(device=torch.device("cuda" if torch.cuda.is_available() else "cpu")) + logger.info(f"GPU memory usage after getting data_batch: {mem_bytes / (1024**3):.2f} GB") + + # Generate latent samples using the diffusion model + sample, latents_to_save = self.model.generate_samples_with_latents_from_batch( + data_batch, + n_sample=1, # Generate one sample + guidance=guidance, + seed=seed, # Fixed seed for reproducibility + is_negative_prompt=True, # Use classifier-free guidance + query_steps=query_steps, + ) + + # Decode the latent sample into a video tensor + video = self.model.decode(sample) + + video_normalized = (video - (-1)) / (1 - (-1)) + video_clamped = (torch.clamp(video_normalized[0], 0, 1) * 255).to(torch.uint8).permute(1, 2, 3, 0).cpu().numpy() + next_frame = torch.clamp(video_normalized[0, :, -1, :, :], 0, 1) + next_frame = (next_frame * 255).to(torch.uint8).permute(1, 2, 0).cpu().numpy() + return next_frame, video_clamped, latents_to_save + + def step_inference( + self, + img_array: np.ndarray, + action: np.ndarray = None, + guidance: int = 3, + seed: int = 1, + num_latent_conditional_frames: int = 1, + ): + """ + Runs a single inference step to generate the next video frame and the full video given an input image and action. + + Args: + img_array (np.ndarray): Input image as a numpy array (H, W, C), typically the first frame. + action (np.ndarray, optional): Action vector to condition the model. Should be shape (action_dim,) or (chunk_size, action_dim). + guidance (int, optional): Guidance scale for classifier-free guidance. Default is 3. + seed (int, optional): Random seed for reproducibility. Default is 1. + num_latent_conditional_frames (int, optional): Number of conditional frames to use for the model. Default is 1. + + Returns: + next_frame (np.ndarray): The next predicted frame as a numpy array (H, W, C), uint8. + video_clamped (np.ndarray): The generated video as a numpy array (T, H, W, C), uint8. + """ + num_video_frames = action.shape[0] + 1 + + img_tensor = torchvision.transforms.functional.to_tensor(img_array).unsqueeze(0) # (1, H, W, C) + vid_input = torch.cat([img_tensor, torch.zeros_like(img_tensor).repeat(num_video_frames - 1, 1, 1, 1)], dim=0) + vid_input = (vid_input * 255.0).to(torch.uint8) # Convert to uint8 range if needed (might depend on model) + vid_input = vid_input.unsqueeze(0).permute(0, 2, 1, 3, 4) # Add batch dim B=1 and permute + + # Prepare the data batch with text embeddings + data_batch = self._get_data_batch_input( + vid_input, + prompt="", + num_conditional_frames=num_latent_conditional_frames, + negative_prompt="", + use_neg_prompt=False, + ) + + data_batch["action"] = torch.from_numpy(action).cuda().to(dtype=torch.bfloat16)[None, ...] + + mem_bytes = torch.cuda.memory_allocated(device=torch.device("cuda" if torch.cuda.is_available() else "cpu")) + logger.info(f"GPU memory usage after getting data_batch: {mem_bytes / (1024**3):.2f} GB") + + # Generate latent samples using the diffusion model + # Video should be of shape torch.Size([1, 3, 93, 192, 320]) # Note: Shape check comment + sample = self.model.generate_samples_from_batch( + data_batch, + n_sample=1, # Generate one sample + guidance=guidance, + seed=seed, # Fixed seed for reproducibility + is_negative_prompt=True, # Use classifier-free guidance + ) + + # Decode the latent sample into a video tensor + video = self.model.decode(sample) + + video_normalized = (video - (-1)) / (1 - (-1)) + video_clamped = (torch.clamp(video_normalized[0], 0, 1) * 255).to(torch.uint8).permute(1, 2, 3, 0).cpu().numpy() + next_frame = torch.clamp(video_normalized[0, :, -1, :, :], 0, 1) + next_frame = (next_frame * 255).to(torch.uint8).permute(1, 2, 0).cpu().numpy() + return next_frame, video_clamped + + def step_inference_multi_frame( + self, + video_array: np.ndarray, + action: np.ndarray = None, + guidance: int = 3, + seed: int = 1, + num_latent_conditional_frames: int = 2, + ): + """ + Runs a single inference step to generate the next video frame and the full video given an input image and action. + + Args: + video_array (np.ndarray): Input video as a numpy array (T, H, W, C). + action (np.ndarray, optional): Action vector to condition the model. Should be shape (action_dim,) or (chunk_size, action_dim). + + guidance (int, optional): Guidance scale for classifier-free guidance. Default is 3. + seed (int, optional): Random seed for reproducibility. Default is 1. + num_latent_conditional_frames (int, optional): Number of conditional frames to use for the model. Default is 1. + + Returns: + next_frame (np.ndarray): The next predicted frame as a numpy array (H, W, C), uint8. + video_clamped (np.ndarray): The generated video as a numpy array (T, H, W, C), uint8. + """ + num_video_frames = action.shape[0] + 1 + (num_latent_conditional_frames - 1) * 4 + num_cond_image_frames = (num_latent_conditional_frames - 1) * 4 + 1 + + assert num_cond_image_frames == video_array.shape[0], ( + "Number of conditional frames is not equal to the number of frames in the video" + ) + assert action.shape[0] == num_video_frames - num_cond_image_frames, ( + "Number of action frames is not equal to the number of frames in the video" + ) + + video_tensor = torch.stack( + [torchvision.transforms.functional.to_tensor(v) for v in video_array] + ) # (T, C, H, W) + vid_input = torch.cat( + [ + video_tensor, + torch.zeros_like(video_tensor[0][None, ...]).repeat(num_video_frames - num_cond_image_frames, 1, 1, 1), + ], + dim=0, + ) + vid_input = (vid_input * 255.0).to(torch.uint8) # Convert to uint8 range if needed (might depend on model) + vid_input = vid_input.unsqueeze(0).permute(0, 2, 1, 3, 4) # Add batch dim B=1 and permute + + # Prepare the data batch with text embeddings + data_batch = self._get_data_batch_input( + vid_input, + prompt="", + num_conditional_frames=num_latent_conditional_frames, + negative_prompt="", + use_neg_prompt=False, + ) + + zero_action = np.zeros( + (4 * (num_latent_conditional_frames - 1), action.shape[1]) + ) # (4 * (num_latent_conditional_frames-1), action_dim) + action_padded = np.concatenate([zero_action, action], axis=0) + + data_batch["action"] = torch.from_numpy(action_padded).cuda().to(dtype=torch.bfloat16)[None, ...] + + mem_bytes = torch.cuda.memory_allocated(device=torch.device("cuda" if torch.cuda.is_available() else "cpu")) + logger.info(f"GPU memory usage after getting data_batch: {mem_bytes / (1024**3):.2f} GB") + + # Generate latent samples using the diffusion model + # Video should be of shape torch.Size([1, 3, 93, 192, 320]) # Note: Shape check comment + sample = self.model.generate_samples_from_batch( + data_batch, + n_sample=1, # Generate one sample + guidance=guidance, + seed=seed, # Fixed seed for reproducibility + is_negative_prompt=True, # Use classifier-free guidance + ) + + # Decode the latent sample into a video tensor + video = self.model.decode(sample) + + video_normalized = (video - (-1)) / (1 - (-1)) + video_clamped = (torch.clamp(video_normalized[0], 0, 1) * 255).to(torch.uint8).permute(1, 2, 3, 0).cpu().numpy() + next_frame = torch.clamp(video_normalized[0, :, -1, :, :], 0, 1) + next_frame = (next_frame * 255).to(torch.uint8).permute(1, 2, 0).cpu().numpy() + return next_frame, video_clamped + + def cleanup(self): + """Clean up distributed resources.""" + if self.context_parallel_size > 1: + if parallel_state.is_initialized(): + parallel_state.destroy_model_parallel() + dist.destroy_process_group() diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/models/action_conditioned_video2world_model.py b/REGEN-main/cosmos_policy/_src/predict2/action/models/action_conditioned_video2world_model.py new file mode 100644 index 0000000000000000000000000000000000000000..17d669529139dc2f9c874db9f5c229aa0a397aa2 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/models/action_conditioned_video2world_model.py @@ -0,0 +1,318 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math +from enum import Enum +from typing import Any, Callable, Dict, Tuple + +import attrs +import torch +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor + +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.configs.video2world.defaults.conditioner import Video2WorldCondition +from cosmos_policy._src.predict2.models.text2world_model import ( + DenoisePrediction, + Text2WorldCondition, + Text2WorldModelConfig, +) +from cosmos_policy._src.predict2.models.text2world_model import DiffusionModel as Text2WorldModel + +NUM_CONDITIONAL_FRAMES_KEY: str = "num_conditional_frames" + + +class ConditioningStrategy(str, Enum): + FRAME_REPLACE = "frame_replace" # First few frames of the video are replaced with the conditional frames + + def __str__(self) -> str: + return self.value + + +class HighSigmaStrategy(str, Enum): + NONE = "none" + UNIFORM80_2000 = "uniform80_2000" + LOGUNIFORM200_100000 = "LOGUNIFORM200_100000" + SHIFT24 = "shift24" + BALANCED_TWO_HEADS_V1 = "balanced_two_heads_v1" + HARDCODED_20steps = "hardcoded_20steps" + + def __str__(self) -> str: + return self.value + + +@attrs.define(slots=False) +class ActionConditionedVideo2WorldConfig(Text2WorldModelConfig): + min_num_conditional_frames: int = 1 # Minimum number of latent conditional frames + max_num_conditional_frames: int = 2 # Maximum number of latent conditional frames + sigma_conditional: float = 0.0001 # Noise level used for conditional frames + conditioning_strategy: str = str(ConditioningStrategy.FRAME_REPLACE) # What strategy to use for conditioning + denoise_replace_gt_frames: bool = True # Whether to denoise the ground truth frames + high_sigma_strategy: str = str(HighSigmaStrategy.UNIFORM80_2000) # What strategy to use for high sigma + high_sigma_ratio: float = 0.05 # Ratio of high sigma frames + low_sigma_ratio: float = 0.05 # Ratio of low sigma frames + action_dim: int = 10 * 8 + + def __attrs_post_init__(self): + super().__attrs_post_init__() + assert self.conditioning_strategy in [ + str(ConditioningStrategy.FRAME_REPLACE), + ] + assert self.high_sigma_strategy in [ + str(HighSigmaStrategy.NONE), + str(HighSigmaStrategy.UNIFORM80_2000), + str(HighSigmaStrategy.LOGUNIFORM200_100000), + str(HighSigmaStrategy.BALANCED_TWO_HEADS_V1), + str(HighSigmaStrategy.SHIFT24), + str(HighSigmaStrategy.HARDCODED_20steps), + ] + + +LOG_200 = math.log(200) +LOG_100000 = math.log(100000) + + +class ActionConditionedVideo2WorldModel(Text2WorldModel): + def get_data_and_condition( + self, data_batch: dict[str, torch.Tensor] + ) -> Tuple[Tensor, Tensor, Video2WorldCondition]: + # generate random number of conditional frames for training + raw_state, latent_state, condition = super().get_data_and_condition(data_batch) + condition = condition.set_video_condition( + gt_frames=latent_state.to(**self.tensor_kwargs), + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=data_batch.get(NUM_CONDITIONAL_FRAMES_KEY, None), + ) + return raw_state, latent_state, condition + + def draw_training_sigma_and_epsilon(self, x0_size: int, condition: Any) -> torch.Tensor: + sigma_B_1, epsilon = super().draw_training_sigma_and_epsilon(x0_size, condition) + is_video_batch = condition.data_type == DataType.VIDEO + # if is_video_batch, with 5% ratio, we regenerate sigma_B_1 with uniformally from 80 to 2000 + # with remaining 95% ratio, we keep the original sigma_B_1 + if is_video_batch: + if self.config.high_sigma_strategy == str(HighSigmaStrategy.UNIFORM80_2000): + mask = torch.rand(sigma_B_1.shape, device=sigma_B_1.device) < self.config.high_sigma_ratio + new_sigma = torch.rand(sigma_B_1.shape, device=sigma_B_1.device).type_as(sigma_B_1) * 1920 + 80 + sigma_B_1 = torch.where(mask, new_sigma, sigma_B_1) + elif self.config.high_sigma_strategy == str(HighSigmaStrategy.LOGUNIFORM200_100000): + mask = torch.rand(sigma_B_1.shape, device=sigma_B_1.device) < self.config.high_sigma_ratio + log_new_sigma = ( + torch.rand(sigma_B_1.shape, device=sigma_B_1.device).type_as(sigma_B_1) * (LOG_100000 - LOG_200) + + LOG_200 + ) + sigma_B_1 = torch.where(mask, log_new_sigma.exp(), sigma_B_1) + elif self.config.high_sigma_strategy == str(HighSigmaStrategy.SHIFT24): + # sample t from uniform distribution between 0 and 1, with same shape as sigma_B_1 + _t = torch.rand(sigma_B_1.shape, device=sigma_B_1.device).double() + _t = 24 * _t / (24 * _t + 1 - _t) + sigma_B_1 = (_t / (1.0 - _t)).float() + + mask = torch.rand(sigma_B_1.shape, device=sigma_B_1.device) < self.config.high_sigma_ratio + new_sigma = torch.rand(sigma_B_1.shape, device=sigma_B_1.device).type_as(sigma_B_1) * 1920 + 80 + sigma_B_1 = torch.where(mask, new_sigma, sigma_B_1) + elif self.config.high_sigma_strategy == str(HighSigmaStrategy.BALANCED_TWO_HEADS_V1): + # replace high sigma parts + mask = torch.rand(sigma_B_1.shape, device=sigma_B_1.device) < self.config.high_sigma_ratio + log_new_sigma = ( + torch.rand(sigma_B_1.shape, device=sigma_B_1.device).type_as(sigma_B_1) * (LOG_100000 - LOG_200) + + LOG_200 + ) + sigma_B_1 = torch.where(mask, log_new_sigma.exp(), sigma_B_1) + # replace low sigma parts + mask = torch.rand(sigma_B_1.shape, device=sigma_B_1.device) < self.config.low_sigma_ratio + low_sigma_B_1 = torch.rand(sigma_B_1.shape, device=sigma_B_1.device).type_as(sigma_B_1) * 2.0 + 0.00001 + sigma_B_1 = torch.where(mask, low_sigma_B_1, sigma_B_1) + elif self.config.high_sigma_strategy == str(HighSigmaStrategy.HARDCODED_20steps): + if not hasattr(self, "hardcoded_20steps_sigma"): + from cosmos_policy._src.imaginaire.modules.res_sampler import get_rev_ts + + hardcoded_20steps_sigma = get_rev_ts( + t_min=self.sde.sigma_min, t_max=self.sde.sigma_max, num_steps=20, ts_order=7.0 + ) + # add extra 100000 to the beginning + self.hardcoded_20steps_sigma = torch.cat( + [torch.tensor([100000.0], device=hardcoded_20steps_sigma.device), hardcoded_20steps_sigma], + dim=0, + ) + sigma_B_1 = self.hardcoded_20steps_sigma[ + torch.randint(0, len(self.hardcoded_20steps_sigma), sigma_B_1.shape) + ].type_as(sigma_B_1) + elif self.config.high_sigma_strategy == str(HighSigmaStrategy.NONE): + pass + else: + raise ValueError(f"High sigma strategy {self.config.high_sigma_strategy} is not supported") + return sigma_B_1, epsilon + + def denoise( + self, xt_B_C_T_H_W: torch.Tensor, sigma: torch.Tensor, condition: Text2WorldCondition + ) -> DenoisePrediction: + """ + Performs denoising on the input noise data, noise level, and condition + + Args: + xt (torch.Tensor): The input noise data. + sigma (torch.Tensor): The noise level. + condition (Text2WorldCondition): conditional information, generated from self.conditioner + + Returns: + DenoisePrediction: The denoised prediction, it includes clean data predicton (x0), \ + noise prediction (eps_pred). + """ + + if sigma.ndim == 1: + sigma_B_T = rearrange(sigma, "b -> b 1") + elif sigma.ndim == 2: + sigma_B_T = sigma + else: + raise ValueError(f"sigma shape {sigma.shape} is not supported") + + sigma_B_1_T_1_1 = rearrange(sigma_B_T, "b t -> b 1 t 1 1") + # get precondition for the network + c_skip_B_1_T_1_1, c_out_B_1_T_1_1, c_in_B_1_T_1_1, c_noise_B_1_T_1_1 = self.scaling(sigma=sigma_B_1_T_1_1) + + net_state_in_B_C_T_H_W = xt_B_C_T_H_W * c_in_B_1_T_1_1 + + if condition.is_video: + condition_state_in_B_C_T_H_W = condition.gt_frames.type_as(net_state_in_B_C_T_H_W) / self.config.sigma_data + if not condition.use_video_condition: + # When using random dropout, we zero out the ground truth frames + condition_state_in_B_C_T_H_W = condition_state_in_B_C_T_H_W * 0 + + _, C, _, _, _ = xt_B_C_T_H_W.shape + condition_video_mask = condition.condition_video_input_mask_B_C_T_H_W.repeat(1, C, 1, 1, 1).type_as( + net_state_in_B_C_T_H_W + ) + + # Replace the first few frames of the video with the conditional frames + # Update the c_noise as the conditional frames are clean and have very low noise + + # Make the first few frames of x_t be the ground truth frames + net_state_in_B_C_T_H_W = condition_state_in_B_C_T_H_W * condition_video_mask + net_state_in_B_C_T_H_W * ( + 1 - condition_video_mask + ) + # Adjust c_noise for the conditional frames + sigma_cond_B_1_T_1_1 = torch.ones_like(sigma_B_1_T_1_1) * self.config.sigma_conditional + _, _, _, c_noise_cond_B_1_T_1_1 = self.scaling(sigma=sigma_cond_B_1_T_1_1) + condition_video_mask_B_1_T_1_1 = condition_video_mask.mean(dim=[1, 3, 4], keepdim=True) + c_noise_B_1_T_1_1 = c_noise_cond_B_1_T_1_1 * condition_video_mask_B_1_T_1_1 + c_noise_B_1_T_1_1 * ( + 1 - condition_video_mask_B_1_T_1_1 + ) + + # forward pass through the network + net_output_B_C_T_H_W = self.net( + x_B_C_T_H_W=net_state_in_B_C_T_H_W.to( + **self.tensor_kwargs + ), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + timesteps_B_T=c_noise_B_1_T_1_1.squeeze(dim=[1, 3, 4]).to( + **{ + **self.tensor_kwargs, + "dtype": torch.float32 if self.config.use_wan_fp32_strategy else self.tensor_kwargs["dtype"], + }, + ), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + **condition.to_dict(), + ).float() + + x0_pred_B_C_T_H_W = c_skip_B_1_T_1_1 * xt_B_C_T_H_W + c_out_B_1_T_1_1 * net_output_B_C_T_H_W + if condition.is_video and self.config.denoise_replace_gt_frames: + # Set the first few frames to the ground truth frames. This will ensure that the loss is not computed for the first few frames. + x0_pred_B_C_T_H_W = condition.gt_frames.type_as( + x0_pred_B_C_T_H_W + ) * condition_video_mask + x0_pred_B_C_T_H_W * (1 - condition_video_mask) + + # get noise prediction based on sde + eps_pred_B_C_T_H_W = (xt_B_C_T_H_W - x0_pred_B_C_T_H_W) / sigma_B_1_T_1_1 + + return DenoisePrediction(x0_pred_B_C_T_H_W, eps_pred_B_C_T_H_W, None) + + def get_x0_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `x0_fn` based on the provided data batch and guidance factor. + + This function first processes the input data batch through a conditioning workflow (`conditioner`) to obtain conditioned and unconditioned states. It then defines a nested function `x0_fn` which applies a denoising operation on an input `noise_x` at a given noise level `sigma` using both the conditioned and unconditioned states. + + Args: + - data_batch (Dict): A batch of data used for conditioning. The format and content of this dictionary should align with the expectations of the `self.conditioner` + - guidance (float, optional): A scalar value that modulates the influence of the conditioned state relative to the unconditioned state in the output. Defaults to 1.5. + - is_negative_prompt (bool): use negative prompt t5 in uncondition if true + + Returns: + - Callable: A function `x0_fn(noise_x, sigma)` that takes two arguments, `noise_x` and `sigma`, and return x0 predictoin + + The returned function is suitable for use in scenarios where a denoised state is required based on both conditioned and unconditioned inputs, with an adjustable level of guidance influence. + """ + + if NUM_CONDITIONAL_FRAMES_KEY in data_batch: + num_conditional_frames = data_batch[NUM_CONDITIONAL_FRAMES_KEY] + else: + num_conditional_frames = 1 + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + is_image_batch = self.is_image_batch(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + _, x0, _ = self.get_data_and_condition(data_batch) + # override condition with inference mode; num_conditional_frames used Here! + condition = condition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + ) + uncondition = uncondition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + ) + condition = condition.edit_for_inference(is_cfg_conditional=True, num_conditional_frames=num_conditional_frames) + uncondition = uncondition.edit_for_inference( + is_cfg_conditional=False, num_conditional_frames=num_conditional_frames + ) + + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def x0_fn(noise_x: torch.Tensor, sigma: torch.Tensor) -> torch.Tensor: + cond_x0 = self.denoise(noise_x, sigma, condition).x0 + uncond_x0 = self.denoise(noise_x, sigma, uncondition).x0 + raw_x0 = cond_x0 + guidance * (cond_x0 - uncond_x0) + if "guided_image" in data_batch: + # replacement trick that enables inpainting with base model + assert "guided_mask" in data_batch, "guided_mask should be in data_batch if guided_image is present" + guide_image = data_batch["guided_image"] + guide_mask = data_batch["guided_mask"] + raw_x0 = guide_mask * guide_image + (1 - guide_mask) * raw_x0 + return raw_x0 + + return x0_fn diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/models/action_conditioned_video2world_rectified_flow_model.py b/REGEN-main/cosmos_policy/_src/predict2/action/models/action_conditioned_video2world_rectified_flow_model.py new file mode 100644 index 0000000000000000000000000000000000000000..016ee723ce47c5fb03a9895dcb86ab80fb4976f4 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/models/action_conditioned_video2world_rectified_flow_model.py @@ -0,0 +1,294 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from enum import Enum +from typing import Callable, Dict, Optional, Tuple + +import attrs +import torch +import tqdm +from megatron.core import parallel_state +from torch import Tensor + +from cosmos_policy._src.imaginaire.flags import INTERNAL +from cosmos_policy._src.imaginaire.utils import misc +from cosmos_policy._src.imaginaire.utils.context_parallel import broadcast_split_tensor, cat_outputs_cp +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.configs.video2world.defaults.conditioner import Video2WorldCondition +from cosmos_policy._src.predict2.models.text2world_model import DenoisePrediction +from cosmos_policy._src.predict2.models.text2world_model_rectified_flow import ( + Text2WorldCondition, + Text2WorldModelRectifiedFlow, + Text2WorldModelRectifiedFlowConfig, +) + +NUM_CONDITIONAL_FRAMES_KEY: str = "num_conditional_frames" + + +class ConditioningStrategy(str, Enum): + FRAME_REPLACE = "frame_replace" # First few frames of the video are replaced with the conditional frames + + def __str__(self) -> str: + return self.value + + +@attrs.define(slots=False) +class Video2WorldModelRectifiedFlowConfig(Text2WorldModelRectifiedFlowConfig): + min_num_conditional_frames: int = 1 # Minimum number of latent conditional frames + max_num_conditional_frames: int = 2 # Maximum number of latent conditional frames + conditional_frame_timestep: float = ( + -1.0 + ) # Noise level used for conditional frames; default is -1 which will not take effective + conditioning_strategy: str = str(ConditioningStrategy.FRAME_REPLACE) # What strategy to use for conditioning + denoise_replace_gt_frames: bool = True # Whether to denoise the ground truth frames + conditional_frames_probs: Optional[Dict[int, float]] = None # Probability distribution for conditional frames + + def __attrs_post_init__(self): + super().__attrs_post_init__() + assert self.conditioning_strategy in [ + str(ConditioningStrategy.FRAME_REPLACE), + ] + + +class ActionVideo2WorldModelRectifiedFlow(Text2WorldModelRectifiedFlow): + def get_data_and_condition( + self, data_batch: dict[str, torch.Tensor] + ) -> Tuple[Tensor, Tensor, Video2WorldCondition]: + # generate random number of conditional frames for training + raw_state, latent_state, condition = super().get_data_and_condition(data_batch) + condition = condition.set_video_condition( + gt_frames=latent_state.to(**self.tensor_kwargs), + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=data_batch.get(NUM_CONDITIONAL_FRAMES_KEY, None), + conditional_frames_probs=self.config.conditional_frames_probs, + ) + return raw_state, latent_state, condition + + @torch.no_grad() + def generate_samples_with_latents_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + is_negative_prompt: bool = False, + num_steps: int = 35, + shift: float = 5.0, + query_steps=[0, 9, 18, 27, 34], + **kwargs, + ) -> torch.Tensor: + """ + Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples. + Args: + data_batch (dict): raw data batch draw from the training data loader. + iteration (int): Current iteration number. + guidance (float): guidance weights + seed (int): random seed + state_shape (tuple): shape of the state, default to data batch if not provided + n_sample (int): number of samples to generate + is_negative_prompt (bool): use negative prompt t5 in uncondition if true + num_steps (int): number of steps for the diffusion process + """ + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + state_shape = [ + self.config.state_ch, + self.tokenizer.get_latent_num_frames(_T), + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + noise = misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + + seed_g = torch.Generator(device=self.tensor_kwargs["device"]) + seed_g.manual_seed(seed) + + self.sample_scheduler.set_timesteps( + num_steps, + device=self.tensor_kwargs["device"], + shift=shift, + use_kerras_sigma=self.config.use_kerras_sigma_at_inference, + ) + + timesteps = self.sample_scheduler.timesteps + + velocity_fn = self.get_velocity_fn_from_batch(data_batch, guidance, is_negative_prompt=is_negative_prompt) + if self.net.is_context_parallel_enabled: + noise = broadcast_split_tensor(tensor=noise, seq_dim=2, process_group=self.get_context_parallel_group()) + latents = noise + + latent_to_save = {} + if INTERNAL: + timesteps_iter = timesteps + else: + timesteps_iter = tqdm.tqdm(timesteps, desc="Generating samples", total=len(timesteps)) + + for num_step, t in enumerate(timesteps_iter): + if num_step in query_steps: + latent_to_save[num_step] = latents + print(f"Saving latent at step {num_step}, timestep {t}") + + latent_model_input = latents + timestep = [t] + + timestep = torch.stack(timestep) + + velocity_pred = velocity_fn(noise, latent_model_input, timestep.unsqueeze(0)) + temp_x0 = self.sample_scheduler.step( + velocity_pred.unsqueeze(0), t, latents[0].unsqueeze(0), return_dict=False, generator=seed_g + )[0] + latents = temp_x0.squeeze(0) + + latent_to_save[num_step] = latents + + if self.net.is_context_parallel_enabled: + latents = cat_outputs_cp(latents, seq_dim=2, cp_group=self.get_context_parallel_group()) + + return latents, latent_to_save + + def denoise( + self, + noise: torch.Tensor, + xt_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + condition: Text2WorldCondition, + ) -> DenoisePrediction: + """ + Args: + xt (torch.Tensor): The input noise data. + sigma (torch.Tensor): The noise level. + condition (Text2WorldCondition): conditional information, generated from self.conditioner + + Returns: + velocity prediction + """ + if condition.is_video: + condition_state_in_B_C_T_H_W = condition.gt_frames.type_as(xt_B_C_T_H_W) + if not condition.use_video_condition: + # When using random dropout, we zero out the ground truth frames + condition_state_in_B_C_T_H_W = condition_state_in_B_C_T_H_W * 0 + + _, C, _, _, _ = xt_B_C_T_H_W.shape + condition_video_mask = condition.condition_video_input_mask_B_C_T_H_W.repeat(1, C, 1, 1, 1).type_as( + xt_B_C_T_H_W + ) + + # Make the first few frames of x_t be the ground truth frames + xt_B_C_T_H_W = condition_state_in_B_C_T_H_W * condition_video_mask + xt_B_C_T_H_W * ( + 1 - condition_video_mask + ) + + # forward pass through the network + net_output_B_C_T_H_W = self.net( + x_B_C_T_H_W=xt_B_C_T_H_W.to(**self.tensor_kwargs), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + timesteps_B_T=timesteps_B_T, # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + **condition.to_dict(), + ).float() + + if condition.is_video and self.config.denoise_replace_gt_frames: + gt_frames_x0 = condition.gt_frames.type_as(net_output_B_C_T_H_W) + gt_frames_velocity = noise - gt_frames_x0 + net_output_B_C_T_H_W = gt_frames_velocity * condition_video_mask + net_output_B_C_T_H_W * ( + 1 - condition_video_mask + ) + + return net_output_B_C_T_H_W + + def get_velocity_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `x0_fn` based on the provided data batch and guidance factor. + + This function first processes the input data batch through a conditioning workflow (`conditioner`) to obtain conditioned and unconditioned states. It then defines a nested function `x0_fn` which applies a denoising operation on an input `noise_x` at a given noise level `sigma` using both the conditioned and unconditioned states. + + Args: + - data_batch (Dict): A batch of data used for conditioning. The format and content of this dictionary should align with the expectations of the `self.conditioner` + - guidance (float, optional): A scalar value that modulates the influence of the conditioned state relative to the unconditioned state in the output. Defaults to 1.5. + - is_negative_prompt (bool): use negative prompt t5 in uncondition if true + + Returns: + - Callable: A function `x0_fn(noise_x, sigma)` that takes two arguments, `noise_x` and `sigma`, and return velocity predictoin + + The returned function is suitable for use in scenarios where a denoised state is required based on both conditioned and unconditioned inputs, with an adjustable level of guidance influence. + """ + + if NUM_CONDITIONAL_FRAMES_KEY in data_batch: + num_conditional_frames = data_batch[NUM_CONDITIONAL_FRAMES_KEY] + else: + num_conditional_frames = 1 + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + is_image_batch = self.is_image_batch(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + _, x0, _ = self.get_data_and_condition(data_batch) + # override condition with inference mode; num_conditional_frames used Here! + condition = condition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + conditional_frames_probs=self.config.conditional_frames_probs, + ) + uncondition = uncondition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + conditional_frames_probs=self.config.conditional_frames_probs, + ) + condition = condition.edit_for_inference(is_cfg_conditional=True, num_conditional_frames=num_conditional_frames) + uncondition = uncondition.edit_for_inference( + is_cfg_conditional=False, num_conditional_frames=num_conditional_frames + ) + + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def velocity_fn(noise: torch.Tensor, noise_x: torch.Tensor, timestep: torch.Tensor) -> torch.Tensor: + cond_v = self.denoise(noise, noise_x, timestep, condition) + uncond_v = self.denoise(noise, noise_x, timestep, uncondition) + velocity_pred = cond_v + guidance * (cond_v - uncond_v) + return velocity_pred + + return velocity_fn diff --git a/REGEN-main/cosmos_policy/_src/predict2/action/networks/action_conditioned_minimal_v1_lvg_dit.py b/REGEN-main/cosmos_policy/_src/predict2/action/networks/action_conditioned_minimal_v1_lvg_dit.py new file mode 100644 index 0000000000000000000000000000000000000000..2323de979976b89b462a61f28a51ac10a8ecacb2 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/action/networks/action_conditioned_minimal_v1_lvg_dit.py @@ -0,0 +1,352 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import List, Optional, Tuple + +import torch +import torch.amp as amp +import torch.nn as nn +from einops import rearrange + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.networks.minimal_v4_dit import MiniTrainDIT + + +class Mlp(nn.Module): + def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.0): + super().__init__() + out_features = out_features or in_features + hidden_features = hidden_features or in_features + self.fc1 = nn.Linear(in_features, hidden_features) + self.activation = act_layer() + self.fc2 = nn.Linear(hidden_features, out_features) + self.drop = nn.Dropout(drop) + + def forward(self, x): + x = self.fc1(x) + x = self.activation(x) + x = self.drop(x) + x = self.fc2(x) + x = self.drop(x) + return x + + +class ActionConditionedMinimalV1LVGDiT(MiniTrainDIT): + def __init__(self, *args, timestep_scale: float = 1.0, **kwargs): + assert "in_channels" in kwargs, "in_channels must be provided" + kwargs["in_channels"] += 1 # Add 1 for the condition mask + + action_dim = kwargs.get("action_dim", 10 * 8) + if "action_dim" in kwargs: + del kwargs["action_dim"] + + num_action_per_chunk = kwargs.get("num_action_per_chunk", 12) + if "num_action_per_chunk" in kwargs: + del kwargs["num_action_per_chunk"] + + # NOTE: this is not used in the original code, but we need it for the rectified flow model + + self.timestep_scale = timestep_scale + log.info(f"timestep_scale: {timestep_scale}") + + super().__init__(*args, **kwargs) + + # add action embedding + self.action_embedder_B_D = Mlp( + in_features=action_dim * num_action_per_chunk, + hidden_features=self.model_channels * 4, + out_features=self.model_channels, + act_layer=lambda: nn.GELU(approximate="tanh"), + drop=0, + ) + self.action_embedder_B_3D = Mlp( + in_features=action_dim * num_action_per_chunk, + hidden_features=self.model_channels * 4, + out_features=self.model_channels * 3, + act_layer=lambda: nn.GELU(approximate="tanh"), + drop=0, + ) + + def forward( + self, + x_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + crossattn_emb: torch.Tensor, + condition_video_input_mask_B_C_T_H_W: Optional[torch.Tensor] = None, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + data_type: Optional[DataType] = DataType.VIDEO, + img_context_emb: Optional[torch.Tensor] = None, + action: Optional[torch.Tensor] = None, + intermediate_feature_ids: Optional[List[int]] = None, + **kwargs, + ) -> torch.Tensor | List[torch.Tensor] | Tuple[torch.Tensor, List[torch.Tensor]]: + del kwargs + + if data_type == DataType.VIDEO: + x_B_C_T_H_W = torch.cat([x_B_C_T_H_W, condition_video_input_mask_B_C_T_H_W.type_as(x_B_C_T_H_W)], dim=1) + else: + B, _, T, H, W = x_B_C_T_H_W.shape + x_B_C_T_H_W = torch.cat( + [x_B_C_T_H_W, torch.zeros((B, 1, T, H, W), dtype=x_B_C_T_H_W.dtype, device=x_B_C_T_H_W.device)], dim=1 + ) + + # NOTE: we need to scale the timesteps, which is added for rectified flow model + timesteps_B_T = timesteps_B_T * self.timestep_scale + + assert action is not None, "action must be provided" + action = rearrange(action, "b t d -> b 1 (t d)") + action_emb_B_D = self.action_embedder_B_D(action) + action_emb_B_3D = self.action_embedder_B_3D(action) + + assert isinstance(data_type, DataType), ( + f"Expected DataType, got {type(data_type)}. We need discuss this flag later." + ) + x_B_T_H_W_D, rope_emb_L_1_1_D, extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D = self.prepare_embedded_sequence( + x_B_C_T_H_W, + fps=fps, + padding_mask=padding_mask, + ) + + if self.use_crossattn_projection: + crossattn_emb = self.crossattn_proj(crossattn_emb) + + if img_context_emb is not None: + assert self.extra_image_context_dim is not None, ( + "extra_image_context_dim must be set if img_context_emb is provided" + ) + img_context_emb = self.img_context_proj(img_context_emb) + context_input = (crossattn_emb, img_context_emb) + else: + context_input = crossattn_emb + + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if timesteps_B_T.ndim == 1: + timesteps_B_T = timesteps_B_T.unsqueeze(1) + t_embedding_B_T_D, adaln_lora_B_T_3D = self.t_embedder(timesteps_B_T) + + # add action embedding to the timestep embedding and adaln_lora + t_embedding_B_T_D = t_embedding_B_T_D + action_emb_B_D + adaln_lora_B_T_3D = adaln_lora_B_T_3D + action_emb_B_3D + + t_embedding_B_T_D = self.t_embedding_norm(t_embedding_B_T_D) + + # for logging purpose + affline_scale_log_info = {} + affline_scale_log_info["t_embedding_B_T_D"] = t_embedding_B_T_D.detach() + self.affline_scale_log_info = affline_scale_log_info + self.affline_emb = t_embedding_B_T_D + self.crossattn_emb = crossattn_emb + + if extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D is not None: + assert x_B_T_H_W_D.shape == extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D.shape, ( + f"{x_B_T_H_W_D.shape} != {extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D.shape}" + ) + + B, T, H, W, D = x_B_T_H_W_D.shape + + intermediate_features_outputs = [] + for i, block in enumerate(self.blocks): + x_B_T_H_W_D = block( + x_B_T_H_W_D, + t_embedding_B_T_D, + context_input, + rope_emb_L_1_1_D=rope_emb_L_1_1_D, + adaln_lora_B_T_3D=adaln_lora_B_T_3D, + extra_per_block_pos_emb=extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D, + ) + if intermediate_feature_ids and i in intermediate_feature_ids: + x_reshaped_for_disc = rearrange(x_B_T_H_W_D, "b tp hp wp d -> b (tp hp wp) d") + intermediate_features_outputs.append(x_reshaped_for_disc) + + x_B_T_H_W_O = self.final_layer(x_B_T_H_W_D, t_embedding_B_T_D, adaln_lora_B_T_3D=adaln_lora_B_T_3D) + x_B_C_Tt_Hp_Wp = self.unpatchify(x_B_T_H_W_O) + if intermediate_feature_ids: + if len(intermediate_features_outputs) != len(intermediate_feature_ids): + log.warning( + f"Collected {len(intermediate_features_outputs)} intermediate features, " + f"but expected {len(intermediate_feature_ids)}. " + f"Requested IDs: {intermediate_feature_ids}" + ) + return x_B_C_Tt_Hp_Wp, intermediate_features_outputs + + return x_B_C_Tt_Hp_Wp + + +class ActionChunkConditionedMinimalV1LVGDiT(MiniTrainDIT): + def __init__(self, *args, timestep_scale: float = 1.0, **kwargs): + assert "in_channels" in kwargs, "in_channels must be provided" + kwargs["in_channels"] += 1 # Add 1 for the condition mask + + action_dim = kwargs.get("action_dim", 10 * 8) + if "action_dim" in kwargs: + del kwargs["action_dim"] + + self._num_action_per_latent_frame = kwargs.get("temporal_compression_ratio", 4) + if "temporal_compression_ratio" in kwargs: + del kwargs["temporal_compression_ratio"] + + if "num_action_per_chunk" in kwargs: + del kwargs["num_action_per_chunk"] + + self._hidden_dim_in_action_embedder = kwargs.get("hidden_dim_in_action_embedder", None) + if "hidden_dim_in_action_embedder" in kwargs: + del kwargs["hidden_dim_in_action_embedder"] + + # NOTE: this is not used in the original code, but we need it for the rectified flow model + self.timestep_scale = timestep_scale + + super().__init__(*args, **kwargs) + + if self._hidden_dim_in_action_embedder is None: + self._hidden_dim_in_action_embedder = self.model_channels * 4 + + log.info(f"hidden_dim_in_action_embedder: {self._hidden_dim_in_action_embedder}") + + # add action embedding + self.action_embedder_B_D = Mlp( + in_features=action_dim * self._num_action_per_latent_frame, + hidden_features=self._hidden_dim_in_action_embedder, + out_features=self.model_channels, + act_layer=lambda: nn.GELU(approximate="tanh"), + drop=0, + ) + self.action_embedder_B_3D = Mlp( + in_features=action_dim * self._num_action_per_latent_frame, + hidden_features=self._hidden_dim_in_action_embedder, + out_features=self.model_channels * 3, + act_layer=lambda: nn.GELU(approximate="tanh"), + drop=0, + ) + + def forward( + self, + x_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + crossattn_emb: torch.Tensor, + condition_video_input_mask_B_C_T_H_W: Optional[torch.Tensor] = None, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + data_type: Optional[DataType] = DataType.VIDEO, + img_context_emb: Optional[torch.Tensor] = None, + action: Optional[torch.Tensor] = None, + intermediate_feature_ids: Optional[List[int]] = None, + **kwargs, + ) -> torch.Tensor | List[torch.Tensor] | Tuple[torch.Tensor, List[torch.Tensor]]: + del kwargs + + if data_type == DataType.VIDEO: + x_B_C_T_H_W = torch.cat([x_B_C_T_H_W, condition_video_input_mask_B_C_T_H_W.type_as(x_B_C_T_H_W)], dim=1) + else: + B, _, T, H, W = x_B_C_T_H_W.shape + x_B_C_T_H_W = torch.cat( + [x_B_C_T_H_W, torch.zeros((B, 1, T, H, W), dtype=x_B_C_T_H_W.dtype, device=x_B_C_T_H_W.device)], dim=1 + ) + + timesteps_B_T = timesteps_B_T * self.timestep_scale + + # calculate action embedding + num_actions = action.shape[1] + assert action is not None, "action must be provided" + action = rearrange(action, "b t d -> b 1 (t d)") + action = rearrange(action, "b 1 (t d) -> b t d", t=num_actions // self._num_action_per_latent_frame) + action_emb_B_D = self.action_embedder_B_D(action) + action_emb_B_3D = self.action_embedder_B_3D(action) + + zero_pad_action_emb_B_D = torch.zeros_like(action_emb_B_D[:, :1, :], device=action_emb_B_D.device) + zero_pad_action_emb_B_3D = torch.zeros_like(action_emb_B_3D[:, :1, :], device=action_emb_B_3D.device) + + action_emb_B_D = torch.cat([zero_pad_action_emb_B_D, action_emb_B_D], dim=1) + action_emb_B_3D = torch.cat([zero_pad_action_emb_B_3D, action_emb_B_3D], dim=1) + + # NOTE: adjust the action embedding according to the number of frames + if condition_video_input_mask_B_C_T_H_W is not None and data_type == DataType.VIDEO: + condition_video_input_mask_B_T = (1 - condition_video_input_mask_B_C_T_H_W[:, 0, :, 0, 0]).unsqueeze(-1) + action_emb_B_D = action_emb_B_D * condition_video_input_mask_B_T + action_emb_B_3D = action_emb_B_3D * condition_video_input_mask_B_T + # ------------------------------------------------------------- + + assert isinstance(data_type, DataType), ( + f"Expected DataType, got {type(data_type)}. We need discuss this flag later." + ) + x_B_T_H_W_D, rope_emb_L_1_1_D, extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D = self.prepare_embedded_sequence( + x_B_C_T_H_W, + fps=fps, + padding_mask=padding_mask, + ) + + if self.use_crossattn_projection: + crossattn_emb = self.crossattn_proj(crossattn_emb) + + if img_context_emb is not None: + assert self.extra_image_context_dim is not None, ( + "extra_image_context_dim must be set if img_context_emb is provided" + ) + img_context_emb = self.img_context_proj(img_context_emb) + context_input = (crossattn_emb, img_context_emb) + else: + context_input = crossattn_emb + + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if timesteps_B_T.ndim == 1: + timesteps_B_T = timesteps_B_T.unsqueeze(1) + t_embedding_B_T_D, adaln_lora_B_T_3D = self.t_embedder(timesteps_B_T) + + t_embedding_B_T_D = t_embedding_B_T_D + action_emb_B_D + adaln_lora_B_T_3D = adaln_lora_B_T_3D + action_emb_B_3D + + t_embedding_B_T_D = self.t_embedding_norm(t_embedding_B_T_D) + + # for logging purpose + affline_scale_log_info = {} + affline_scale_log_info["t_embedding_B_T_D"] = t_embedding_B_T_D.detach() + self.affline_scale_log_info = affline_scale_log_info + self.affline_emb = t_embedding_B_T_D + self.crossattn_emb = crossattn_emb + + if extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D is not None: + assert x_B_T_H_W_D.shape == extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D.shape, ( + f"{x_B_T_H_W_D.shape} != {extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D.shape}" + ) + + B, T, H, W, D = x_B_T_H_W_D.shape + + intermediate_features_outputs = [] + for i, block in enumerate(self.blocks): + x_B_T_H_W_D = block( + x_B_T_H_W_D, + t_embedding_B_T_D, + context_input, + rope_emb_L_1_1_D=rope_emb_L_1_1_D, + adaln_lora_B_T_3D=adaln_lora_B_T_3D, + extra_per_block_pos_emb=extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D, + ) + if intermediate_feature_ids and i in intermediate_feature_ids: + x_reshaped_for_disc = rearrange(x_B_T_H_W_D, "b tp hp wp d -> b (tp hp wp) d") + intermediate_features_outputs.append(x_reshaped_for_disc) + + x_B_T_H_W_O = self.final_layer(x_B_T_H_W_D, t_embedding_B_T_D, adaln_lora_B_T_3D=adaln_lora_B_T_3D) + x_B_C_Tt_Hp_Wp = self.unpatchify(x_B_T_H_W_O) + if intermediate_feature_ids: + if len(intermediate_features_outputs) != len(intermediate_feature_ids): + log.warning( + f"Collected {len(intermediate_features_outputs)} intermediate features, " + f"but expected {len(intermediate_feature_ids)}. " + f"Requested IDs: {intermediate_feature_ids}" + ) + return x_B_C_Tt_Hp_Wp, intermediate_features_outputs + + return x_B_C_Tt_Hp_Wp diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/compile_tokenizer.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/compile_tokenizer.py new file mode 100644 index 0000000000000000000000000000000000000000..9277481b1ffe030a19e15561db80f4de0ed51bed --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/compile_tokenizer.py @@ -0,0 +1,61 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import torch + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.callback import Callback +from cosmos_policy._src.predict2.models.text2world_model import DiffusionModel + + +class CompileTokenizer(Callback): + def __init__(self, enabled: bool = False, compile_after_iterations: int = 4, dynamic: bool = False): + super().__init__() + self.enabled = enabled + self.compiled = False + self.compile_after_iterations = compile_after_iterations + self.skip_counter = 0 + self.dynamic = ( + dynamic # If there are issues with constant recompilations you may set this value to None or True + ) + + def on_training_step_start( + self, model: DiffusionModel, data_batch: dict[str, torch.Tensor], iteration: int = 0 + ) -> None: + if not self.enabled or self.compiled: + return + + if isinstance(model.tokenizer, torch.jit.ScriptModule): + log.critical( + f"The Tokenizer model {type(model.tokenizer)} is a JIT model, which is not compilable. The Tokenizer will not be compiled." + ) + + if self.skip_counter == self.compile_after_iterations: + try: + # PyTorch >= 2.7 + torch._dynamo.config.recompile_limit = 32 + except AttributeError: + try: + torch._dynamo.config.cache_size_limit = 32 + except AttributeError: + log.warning( + "Tokenizer compilation requested, but Torch Dynamo is unavailable – skipping compilation." + ) + self.enabled = False + return + + model.tokenizer.encode = torch.compile(model.tokenizer.encode, dynamic=self.dynamic) + self.compiled = True + self.skip_counter += 1 diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/dataloading_monitor.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/dataloading_monitor.py new file mode 100644 index 0000000000000000000000000000000000000000..57da46786c4c1b8b63a4add0483bd9e9bba42b1c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/dataloading_monitor.py @@ -0,0 +1,95 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import time + +import numpy as np +import torch +import wandb + +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import distributed +from cosmos_policy._src.imaginaire.utils.callback import Callback +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + + +class DetailedDataLoadingSpeedMonitor(Callback): + def __init__( + self, + every_n: int, + step_size: int = 1, + save_s3: bool = False, + ): + self.every_n = every_n + self.step_size = step_size + self.should_run = False + self.start_dataloading_time = None + self.dataloading_time = None + self.name = self.__class__.__name__ + self.save_s3 = save_s3 + self.time_delta_list = [] + + def on_before_dataloading(self, iteration: int = 0) -> None: + # We want to run it one iteration before on_training_step_start should_run is set to True. + global_step = iteration // self.step_size + self.should_run = (global_step + 1) % self.every_n == 0 + self.start_dataloading_time = time.time() + + def on_after_dataloading(self, iteration: int = 0) -> None: + self.time_delta_list.append(time.time() - self.start_dataloading_time) + + def on_training_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + if self.should_run: + self.should_run = False + cur_rank_mean, cur_rank_max = np.mean(self.time_delta_list), np.max(self.time_delta_list) + self.time_delta_list = [] # Reset the list + + dataloading_time_gather_list = distributed.all_gather_tensor( + torch.tensor([cur_rank_mean, cur_rank_max]).cuda() + ) + wandb_info = { + f"{self.name}_mean/dataloading_{k:03d}": v[0].item() for k, v in enumerate(dataloading_time_gather_list) + } + wandb_info.update( + { + f"{self.name}_max/dataloading_{k:03d}": v[1].item() + for k, v in enumerate(dataloading_time_gather_list) + } + ) + mean_times = torch.stack(dataloading_time_gather_list)[:, 0] + slowest_dataloading_rank_id = torch.argmax(mean_times) + max_dataloading = torch.max(mean_times) + wandb_info.update( + { + "slowest_rank/slowest_dataloading_rank": slowest_dataloading_rank_id.item(), + "slowest_rank/slowest_dataloading_time": max_dataloading.item(), + } + ) + + if wandb.run: + wandb.log(wandb_info, step=iteration) + + if self.save_s3 and distributed.is_rank0(): + easy_io.dump( + wandb_info, + f"s3://rundir/{self.name}/iter_{iteration:09d}.yaml", + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/device_monitor.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/device_monitor.py new file mode 100644 index 0000000000000000000000000000000000000000..51ab3ae0b63287d7e1177d752b8193cd8a652ce0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/device_monitor.py @@ -0,0 +1,194 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +from typing import Any, Dict, List, Tuple + +import pandas as pd +import psutil +import pynvml +import torch +import wandb + +from cosmos_policy._src.imaginaire.callbacks.every_n import EveryN +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + + +def log_prof_data( + data_list: List[Dict[str, Any]], + iteration: int, +) -> Tuple[pd.DataFrame]: + # Create a table to log data with rank information + columns = ["iteration", "rank"] + list(data_list[0].keys()) + data = [] + + # Initialize dictionaries to store min and max values for each metric + min_values = {key: float("inf") for key in columns[2:]} + max_values = {key: float("-inf") for key in columns[2:]} + sum_values = {key: 0.0 for key in columns[2:]} + + count = 0 + + for _rank, prof_data in enumerate(data_list): + row = [iteration, _rank] + [prof_data[key] for key in columns[2:]] + data.append(row) + count += 1 + + # Update min, max, and sum values + for key in columns[2:]: + min_values[key] = min(min_values[key], prof_data[key]) + max_values[key] = max(max_values[key], prof_data[key]) + sum_values[key] += prof_data[key] + + # Calculate average values + avg_values = {key: sum_values[key] / count for key in columns[2:]} + + df = pd.DataFrame(data, columns=columns) + summary_df = pd.DataFrame({"Avg": avg_values, "Max": max_values, "Min": min_values}) + + if wandb.run: + # Tables are stored as W&B artifacts. Logging one every few iterations + # eventually exhausts the run's file stream and stops all metric uploads. + # Keep the scalar summaries, which are sufficient for live monitoring. + summary = {} + for key in columns[2:]: + summary[f"DeviceMonitor/min_{key}"] = min_values[key] + summary[f"DeviceMonitor/max_{key}"] = max_values[key] + summary[f"DeviceMonitor/avg_{key}"] = avg_values[key] + + wandb.log(summary, step=iteration) + return df, summary_df + + +class DeviceMonitor(EveryN): + """ + A callback to monitor device (CPU/GPU) usage and log it at regular intervals. + + Args: + every_n (int, optional): The frequency at which the callback is invoked. Defaults to 200. + step_size (int, optional): The step size for the callback. Defaults to 1. + save_s3 (bool, optional): Whether to save the monitoring data to S3. Defaults to False. + """ + + def __init__( + self, + every_n: int = 200, + step_size: int = 1, + save_s3: bool = False, + upload_every_n_mul: int = 1, + log_memory_detail: bool = True, + ): + super().__init__(every_n=every_n, step_size=step_size) + self.name = self.__class__.__name__ + self.save_s3 = save_s3 + self.s3_save_fp = f"s3://rundir/{self.name}" + self.upload_every_n = upload_every_n_mul * every_n + + self.log_memory_detail = log_memory_detail + + def on_train_start(self, model, iteration=0): + torch.cuda.reset_peak_memory_stats() + self.world_size = distributed.get_world_size() + self.rank = distributed.get_rank() + config_job = self.config.job + self.local_dir = f"{config_job.path_local}/{self.name}" + if self.rank == 0: + os.makedirs(self.local_dir, exist_ok=True) + log.info(f"{self.name} callback: local_dir: {self.local_dir}") + + local_rank = int(os.getenv("LOCAL_RANK", 0)) + self.handle = pynvml.nvmlDeviceGetHandleByIndex(local_rank) + + def every_n_impl( + self, + trainer: ImaginaireTrainer, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int, + ) -> None: + cur_process = psutil.Process(os.getpid()) + cpu_memory_usage = sum(p.memory_info().rss for p in [cur_process] + cur_process.children(recursive=True)) + cpu_mem_gb = cpu_memory_usage / (1024**3) + + peak_gpu_mem_gb = torch.cuda.max_memory_allocated() / (1024**3) + peak_gpu_mem_reserved_gb = torch.cuda.max_memory_reserved() / (1024**3) + temp = torch.cuda.temperature() + try: + power = torch.cuda.power_draw() + except Exception as e: + log.warning(f"Failed to get power draw with error {e}") + power = 0 + util = torch.cuda.utilization() + clock = torch.cuda.clock_rate() + + memory_info = pynvml.nvmlDeviceGetMemoryInfo(self.handle) + nvml_used_gpu_mem_gb = memory_info.used / (1024**3) + nvml_free_gpu_mem_gb = memory_info.free / (1024**3) + + prof_data = { + "cpu_mem_gb": cpu_mem_gb, + "peak_gpu_mem_gb": peak_gpu_mem_gb, + "peak_gpu_mem_reserved_gb": peak_gpu_mem_reserved_gb, + "nvml_used_gpu_mem_gb": nvml_used_gpu_mem_gb, + "nvml_free_gpu_mem_gb": nvml_free_gpu_mem_gb, + "temp": temp, + "power": power, + "util": util, + "clock": clock, + } + + data_list = [prof_data] * self.world_size + # this is blocking by default + if self.world_size > 1: + torch.distributed.all_gather_object(data_list, prof_data) + torch.distributed.barrier() + + df, summary_df = log_prof_data(data_list, iteration) + if self.save_s3 and self.rank == 0: + global_step = iteration // self.step_size + should_run = global_step % self.upload_every_n == 0 + if should_run: + df.to_csv(os.path.join(self.local_dir, f"prof_data_{iteration:09d}.csv"), index=False) + summary_df.to_csv(os.path.join(self.local_dir, f"summary_{iteration:09d}.csv"), index=True) + easy_io.copyfile_from_local( + os.path.join(self.local_dir, f"prof_data_{iteration:09d}.csv"), + os.path.join(self.s3_save_fp, f"prof_data_{iteration:09d}.csv"), + ) + easy_io.copyfile_from_local( + os.path.join(self.local_dir, f"summary_{iteration:09d}.csv"), + os.path.join(self.s3_save_fp, f"summary_{iteration:09d}.csv"), + ) + if self.rank == 0: + log.info(f"{self.name} Stats:\n{summary_df.to_string()}") + if self.log_memory_detail: + memory_stats = torch.cuda.memory_stats() + if wandb.run: + wandb_memory_info = {f"mem/{key}": memory_stats[key] for key in memory_stats.keys()} + wandb.log(wandb_memory_info, step=iteration) + if self.save_s3: + global_step = iteration // self.step_size + should_run = global_step % self.upload_every_n == 0 + if should_run: + easy_io.dump( + memory_stats, + os.path.join(self.s3_save_fp, f"memory_stats_{iteration:09d}.yaml"), + ) + + torch.cuda.reset_peak_memory_stats() diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/every_n_draw_sample.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/every_n_draw_sample.py new file mode 100644 index 0000000000000000000000000000000000000000..1551a5fff308db1e1664223223882103f267de84 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/every_n_draw_sample.py @@ -0,0 +1,375 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math +import os +from contextlib import nullcontext +from functools import partial +from typing import List, Optional + +import numpy as np +import torch +import torch.distributed as dist +import torch.nn.functional as F +import torchvision +import torchvision.transforms.functional as torchvision_F +import wandb +from einops import rearrange, repeat +from megatron.core import parallel_state + +from cosmos_policy._src.imaginaire.callbacks.every_n import EveryN +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import distributed, log, misc +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.imaginaire.utils.parallel_state_helper import is_tp_cp_pp_rank0 +from cosmos_policy._src.imaginaire.visualize.video import save_img_or_video + + +def resize_image(image: torch.Tensor, size: int = 1024) -> torch.Tensor: + """ + Resize the image to the given size. This is done so that wandb can display the image correctly. + """ + _, h, w = image.shape + ratio = size / max(h, w) + new_h, new_w = int(ratio * h), int(ratio * w) + return torchvision_F.resize(image, (new_h, new_w)) + + +def is_primitive(value): + return isinstance(value, (int, float, str, bool, type(None))) + + +def convert_to_primitive(value): + if isinstance(value, (list, tuple)): + return [convert_to_primitive(v) for v in value if is_primitive(v) or isinstance(v, (list, dict))] + elif isinstance(value, dict): + return {k: convert_to_primitive(v) for k, v in value.items() if is_primitive(v) or isinstance(v, (list, dict))} + elif is_primitive(value): + return value + else: + return "non-primitive" # Skip non-primitive types + + +class EveryNDrawSample(EveryN): + """ + This callback sample condition inputs from training data, run inference and save the results to wandb and s3. + + Args: + every_n (int): The frequency at which the callback is invoked. + step_size (int, optional): The step size for the callback. Defaults to 1. + n_viz_sample (int, optional): for each batch, min(n_viz_sample, batch_size) samples will be saved to wandb. Defaults to 3. + n_sample_to_save (int, optional): number of samples to save. The actual number of samples to save is min(n_sample_to_save, data parallel instances). Defaults to 128. + num_sampling_step (int, optional): number of sampling steps. Defaults to 35. + guidance (List[float], optional): guidance scale. Defaults to [0.0, 3.0, 7.0]. + do_x0_prediction (bool, optional): whether to do x0 prediction. Defaults to True. + n_sigmas_for_x0_prediction (int, optional): number of sigmas to use for x0 prediction. Defaults to 4. + save_s3 (bool, optional): whether to save to s3. Defaults to False. + is_ema (bool, optional): whether the callback is run for ema model. Defaults to False. + use_negative_prompt (bool, optional): whether to use negative prompt. Defaults to False. + fps (int, optional): frames per second when saving the video. Defaults to 16. + """ + + def __init__( + self, + every_n: int, + step_size: int = 1, + n_viz_sample: int = 3, + n_sample_to_save: int = 128, + num_sampling_step: int = 35, + guidance: List[float] = [0.0, 3.0, 7.0], + do_x0_prediction: bool = True, + n_sigmas_for_x0_prediction: int = 4, + save_s3: bool = False, + is_ema: bool = False, + use_negative_prompt: bool = False, + prompt_type: str = "t5_xxl", + fps: int = 16, + run_at_start: bool = False, + ): + # s3: # files: min(n_sample_to_save, data instance) # per file: min(batch_size, n_viz_sample) + # wandb: 1 file, # per file: min(batch_size, n_viz_sample) + super().__init__(every_n, step_size, run_at_start=run_at_start) + + self.n_viz_sample = n_viz_sample + self.n_sample_to_save = n_sample_to_save + self.save_s3 = save_s3 + self.do_x0_prediction = do_x0_prediction + self.n_sigmas_for_x0_prediction = n_sigmas_for_x0_prediction + self.name = self.__class__.__name__ + self.is_ema = is_ema + self.use_negative_prompt = use_negative_prompt + self.prompt_type = prompt_type + self.guidance = guidance + self.num_sampling_step = num_sampling_step + self.rank = distributed.get_rank() + self.fps = fps + + def on_train_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + config_job = self.config.job + self.local_dir = f"{config_job.path_local}/{self.name}" + if distributed.get_rank() == 0: + os.makedirs(self.local_dir, exist_ok=True) + log.info(f"Callback: local_dir: {self.local_dir}") + + if parallel_state.is_initialized(): + self.data_parallel_id = parallel_state.get_data_parallel_rank() + else: + self.data_parallel_id = self.rank + + if self.use_negative_prompt: + if self.prompt_type == "t5_xxl": + self.negative_prompt_data = easy_io.load( + "s3://bucket/edify_video/v4/validation/item_dataset/negative_prompt/000000.pkl" + ) + elif self.prompt_type == "umt5_xxl": + self.negative_prompt_data = easy_io.load( + "s3://bucket/edify_video/v4/validation/item_dataset/negative_prompt/umt5_neg.pt" + ) + else: + raise ValueError(f"Invalid prompt type: {self.prompt_type}") + + @misc.timer("EveryNDrawSample: x0") + @torch.no_grad() + def x0_pred(self, trainer, model, data_batch, output_batch, loss, iteration): + tag = "ema" if self.is_ema else "reg" + + log.debug("starting data and condition model", rank0_only=False) + + raw_data, x0, condition = model.get_data_and_condition(data_batch) + _, condition, x0, _ = model.broadcast_split_for_model_parallelsim(None, condition, x0, None) + + log.debug("done data and condition model", rank0_only=False) + batch_size = x0.shape[0] + sigmas = np.exp( + np.linspace( + math.log(model.sde.sigma_min), math.log(model.sde.sigma_max), self.n_sigmas_for_x0_prediction + 1 + )[1:] + ) + + to_show = [] + generator = torch.Generator(device="cuda") + generator.manual_seed(0) + random_noise = torch.randn(*x0.shape, generator=generator, **model.tensor_kwargs) + _ones = torch.ones(batch_size, **model.tensor_kwargs) + mse_loss_list = [] + for _, sigma in enumerate(sigmas): + x_sigma = sigma * random_noise + x0 + log.debug(f"starting denoising {sigma}", rank0_only=False) + sample = model.denoise(x_sigma, _ones * sigma, condition).x0 + log.debug(f"done denoising {sigma}", rank0_only=False) + mse_loss = distributed.dist_reduce_tensor(F.mse_loss(sample, x0)) + mse_loss_list.append(mse_loss) + + if hasattr(model, "decode"): + sample = model.decode(sample) + to_show.append(sample.float().cpu()) + to_show.append( + raw_data.float().cpu(), + ) + + base_fp_wo_ext = f"{tag}_ReplicateID{self.data_parallel_id:04d}_x0_Iter{iteration:09d}" + + local_path = self.run_save(to_show, batch_size, base_fp_wo_ext) + return local_path, torch.tensor(mse_loss_list).cuda(), sigmas + + @torch.no_grad() + def every_n_impl(self, trainer, model, data_batch, output_batch, loss, iteration): + if self.is_ema: + if not model.config.ema.enabled: + return + context = partial(model.ema_scope, "every_n_sampling") + else: + context = nullcontext + + tag = "ema" if self.is_ema else "reg" + sample_counter = getattr(trainer, "sample_counter", iteration) + batch_info = { + "data": { + k: convert_to_primitive(v) + for k, v in data_batch.items() + if is_primitive(v) or isinstance(v, (list, dict)) + }, + "sample_counter": sample_counter, + "iteration": iteration, + } + if is_tp_cp_pp_rank0(): + if self.save_s3 and self.data_parallel_id < self.n_sample_to_save: + easy_io.dump( + batch_info, + f"s3://rundir/{self.name}/BatchInfo_ReplicateID{self.data_parallel_id:04d}_Iter{iteration:09d}.json", + ) + + log.debug("entering, every_n_impl", rank0_only=False) + with context(): + if self.do_x0_prediction: + log.debug("entering, x0_pred", rank0_only=False) + x0_img_fp, mse_loss, sigmas = self.x0_pred( + trainer, + model, + data_batch, + output_batch, + loss, + iteration, + ) + log.debug("done, x0_pred", rank0_only=False) + if self.save_s3 and self.rank == 0: + easy_io.dump( + { + "mse_loss": mse_loss.tolist(), + "sigmas": sigmas.tolist(), + "iteration": iteration, + }, + f"s3://rundir/{self.name}/{tag}_MSE_Iter{iteration:09d}.json", + ) + + log.debug("entering, sample", rank0_only=False) + sample_img_fp = self.sample( + trainer, + model, + data_batch, + output_batch, + loss, + iteration, + ) + log.debug("done, sample", rank0_only=False) + + log.debug("waiting for all ranks to finish", rank0_only=False) + dist.barrier() + if wandb.run: + sample_counter = getattr(trainer, "sample_counter", iteration) + data_type = "image" if model.is_image_batch(data_batch) else "video" + tag += f"_{data_type}" + info = { + "trainer/global_step": iteration, + "sample_counter": sample_counter, + } + if self.do_x0_prediction: + info[f"{self.name}/{tag}_x0"] = wandb.Image(x0_img_fp, caption=f"{sample_counter}") + # convert mse_loss to a dict + mse_loss = mse_loss.tolist() + info.update({f"x0_pred_mse_{tag}/Sigma{sigmas[i]:0.5f}": mse_loss[i] for i in range(len(mse_loss))}) + + info[f"{self.name}/{tag}_sample"] = wandb.Image(sample_img_fp, caption=f"{sample_counter}") + wandb.log( + info, + step=iteration, + ) + torch.cuda.empty_cache() + + @misc.timer("EveryNDrawSample: sample") + def sample(self, trainer, model, data_batch, output_batch, loss, iteration): + tag = "ema" if self.is_ema else "reg" + + # Obtain text embeddings online + text_encoder_config = getattr(model.config, "text_encoder_config", None) + if text_encoder_config is not None and text_encoder_config.compute_online: + text_embeddings = model.text_encoder.compute_text_embeddings_online(data_batch, model.input_caption_key) + data_batch["t5_text_embeddings"] = text_embeddings + data_batch["t5_text_mask"] = torch.ones(text_embeddings.shape[0], text_embeddings.shape[1], device="cuda") + + raw_data, x0, condition = model.get_data_and_condition(data_batch) + if self.use_negative_prompt: + batch_size = x0.shape[0] + if self.negative_prompt_data["t5_text_embeddings"].shape != data_batch["t5_text_embeddings"].shape: + data_batch["neg_t5_text_embeddings"] = misc.to( + repeat( + self.negative_prompt_data["t5_text_embeddings"], + "... -> b ...", + b=batch_size, + ), + **model.tensor_kwargs, + ) + else: + data_batch["neg_t5_text_embeddings"] = misc.to( + self.negative_prompt_data["t5_text_embeddings"], + **model.tensor_kwargs, + ) + + assert data_batch["neg_t5_text_embeddings"].shape == data_batch["t5_text_embeddings"].shape, ( + f"{data_batch['neg_t5_text_embeddings'].shape} != {data_batch['t5_text_embeddings'].shape}" + ) + data_batch["neg_t5_text_mask"] = data_batch["t5_text_mask"] + + to_show = [] + for guidance in self.guidance: + sample = model.generate_samples_from_batch( + data_batch, + guidance=guidance, + # make sure no mismatch and also works for cp + state_shape=x0.shape[1:], + n_sample=x0.shape[0], + num_steps=self.num_sampling_step, + is_negative_prompt=True if self.use_negative_prompt else False, + ) + if hasattr(model, "decode"): + sample = model.decode(sample) + to_show.append(sample.float().cpu()) + + to_show.append(raw_data.float().cpu()) + + base_fp_wo_ext = f"{tag}_ReplicateID{self.data_parallel_id:04d}_Sample_Iter{iteration:09d}" + + batch_size = x0.shape[0] + if is_tp_cp_pp_rank0(): + local_path = self.run_save(to_show, batch_size, base_fp_wo_ext) + return local_path + return None + + def run_save(self, to_show, batch_size, base_fp_wo_ext) -> Optional[str]: + to_show = (1.0 + torch.stack(to_show, dim=0).clamp(-1, 1)) / 2.0 # [n, b, c, t, h, w] + is_single_frame = to_show.shape[3] == 1 + n_viz_sample = min(self.n_viz_sample, batch_size) + + # ! we only save first n_sample_to_save video! + if self.save_s3 and self.data_parallel_id < self.n_sample_to_save: + save_img_or_video( + rearrange(to_show, "n b c t h w -> c t (n h) (b w)"), + f"s3://rundir/{self.name}/{base_fp_wo_ext}", + fps=self.fps, + ) + + file_base_fp = f"{base_fp_wo_ext}_resize.jpg" + local_path = f"{self.local_dir}/{file_base_fp}" + + if self.rank == 0 and wandb.run: + if is_single_frame: # image case + to_show = rearrange( + to_show[:, :n_viz_sample], + "n b c t h w -> t c (n h) (b w)", + ) + image_grid = torchvision.utils.make_grid(to_show, nrow=1, padding=0, normalize=False) + # resize so that wandb can handle it + torchvision.utils.save_image(resize_image(image_grid, 1024), local_path, nrow=1, scale_each=True) + else: + to_show = to_show[:, :n_viz_sample] # [n, b, c, 3, h, w] + + # resize 3 frames frames so that we can display them on wandb + _T = to_show.shape[3] + three_frames_list = [0, _T // 2, _T - 1] + to_show = to_show[:, :, :, three_frames_list] + log_image_size = 1024 + to_show = rearrange( + to_show, + "n b c t h w -> 1 c (n h) (b t w)", + ) + + # resize so that wandb can handle it + image_grid = torchvision.utils.make_grid(to_show, nrow=1, padding=0, normalize=False) + torchvision.utils.save_image( + resize_image(image_grid, log_image_size), local_path, nrow=1, scale_each=True + ) + + return local_path + return None diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/grad_clip.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/grad_clip.py new file mode 100644 index 0000000000000000000000000000000000000000..f2efda0c98fff1f355da94232199f59c7f9efe6e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/grad_clip.py @@ -0,0 +1,112 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from dataclasses import dataclass +from typing import List, Tuple + +import torch +import wandb + +from cosmos_policy._src.imaginaire.utils import distributed +from cosmos_policy._src.imaginaire.utils.callback import Callback +from cosmos_policy._src.predict2.models.text2world_model import DiffusionModel + + +@torch.jit.script +def _fused_nan_to_num(params: List[torch.Tensor]): + for param in params: + torch.nan_to_num(param, nan=0.0, posinf=0.0, neginf=0.0, out=param) + + +@dataclass +class _MagnitudeRecord: + state: float = 0 + iter_count: int = 0 + + def reset(self) -> None: + self.state = 0 + self.iter_count = 0 + + def update(self, cur_state: torch.Tensor) -> None: + self.state += cur_state + self.iter_count += 1 + + def get_stat(self) -> Tuple[float, float]: + if self.iter_count > 0: + avg_state = self.state / self.iter_count + avg_state = avg_state.item() + else: + avg_state = 0 + self.reset() + return avg_state + + +class GradClip(Callback): + """ + This callback is used to clip the gradient norm of the model. + It also logs the average gradient norm of the model to wandb. + """ + + def __init__(self, clip_norm=1.0, force_finite: bool = True): + self.clip_norm = clip_norm + self.force_finite = force_finite + + self.img_mag_log = _MagnitudeRecord() + self.video_mag_log = _MagnitudeRecord() + self._cur_state = None + + def on_training_step_start( + self, model: DiffusionModel, data_batch: dict[str, torch.Tensor], iteration: int = 0 + ) -> None: + if model.is_image_batch(data_batch): + self._cur_state = self.img_mag_log + else: + self._cur_state = self.video_mag_log + + def on_before_optimizer_step( + self, + model_ddp: distributed.DistributedDataParallel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int = 0, + ) -> None: + del optimizer, scheduler + if isinstance(model_ddp, distributed.DistributedDataParallel): + model = model_ddp.module + else: + model = model_ddp + params = [] + + if self.force_finite: + for param in model.parameters(): + if param.grad is not None: + params.append(param.grad) + _fused_nan_to_num(params) + + total_norm = model.clip_grad_norm_(self.clip_norm) + + self._cur_state.update(total_norm) + if iteration % self.config.trainer.logging_iter == 0: + avg_img_mag, avg_video_mag = self.img_mag_log.get_stat(), self.video_mag_log.get_stat() + if wandb.run: + wandb.log( + { + "clip_grad_norm/image": avg_img_mag, + "clip_grad_norm/video": avg_video_mag, + "iteration": iteration, + }, + step=iteration, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/heart_beat.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/heart_beat.py new file mode 100644 index 0000000000000000000000000000000000000000..00d80116e656ae15c4b9e4144f0d23943cfeecf5 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/heart_beat.py @@ -0,0 +1,106 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import time +from datetime import datetime + +import pytz +import torch + +from cosmos_policy._src.imaginaire.callbacks.every_n import EveryN +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer +from cosmos_policy._src.imaginaire.utils import distributed +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + + +class HeartBeat(EveryN): + """ + A callback that logs a heartbeat message at regular intervals to indicate that the training process is still running. + + Args: + every_n (int): The frequency at which the callback is invoked. + step_size (int, optional): The step size for the callback. Defaults to 1. + update_interval_in_minute (int, optional): The interval in minutes for logging the heartbeat. Defaults to 20 minutes. + save_s3 (bool, optional): Whether to save the heartbeat information to S3. Defaults to False. + """ + + def __init__(self, every_n: int, step_size: int = 1, update_interval_in_minute: int = 20, save_s3: bool = False): + super().__init__(every_n=every_n, step_size=step_size) + self.name = self.__class__.__name__ + self.update_interval_in_minute = update_interval_in_minute + self.save_s3 = save_s3 + self.pst = pytz.timezone("America/Los_Angeles") + self.is_hitted = False + + @distributed.rank0_only + def on_train_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + self.time = time.time() + if self.save_s3: + current_time_pst = datetime.now(self.pst).strftime("%Y_%m_%d-%H_%M_%S") + info = { + "iteration": iteration, + "time": current_time_pst, + } + easy_io.dump(info, f"s3://rundir/{self.name}_start.yaml") + easy_io.dump(info, f"s3://timestamps_rundir/{self.name}_start.yaml") + + def on_training_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + if not self.is_hitted: + self.is_hitted = True + if distributed.get_rank() == 0: + self.report(iteration) + super().on_training_step_end(model, data_batch, output_batch, loss, iteration) + + @distributed.rank0_only + def every_n_impl( + self, + trainer: ImaginaireTrainer, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int, + ) -> None: + if time.time() - self.time > 60 * self.update_interval_in_minute: + self.report(iteration) + + def report(self, iteration: int = 0): + self.time = time.time() + if self.save_s3: + current_time_pst = datetime.now(self.pst).strftime("%Y_%m_%d-%H_%M_%S") + info = { + "iteration": iteration, + "time": current_time_pst, + } + easy_io.dump(info, f"s3://rundir/{self.name}.yaml") + + @distributed.rank0_only + def on_train_end(self, model: ImaginaireModel, iteration: int = 0) -> None: + if self.save_s3: + current_time_pst = datetime.now(self.pst).strftime("%Y_%m_%d-%H_%M_%S") + info = { + "iteration": iteration, + "time": current_time_pst, + } + easy_io.dump(info, f"s3://rundir/{self.name}_end.yaml") + easy_io.dump(info, f"s3://timestamps_rundir/{self.name}_end.yaml") diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/iter_speed.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/iter_speed.py new file mode 100644 index 0000000000000000000000000000000000000000..a6f9a81d6f77eae6cd0054dc3734c6819c30c6a0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/iter_speed.py @@ -0,0 +1,106 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import time + +import torch +import wandb +from torch import Tensor + +from cosmos_policy._src.imaginaire.callbacks.every_n import EveryN +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.distributed import rank0_only +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + + +class IterSpeed(EveryN): + """ + Args: + hit_thres (int): Number of iterations to wait before logging. + save_s3 (bool): Whether to save to S3. + save_s3_every_log_n (int): Save to S3 every n log iterations, which means save_s3_every_log_n n * every_n global iterations. + """ + + def __init__(self, *args, hit_thres: int = 5, save_s3: bool = True, save_s3_every_log_n: int = 10, **kwargs): + super().__init__(*args, **kwargs) + self.time = None + self.hit_counter = 0 + self.hit_thres = hit_thres + self.save_s3 = save_s3 + self.save_s3_every_log_n = save_s3_every_log_n + self.name = self.__class__.__name__ + self.last_hit_time = time.time() + + def on_training_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + if self.hit_counter < self.hit_thres: + log.info( + f"Iteration {iteration}: " + f"Hit counter: {self.hit_counter + 1}/{self.hit_thres} | " + f"Loss: {loss.item():.4f} | " + f"Time: {time.time() - self.last_hit_time:.2f}s" + ) + self.hit_counter += 1 + self.last_hit_time = time.time() + #! useful for large scale training and avoid oom crash in the first two iterations!!! + torch.cuda.synchronize() + return + super().on_training_step_end(model, data_batch, output_batch, loss, iteration) + + @rank0_only + def every_n_impl( + self, + trainer: ImaginaireTrainer, + model: ImaginaireModel, + data_batch: dict[str, Tensor], + output_batch: dict[str, Tensor], + loss: Tensor, + iteration: int, + ) -> None: + if self.time is None: + self.time = time.time() + return + cur_time = time.time() + iter_speed = (cur_time - self.time) / self.every_n / self.step_size + + log.info(f"{iteration} : iter_speed {iter_speed:.2f} seconds per iteration | Loss: {loss.item():.4f}") + + if wandb.run: + sample_counter = getattr(trainer, "sample_counter", iteration) + wandb.log( + { + "timer/iter_speed": iter_speed, + "sample_counter": sample_counter, + }, + step=iteration, + ) + self.time = cur_time + if self.save_s3: + if iteration % (self.save_s3_every_log_n * self.every_n) == 0: + easy_io.dump( + { + "iter_speed": iter_speed, + "iteration": iteration, + }, + f"s3://rundir/{self.name}/iter_{iteration:09d}.yaml", + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/val_loss_computation.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/val_loss_computation.py new file mode 100644 index 0000000000000000000000000000000000000000..7dc89a2f421afc8ff894981a212fc02bf091c01b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/val_loss_computation.py @@ -0,0 +1,330 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +ValLossComputation Callback + +A validation loss computation callback that inherits from EveryNDrawSample +and reuses its x0_pred functionality with a different dataset (PromptVideoItemDataset). + +Key Features: +- Inherits from EveryNDrawSample: Reuses x0_pred method and infrastructure +- Real Video Evaluation: Uses PromptVideoItemDataset with actual video ground truth data +- Same Noise Schedule: Uses identical log-space noise schedule as EveryNDrawSample.x0_pred +- EMA Support: Inherited EMA functionality from parent class +- Distributed Training: Inherited distributed handling from parent class +- Comprehensive Logging: Logs validation loss metrics to console and WandB +""" + +from functools import partial + +import torch +import wandb +from megatron.core import parallel_state +from torch.utils.data import DataLoader + +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.predict2.callbacks.every_n_draw_sample import EveryNDrawSample +from cosmos_policy._src.predict2.datasets.data_sources.item_datasets_for_validation import get_itemdataset_option +from cosmos_policy._src.predict2.datasets.item_dataset import ( + ItemDatasetConfig, + PromptVideoItemDataset, + calculate_indices, +) +from cosmos_policy._src.predict2.models.text2world_model import DiffusionModel + + +class ValLossComputation(EveryNDrawSample): + def on_training_step_end(self, model, data_batch, output_batch, loss, iteration=0): + """Override to add debug logging before calling parent.""" + log.critical(f"DEBUG: ValLossComputation.on_training_step_end called at iteration {iteration}", rank0_only=True) + super().on_training_step_end(model, data_batch, output_batch, loss, iteration) + + """ + Validation loss computation callback inheriting from EveryNDrawSample. + + This callback reuses EveryNDrawSample.x0_pred functionality but evaluates on + PromptVideoItemDataset instead of training data. It computes validation loss + using the same noise schedule and methodology as the parent class. + + Key differences from parent: + - Uses PromptVideoItemDataset for validation data instead of training batches + - Only runs x0_pred (no sampling generation) + - Logs validation loss metrics instead of visual samples + """ + + def __init__( + self, + every_n: int, + step_size: int = 1, + val_dataset_name: str = "ptbench_video_val", + batch_size: int = 1, + n_sigmas_for_x0_prediction: int = 4, # Match EveryNDrawSample parameter + is_ema: bool = True, + is_debug: bool = False, + max_eval_samples: int = 100, + **kwargs, # Pass any additional EveryNDrawSample parameters + ): + """Initialize ValLossComputation by inheriting from EveryNDrawSample.""" + print(f"DEBUG: ValLossComputation.__init__ called with every_n={every_n}, max_eval_samples={max_eval_samples}") + log.critical(f"DEBUG: Initializing ValLossComputation with every_n={every_n}, is_ema={is_ema}", rank0_only=True) + # Initialize parent with x0_prediction enabled, sampling disabled + super().__init__( + every_n=every_n, + step_size=step_size, + n_viz_sample=batch_size, # Match batch_size for consistency + do_x0_prediction=True, # Always enable x0_prediction + n_sigmas_for_x0_prediction=n_sigmas_for_x0_prediction, + is_ema=is_ema, + save_s3=False, # Disable S3 saving for validation + guidance=[0.0], # Minimal guidance (not used since we skip sampling) + **kwargs, + ) + + # Validation-specific parameters + self.val_dataset_name = val_dataset_name + self.batch_size = batch_size + self.is_debug = is_debug + self.max_eval_samples = max_eval_samples + + # Override parent's name for logging + self.name = self.__class__.__name__ + + # Will be initialized in on_train_start + self.val_dataloader = None + + def on_train_start(self, model: DiffusionModel, iteration: int = 0) -> None: + """Initialize validation dataset and call parent's on_train_start.""" + log.critical(f"DEBUG: ValLossComputation.on_train_start called at iteration {iteration}", rank0_only=True) + + # Call parent's on_train_start to initialize base functionality + super().on_train_start(model, iteration) + + # Set up validation dataset + self._setup_validation_dataset(model) + + log.critical( + f"DEBUG: ValLossComputation.on_train_start completed, dataloader: {self.val_dataloader is not None}", + rank0_only=True, + ) + + def _setup_validation_dataset(self, model: DiffusionModel) -> None: + """Set up the validation dataset for evaluation.""" + log.critical(f"DEBUG: Starting _setup_validation_dataset for {self.val_dataset_name}", rank0_only=True) + + # Get video dimensions from model configuration + video_height, video_width = model.get_video_height_width() + num_video_frames = model.tokenizer.get_pixel_num_frames(model.get_num_video_latent_frames()) + log.critical( + f"DEBUG: Video dimensions: {video_height}x{video_width}, frames: {num_video_frames}", rank0_only=True + ) + + # Get dataset configuration - fail fast if not found + dataset_option: ItemDatasetConfig = get_itemdataset_option(self.val_dataset_name) + dataset_path = dataset_option.path + dataset_length = dataset_option.length + log.critical(f"DEBUG: Found dataset option: path={dataset_path}, length={dataset_length}", rank0_only=True) + + log.warning( + f"Using validation dataset: {self.val_dataset_name} at path: {dataset_path}. " + f"It is user's responsibility to set up the correct credentials." + ) + + # Limit evaluation samples and ensure FSDP compatibility + dataset_length = min(dataset_length, self.max_eval_samples) + dataset_length = int(dataset_length // model.config.fsdp_shard_size * model.config.fsdp_shard_size) + log.critical(f"DEBUG: Final dataset_length after FSDP alignment: {dataset_length}", rank0_only=True) + + # Distribute dataset across ranks + if torch.distributed.get_world_size() > parallel_state.get_data_parallel_world_size(): + num_replicate = parallel_state.get_data_parallel_world_size() + data_parallel_id = parallel_state.get_data_parallel_rank() + start_idx, end_idx, is_overflow = calculate_indices(dataset_length, num_replicate, data_parallel_id) + log.critical( + f"DEBUG: Using data parallel: replicate={num_replicate}, id={data_parallel_id}", rank0_only=True + ) + else: + world_size, rank = distributed.get_world_size(), distributed.get_rank() + start_idx, end_idx, is_overflow = calculate_indices(dataset_length, world_size, rank) + log.critical(f"DEBUG: Using world distributed: world_size={world_size}, rank={rank}", rank0_only=True) + + log.critical(f"DEBUG: Calculated indices: {start_idx}-{end_idx}, overflow={is_overflow}", rank0_only=True) + + # Debug mode: only evaluate 2 samples + if self.is_debug: + end_idx = min(start_idx + 2, end_idx) + + if is_overflow: + log.critical("DEBUG: Overflow in calculating indices, SKIPPING.", rank0_only=True) + self.val_dataloader = None + else: + log.critical( + f"DEBUG: Creating PromptVideoItemDataset with indices {start_idx}-{end_idx}...", rank0_only=True + ) + # Create validation dataloader + self.val_dataloader = DataLoader( + PromptVideoItemDataset( + path=dataset_path, + start_index=start_idx, + end_index=end_idx, + height=video_height, + width=video_width, + num_video_frames=num_video_frames, + ), + batch_size=self.batch_size, + num_workers=4, + prefetch_factor=2, + persistent_workers=False, + shuffle=False, + ) + log.critical(f"DEBUG: Successfully created dataloader: {self.val_dataloader is not None}", rank0_only=True) + + log.critical( + f"ValLoss: Finished setting up validation dataloader for {self.val_dataset_name} " + f"with video shape {num_video_frames}x{video_height}x{video_width}", + rank0_only=True, + ) + + def every_n_impl(self, trainer, model, data_batch, output_batch, loss, iteration): + """ + Compute validation loss by reusing parent's x0_pred on validation dataset. + + This method iterates through the validation dataset and applies the parent's + x0_pred method to each batch, then aggregates and logs the validation metrics. + """ + log.critical(f"DEBUG: ValLossComputation.every_n_impl called at iteration {iteration}", rank0_only=True) + del data_batch, output_batch, loss # Not used, we use validation data + + if self.val_dataloader is None: + log.critical( + f"DEBUG: ValLoss: Skipping validation at iteration {iteration} (no dataloader)", rank0_only=True + ) + return + + tag = "ema" if self.is_ema else "reg" + log.critical(f"ValLoss: {tag} Starting validation loss computation at iteration {iteration}", rank0_only=True) + + # Use parent's EMA context handling + if self.is_ema: + if not model.config.ema.enabled: + return + context = partial(model.ema_scope, "val_loss") + else: + from contextlib import nullcontext + + context = nullcontext + + total_mse_losses = [] + num_samples = 0 + + with context(): + # Iterate through validation dataset + for i, val_data_batch in enumerate(self.val_dataloader): + log.debug(f"ValLoss: {tag} Processing validation batch {i}") + + # Prepare validation batch (same as training batch format) + val_data_batch = self._prepare_validation_batch(val_data_batch, model) + + # Create dummy output_batch (required by parent's x0_pred) + dummy_output_batch = {"x0": val_data_batch["video"]} + + # Reuse parent's x0_pred method - this does the heavy lifting! + _, mse_loss, sigmas = self.x0_pred(trainer, model, val_data_batch, dummy_output_batch, None, iteration) + + total_mse_losses.append(mse_loss) + num_samples += val_data_batch["video"].shape[0] + + # Aggregate validation metrics + if total_mse_losses: + # Stack and average MSE losses across all validation batches + all_mse = torch.stack(total_mse_losses) # [num_batches, num_noise_levels] + avg_mse_per_noise = all_mse.mean(dim=0) # [num_noise_levels] + overall_val_loss = avg_mse_per_noise.mean().item() # Single scalar + + # Distributed aggregation + if torch.distributed.is_initialized(): + # Reduce using the same pattern as parent callback + samples_tensor = torch.tensor(num_samples, device="cuda", dtype=torch.float32) + total_samples_tensor = distributed.dist_reduce_tensor(samples_tensor, reduce="sum") + total_samples = int(total_samples_tensor.item()) + + # Reduce validation loss (mean across ranks) + loss_tensor = torch.tensor(overall_val_loss, device="cuda", dtype=torch.float32) + overall_val_loss_tensor = distributed.dist_reduce_tensor(loss_tensor, reduce="mean") + overall_val_loss = overall_val_loss_tensor.item() + + # Also reduce per-noise-level losses for WandB logging + avg_mse_per_noise = distributed.dist_reduce_tensor(avg_mse_per_noise, reduce="mean") + else: + total_samples = num_samples + + # Log validation results + log.critical( + f"ValLoss: {tag} Iteration {iteration} - Validation Loss: {overall_val_loss:.6f}, " + f"Samples: {total_samples}", + rank0_only=False, + ) + + # WandB logging (rank 0 only) + if wandb.run and distributed.get_rank() == 0: + wandb.log( + { + f"val_loss/{tag}": overall_val_loss, + f"val_samples/{tag}": total_samples, + }, + step=iteration, + ) + + # Log per-noise-level MSE (same format as parent class) + for i, sigma_val in enumerate(sigmas): + wandb.log({f"val_mse_{tag}/Sigma{sigma_val:0.5f}": avg_mse_per_noise[i].item()}, step=iteration) + + log.critical(f"ValLoss: {tag} Completed validation loss computation at iteration {iteration}", rank0_only=False) + distributed.barrier() + torch.cuda.empty_cache() + + def _prepare_validation_batch(self, val_data_batch: dict, model: DiffusionModel) -> dict: + """ + Prepare validation batch to match training data format expected by parent's x0_pred. + + This method converts PromptVideoItemDataset format to the format expected by + the parent class's x0_pred method, ensuring compatibility. + """ + # Move to correct device, but DON'T convert video dtype (keep uint8) + for key, value in val_data_batch.items(): + if isinstance(value, torch.Tensor): + if key == "video": + # Keep video as uint8, only move to device + val_data_batch[key] = value.to(device=model.tensor_kwargs["device"]) + else: + # Convert other tensors to model precision + val_data_batch[key] = value.to(**model.tensor_kwargs) + + # Video stays in uint8 format - model.get_data_and_condition() expects this + # The model's _normalize_video_databatch_inplace() will handle uint8 -> float conversion + + # Handle text embeddings (same as parent's sample method) + if model.config.text_encoder_config is not None and model.config.text_encoder_config.compute_online: + if "prompt" in val_data_batch: + val_data_batch["ai_caption"] = val_data_batch["prompt"] # Use prompt as ai_caption + text_embeddings = model.text_encoder.compute_text_embeddings_online( + val_data_batch, model.input_caption_key + ) + val_data_batch["t5_text_embeddings"] = text_embeddings + val_data_batch["t5_text_mask"] = torch.ones( + text_embeddings.shape[0], text_embeddings.shape[1], device="cuda" + ) + + return val_data_batch diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/validation_draw_sample.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/validation_draw_sample.py new file mode 100644 index 0000000000000000000000000000000000000000..e6b27ad7924fb6950daccc2fce97b61f3ac4f7e6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/validation_draw_sample.py @@ -0,0 +1,407 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math +import os +from typing import List, Optional + +import numpy as np +import torch +import torch.distributed as dist +import torch.nn.functional as F +import torchvision +import torchvision.transforms.functional as torchvision_F +import wandb +from einops import rearrange, repeat +from megatron.core import parallel_state + +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import distributed, log, misc +from cosmos_policy._src.imaginaire.utils.callback import Callback +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.imaginaire.utils.parallel_state_helper import is_tp_cp_pp_rank0 +from cosmos_policy._src.imaginaire.visualize.video import save_img_or_video + + +def resize_image(image: torch.Tensor, size: int = 1024) -> torch.Tensor: + """ + Resize the image to the given size. This is done so that wandb can display the image correctly. + """ + _, h, w = image.shape + ratio = size / max(h, w) + new_h, new_w = int(ratio * h), int(ratio * w) + return torchvision_F.resize(image, (new_h, new_w)) + + +def is_primitive(value): + return isinstance(value, (int, float, str, bool, type(None))) + + +def convert_to_primitive(value): + if isinstance(value, (list, tuple)): + return [convert_to_primitive(v) for v in value if is_primitive(v) or isinstance(v, (list, dict))] + elif isinstance(value, dict): + return {k: convert_to_primitive(v) for k, v in value.items() if is_primitive(v) or isinstance(v, (list, dict))} + elif is_primitive(value): + return value + else: + return "non-primitive" # Skip non-primitive types + + +class ValidationDrawSample(Callback): + """ + This callback sample condition inputs from validation data, run inference and save the results to wandb and s3. + + Args: + n_samples (int): The number of samples to run inference on. + n_viz_sample (int, optional): for each batch, min(n_viz_sample, batch_size) samples will be saved to wandb. Defaults to 3. + n_sample_to_save (int, optional): number of samples to save. The actual number of samples to save is min(n_sample_to_save, data parallel instances). Defaults to 128. + num_sampling_step (int, optional): number of sampling steps. Defaults to 35. + guidance (List[float], optional): guidance scale. Defaults to [0.0, 3.0, 7.0]. + do_x0_prediction (bool, optional): whether to do x0 prediction. Defaults to True. + n_sigmas_for_x0_prediction (int, optional): number of sigmas to use for x0 prediction. Defaults to 4. + save_s3 (bool, optional): whether to save to s3. Defaults to False. + is_ema (bool, optional): whether the callback is run for ema model. Defaults to False. + use_negative_prompt (bool, optional): whether to use negative prompt. Defaults to False. + fps (int, optional): frames per second when saving the video. Defaults to 16. + """ + + def __init__( + self, + n_samples: int, + n_viz_sample: int = 3, + n_sample_to_save: int = 128, + num_sampling_step: int = 35, + guidance: List[float] = [0.0, 3.0, 7.0], + do_x0_prediction: bool = True, + n_sigmas_for_x0_prediction: int = 4, + save_s3: bool = False, + is_ema: bool = False, + use_negative_prompt: bool = False, + prompt_type: str = "t5_xxl", + fps: int = 16, + run_at_start: bool = False, + barrier_after_run: bool = True, + ): + # s3: # files: min(n_sample_to_save, data instance) # per file: min(batch_size, n_viz_sample) + # wandb: 1 file, # per file: min(batch_size, n_viz_sample) + self.barrier_after_run = barrier_after_run + self.run_at_start = run_at_start + + self.n_viz_sample = n_viz_sample + self.n_sample_to_save = n_sample_to_save + self.save_s3 = save_s3 + self.do_x0_prediction = do_x0_prediction + self.n_sigmas_for_x0_prediction = n_sigmas_for_x0_prediction + self.name = self.__class__.__name__ + self.is_ema = is_ema + self.use_negative_prompt = use_negative_prompt + self.prompt_type = prompt_type + self.guidance = guidance + self.num_sampling_step = num_sampling_step + self.rank = distributed.get_rank() + self.fps = fps + self.n_samples = n_samples + self.sample_counter = 0 + + def on_train_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + config_job = self.config.job + self.local_dir = f"{config_job.path_local}/{self.name}" + self.wandb_online = self.config.job.wandb_mode == "online" + if distributed.get_rank() == 0: + os.makedirs(self.local_dir, exist_ok=True) + log.info(f"Callback: local_dir: {self.local_dir}") + + if parallel_state.is_initialized(): + self.data_parallel_id = parallel_state.get_data_parallel_rank() + else: + self.data_parallel_id = self.rank + + if self.use_negative_prompt: + if self.prompt_type == "t5_xxl": + self.negative_prompt_data = easy_io.load( + "s3://bucket/edify_video/v4/validation/item_dataset/negative_prompt/000000.pkl" + ) + elif self.prompt_type == "umt5_xxl": + self.negative_prompt_data = easy_io.load( + "s3://bucket/edify_video/v4/validation/item_dataset/negative_prompt/umt5_neg.pt" + ) + else: + raise ValueError(f"Invalid prompt type: {self.prompt_type}") + + @misc.timer("ValidationDrawSample: x0") + @torch.no_grad() + def x0_pred(self, model, data_batch, iteration): + tag = "ema" if self.is_ema else "reg" + + log.debug("starting data and condition model", rank0_only=False) + + raw_data, x0, condition = model.get_data_and_condition(data_batch) + _, condition, x0, _ = model.broadcast_split_for_model_parallelsim(None, condition, x0, None) + + log.debug("done data and condition model", rank0_only=False) + batch_size = x0.shape[0] + sigmas = np.exp( + np.linspace( + math.log(model.sde.sigma_min), math.log(model.sde.sigma_max), self.n_sigmas_for_x0_prediction + 1 + )[1:] + ) + + to_show = [] + generator = torch.Generator(device="cuda") + generator.manual_seed(0) + random_noise = torch.randn(*x0.shape, generator=generator, **model.tensor_kwargs) + _ones = torch.ones(batch_size, **model.tensor_kwargs) + mse_loss_list = [] + for _, sigma in enumerate(sigmas): + x_sigma = sigma * random_noise + x0 + log.debug(f"starting denoising {sigma}", rank0_only=False) + sample = model.denoise(x_sigma, _ones * sigma, condition).x0 + log.debug(f"done denoising {sigma}", rank0_only=False) + mse_loss = distributed.dist_reduce_tensor(F.mse_loss(sample, x0)) + mse_loss_list.append(mse_loss) + + if hasattr(model, "decode"): + sample = model.decode(sample) + to_show.append(sample.float().cpu()) + to_show.append( + raw_data.float().cpu(), + ) + + local_path = self.run_save(to_show, batch_size, iteration, "x0") + return local_path, torch.tensor(mse_loss_list).cuda(), sigmas + + def on_validation_start( + self, model: ImaginaireModel, dataloader_val: torch.utils.data.DataLoader, iteration: int = 0 + ) -> None: + self.sample_counter = 0 + + def on_validation_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + trainer = self.trainer + + past_samples = self.sample_counter * distributed.get_world_size() + current_samples = past_samples + self.rank + + should_run = (iteration > 0 or self.run_at_start) and past_samples < self.n_samples + should_save = current_samples < self.n_samples + + if should_run: + save_log = "" if should_save else " (skipping save/log, running only for GPU synchronization)" + log.debug( + f"Callback {self.__class__.__name__} fired on validation_step_end step {iteration} [{current_samples + 1}/{self.n_samples}]{save_log}", + rank0_only=False, + ) + self.run_sample(trainer, model, data_batch, iteration, should_save) + self.sample_counter += 1 + log.debug( + f"Callback {self.__class__.__name__} finished on validation_step_end step {iteration} [{current_samples + 1}/{self.n_samples}]{save_log}", + rank0_only=False, + ) + + if self.barrier_after_run: + distributed.barrier() + + def run_sample(self, trainer, model, data_batch, iteration, should_save): + tag = "ema" if self.is_ema else "reg" + sample_counter = getattr(trainer, "sample_counter", iteration) + batch_info = { + "data": { + k: convert_to_primitive(v) + for k, v in data_batch.items() + if is_primitive(v) or isinstance(v, (list, dict)) + }, + "sample_counter": sample_counter, + "iteration": iteration, + } + if is_tp_cp_pp_rank0(): + if self.save_s3 and self.data_parallel_id < self.n_sample_to_save: + easy_io.dump( + batch_info, + f"s3://rundir/{self.name}/BatchInfo_ReplicateID{self.data_parallel_id:04d}_Iter{iteration:09d}.json", + ) + + log.debug("entering, every_n_impl", rank0_only=False) + if self.do_x0_prediction: + log.debug("entering, x0_pred", rank0_only=False) + x0_save_dir, mse_loss, sigmas = self.x0_pred( + model, + data_batch, + iteration, + ) + log.debug("done, x0_pred", rank0_only=False) + if self.save_s3 and self.rank == 0: + easy_io.dump( + { + "mse_loss": mse_loss.tolist(), + "sigmas": sigmas.tolist(), + "iteration": iteration, + }, + f"s3://rundir/{self.name}/{tag}_MSE_Iter{iteration:09d}.json", + ) + + log.debug("entering, sample", rank0_only=False) + sample_save_dir = self.sample( + model, + data_batch, + iteration, + should_save, + ) + log.debug("done, sample", rank0_only=False) + + log.debug("waiting for all ranks to finish", rank0_only=False) + dist.barrier() + if self.rank == 0 and wandb.run: + sample_counter = getattr(trainer, "sample_counter", iteration) + data_type = "image" if model.is_image_batch(data_batch) else "video" + tag += f"_{data_type}" + info = { + "trainer/global_step": iteration, + "sample_counter": sample_counter, + } + if self.do_x0_prediction: + imgs = [] + + assert x0_save_dir is not None + for fp in os.listdir(x0_save_dir): + imgs.append(wandb.Image(os.path.join(x0_save_dir, fp), caption=f"{sample_counter}")) + info[f"{self.name}/{tag}_x0"] = imgs + # convert mse_loss to a dict + mse_loss = mse_loss.tolist() + info.update({f"x0_pred_mse_{tag}/Sigma{sigmas[i]:0.5f}": mse_loss[i] for i in range(len(mse_loss))}) + + assert sample_save_dir is not None + imgs = [] + for idx, fp in enumerate(os.listdir(sample_save_dir)): + imgs.append(wandb.Image(os.path.join(sample_save_dir, fp), caption=f"{sample_counter}")) + + info[f"{self.name}/{tag}_sample"] = imgs + wandb.log( + info, + step=iteration, + ) + torch.cuda.empty_cache() + + @misc.timer("ValidationDrawSample: sample") + def sample(self, model, data_batch, iteration, should_save): + tag = "ema" if self.is_ema else "reg" + + # Obtain text embeddings online + text_encoder_config = getattr(model.config, "text_encoder_config", None) + if text_encoder_config is not None and text_encoder_config.compute_online: + text_embeddings = model.text_encoder.compute_text_embeddings_online(data_batch, model.input_caption_key) + data_batch["t5_text_embeddings"] = text_embeddings + data_batch["t5_text_mask"] = torch.ones(text_embeddings.shape[0], text_embeddings.shape[1], device="cuda") + + raw_data, x0, _ = model.get_data_and_condition(data_batch) + if self.use_negative_prompt: + batch_size = x0.shape[0] + if self.negative_prompt_data["t5_text_embeddings"].shape != data_batch["t5_text_embeddings"].shape: + data_batch["neg_t5_text_embeddings"] = misc.to( + repeat( + self.negative_prompt_data["t5_text_embeddings"], + "... -> b ...", + b=batch_size, + ), + **model.tensor_kwargs, + ) + else: + data_batch["neg_t5_text_embeddings"] = misc.to( + self.negative_prompt_data["t5_text_embeddings"], + **model.tensor_kwargs, + ) + + assert data_batch["neg_t5_text_embeddings"].shape == data_batch["t5_text_embeddings"].shape, ( + f"{data_batch['neg_t5_text_embeddings'].shape} != {data_batch['t5_text_embeddings'].shape}" + ) + data_batch["neg_t5_text_mask"] = data_batch["t5_text_mask"] + + to_show = [] + for guidance in self.guidance: + sample = model.generate_samples_from_batch( + data_batch, + guidance=guidance, + # make sure no mismatch and also works for cp + state_shape=x0.shape[1:], + n_sample=x0.shape[0], + num_steps=self.num_sampling_step, + is_negative_prompt=self.use_negative_prompt, + ) + if hasattr(model, "decode"): + sample = model.decode(sample) + to_show.append(sample.float().cpu()) + + to_show.append(raw_data.float().cpu()) + + batch_size = x0.shape[0] + if should_save: + local_path = self.run_save(to_show, batch_size, iteration, "sample") + return local_path + return None + + def run_save(self, to_show, batch_size, iteration, save_name) -> Optional[str]: + to_show = (1.0 + torch.stack(to_show, dim=0).clamp(-1, 1)) / 2.0 # [n, b, c, t, h, w] + is_single_frame = to_show.shape[3] == 1 + n_viz_sample = min(self.n_viz_sample, batch_size) + + save_path = os.path.join(str(iteration), str(self.sample_counter), save_name) + + # ! we only save first n_sample_to_save video! + if self.save_s3 and self.data_parallel_id < self.n_sample_to_save: + save_img_or_video( + rearrange(to_show, "n b c t h w -> c t (n h) (b w)"), + f"s3://rundir/{self.name}/{save_path}/{self.rank}", + fps=self.fps, + ) + + local_save_dir = os.path.join(self.local_dir, save_path) + os.makedirs(local_save_dir, exist_ok=True) + local_path = os.path.join(local_save_dir, f"{self.rank}.jpg") + + if self.wandb_online: + if is_single_frame: # image case + to_show = rearrange( + to_show[:, :n_viz_sample], + "n b c t h w -> t c (n h) (b w)", + ) + image_grid = torchvision.utils.make_grid(to_show, nrow=1, padding=0, normalize=False) + # resize so that wandb can handle it + torchvision.utils.save_image(resize_image(image_grid, 1024), local_path, nrow=1, scale_each=True) + else: + to_show = to_show[:, :n_viz_sample] # [n, b, c, 3, h, w] + + # resize 3 frames frames so that we can display them on wandb + _T = to_show.shape[3] + three_frames_list = [0, _T // 2, _T - 1] + to_show = to_show[:, :, :, three_frames_list] + log_image_size = 1024 + to_show = rearrange( + to_show, + "n b c t h w -> 1 c (n h) (b t w)", + ) + + # resize so that wandb can handle it + image_grid = torchvision.utils.make_grid(to_show, nrow=1, padding=0, normalize=False) + torchvision.utils.save_image( + resize_image(image_grid, log_image_size), local_path, nrow=1, scale_each=True + ) + + return local_save_dir diff --git a/REGEN-main/cosmos_policy/_src/predict2/callbacks/wandb_log.py b/REGEN-main/cosmos_policy/_src/predict2/callbacks/wandb_log.py new file mode 100644 index 0000000000000000000000000000000000000000..4ce9a833aec3e602ce329b623a2db7aa95a6a29b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/callbacks/wandb_log.py @@ -0,0 +1,242 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import os +from dataclasses import dataclass +from typing import Any, Tuple + +import torch +import torch.distributed as dist +import torch.utils.data +import wandb + +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import distributed, log, misc, wandb_util +from cosmos_policy._src.imaginaire.utils.callback import Callback +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + + +@dataclass +class _LossRecord: + loss: float = 0 + iter_count: int = 0 + edm_loss: float = 0 + + def reset(self) -> None: + self.loss = 0 + self.iter_count = 0 + self.edm_loss = 0 + + def get_stat(self) -> Tuple[float, float]: + if self.iter_count > 0: + avg_loss = self.loss / self.iter_count + avg_edm_loss = self.edm_loss / self.iter_count + dist.all_reduce(avg_loss, op=dist.ReduceOp.AVG) + dist.all_reduce(avg_edm_loss, op=dist.ReduceOp.AVG) + avg_loss = avg_loss.item() + avg_edm_loss = avg_edm_loss.item() + else: + avg_loss = 0 + avg_edm_loss = 0 + self.reset() + return avg_loss, avg_edm_loss + + +class WandbCallback(Callback): + """ + This callback is used to log the loss, average loss over logging_iter_multipler, and unstable counts of image and video to wandb. + """ + + def __init__( + self, + logging_iter_multipler: int = 1, + save_logging_iter_multipler: int = 1, + save_s3: bool = False, + ) -> None: + super().__init__() + self.train_image_log = _LossRecord() + self.train_video_log = _LossRecord() + self.final_loss_log = _LossRecord() + + self.img_unstable_count = torch.zeros(1, device="cuda") + self.video_unstable_count = torch.zeros(1, device="cuda") + + self.logging_iter_multipler = logging_iter_multipler + self.save_logging_iter_multipler = save_logging_iter_multipler + assert self.logging_iter_multipler > 0, "logging_iter_multipler should be greater than 0" + self.save_s3 = save_s3 + self.wandb_extra_tag = f"@{logging_iter_multipler}" if logging_iter_multipler > 1 else "" + self.name = "wandb_loss_log" + self.wandb_extra_tag + + @distributed.rank0_only + def on_train_start(self, model: ImaginaireModel, iteration: int = 0) -> None: + wandb_util.init_wandb(self.config, model=model) + config = self.config + job_local_path = config.job.path_local + # read optional job_env saved by `log_reproducible_setup` + if os.path.exists(f"{job_local_path}/job_env.yaml"): + job_info = easy_io.load(f"{job_local_path}/job_env.yaml") + if wandb.run: + wandb.run.config.update({f"JOB_INFO/{k}": v for k, v in job_info.items()}, allow_val_change=True) + + if os.path.exists(f"{config.job.path_local}/config.yaml") and "SLURM_LOG_DIR" in os.environ: + easy_io.copyfile( + f"{config.job.path_local}/config.yaml", + os.path.join(os.environ["SLURM_LOG_DIR"], "config.yaml"), + ) + + def on_before_optimizer_step( + self, + model_ddp: distributed.DistributedDataParallel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int = 0, + ) -> None: # Log the curent learning rate. + if iteration % self.config.trainer.logging_iter == 0 and distributed.is_rank0(): + info = {} + info["sample_counter"] = getattr(self.trainer, "sample_counter", iteration) + + for i, param_group in enumerate(optimizer.param_groups): + info[f"optim/lr_{i}"] = param_group["lr"] + info[f"optim/weight_decay_{i}"] = param_group["weight_decay"] + + wandb.log(info, step=iteration) + + def on_training_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: + skip_update_due_to_unstable_loss = False + if torch.isnan(loss) or torch.isinf(loss): + skip_update_due_to_unstable_loss = True + log.critical( + f"Unstable loss {loss} at iteration {iteration} with is_image_batch: {model.is_image_batch(data_batch)}", + rank0_only=False, + ) + + if not skip_update_due_to_unstable_loss: + if model.is_image_batch(data_batch): + self.train_image_log.loss += loss.detach().float() + self.train_image_log.iter_count += 1 + self.train_image_log.edm_loss += output_batch["edm_loss"].detach().float() + else: + self.train_video_log.loss += loss.detach().float() + self.train_video_log.iter_count += 1 + self.train_video_log.edm_loss += output_batch["edm_loss"].detach().float() + + self.final_loss_log.loss += loss.detach().float() + self.final_loss_log.iter_count += 1 + self.final_loss_log.edm_loss += output_batch["edm_loss"].detach().float() + else: + if model.is_image_batch(data_batch): + self.img_unstable_count += 1 + else: + self.video_unstable_count += 1 + + if iteration % (self.config.trainer.logging_iter * self.logging_iter_multipler) == 0: + if self.logging_iter_multipler > 1: + timer_results = {} + else: + timer_results = self.trainer.training_timer.compute_average_results() + avg_image_loss, avg_image_edm_loss = self.train_image_log.get_stat() + avg_video_loss, avg_video_edm_loss = self.train_video_log.get_stat() + avg_final_loss, avg_final_edm_loss = self.final_loss_log.get_stat() + + dist.all_reduce(self.img_unstable_count, op=dist.ReduceOp.SUM) + dist.all_reduce(self.video_unstable_count, op=dist.ReduceOp.SUM) + + if distributed.is_rank0(): + info = {f"timer/{key}": value for key, value in timer_results.items()} + info.update( + { + f"train{self.wandb_extra_tag}/image_loss": avg_image_loss, + f"train{self.wandb_extra_tag}/image_edm_loss": avg_image_edm_loss, + f"train{self.wandb_extra_tag}/video_loss": avg_video_loss, + f"train{self.wandb_extra_tag}/video_edm_loss": avg_video_edm_loss, + f"train{self.wandb_extra_tag}/loss": avg_final_loss, + f"train{self.wandb_extra_tag}/edm_loss": avg_final_edm_loss, + f"train{self.wandb_extra_tag}/img_unstable_count": self.img_unstable_count.item(), + f"train{self.wandb_extra_tag}/video_unstable_count": self.video_unstable_count.item(), + "iteration": iteration, + "sample_counter": getattr(self.trainer, "sample_counter", iteration), + } + ) + if self.save_s3: + if ( + iteration + % ( + self.config.trainer.logging_iter + * self.logging_iter_multipler + * self.save_logging_iter_multipler + ) + == 0 + ): + easy_io.dump( + info, + f"s3://rundir/{self.name}/Train_Iter{iteration:09d}.json", + ) + + if wandb: + wandb.log(info, step=iteration) + if self.logging_iter_multipler == 1: + self.trainer.training_timer.reset() + + # reset unstable count + self.img_unstable_count.zero_() + self.video_unstable_count.zero_() + + def on_validation_start( + self, model: ImaginaireModel, dataloader_val: torch.utils.data.DataLoader, iteration: int = 0 + ) -> None: + # Cache for collecting data/output batches. + self._val_cache: dict[str, Any] = dict( + data_batches=[], + output_batches=[], + loss=torch.tensor(0.0, device="cuda"), + sample_size=torch.tensor(0, device="cuda"), + ) + + def on_validation_step_end( + self, + model: ImaginaireModel, + data_batch: dict[str, torch.Tensor], + output_batch: dict[str, torch.Tensor], + loss: torch.Tensor, + iteration: int = 0, + ) -> None: # Collect the validation batch and aggregate the overall loss. + # Collect the validation batch and aggregate the overall loss. + batch_size = misc.get_data_batch_size(data_batch) + self._val_cache["loss"] += loss * batch_size + self._val_cache["sample_size"] += batch_size + + def on_validation_end(self, model: ImaginaireModel, iteration: int = 0) -> None: + # Compute the average validation loss across all devices. + dist.all_reduce(self._val_cache["loss"], op=dist.ReduceOp.SUM) + dist.all_reduce(self._val_cache["sample_size"], op=dist.ReduceOp.SUM) + loss = self._val_cache["loss"].item() / self._val_cache["sample_size"] + # Log data/stats of validation set to W&B. + if distributed.is_rank0(): + log.info(f"Validation loss (iteration {iteration}): {loss}") + wandb.log({"val/loss": loss}, step=iteration) + + def on_train_end(self, model: ImaginaireModel, iteration: int = 0) -> None: + wandb.finish() diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/conditioner.py b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/conditioner.py new file mode 100644 index 0000000000000000000000000000000000000000..32634693316fa065a2aab01a5597c3c804f9ff05 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/conditioner.py @@ -0,0 +1,326 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import random +from dataclasses import dataclass +from typing import Dict, Optional + +import torch +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.utils.context_parallel import broadcast_split_tensor +from cosmos_policy._src.predict2.conditioner import ( + ReMapkey, + Text2WorldCondition, +) +from cosmos_policy._src.predict2.configs.video2world.defaults.conditioner import ( + _SHARED_CONFIG, + Video2WorldCondition, + Video2WorldConditioner, + VideoPredictionConditioner, +) + + +@dataclass(frozen=True) +class CameraConditionedCondition(Video2WorldCondition): + camera: Optional[torch.Tensor] = None + + def set_camera_conditioned_video_condition( + self, + gt_frames: torch.Tensor, + num_conditional_frames: Optional[int] = None, + ) -> "CameraConditionedCondition": + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = gt_frames + + # condition_video_input_mask_B_C_T_H_W + B, _, T, H, W = gt_frames.shape + condition_video_input_mask_B_C_T_H_W = torch.zeros( + B, 1, T, H, W, dtype=gt_frames.dtype, device=gt_frames.device + ) + if T == 1: # handle image batch + num_conditional_frames_B = torch.zeros(B, dtype=torch.int32) + else: # handle video batch + if isinstance(num_conditional_frames, torch.Tensor): + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames.cpu() + else: + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames + for idx in range(B): + condition_video_input_mask_B_C_T_H_W[ + idx, :, num_conditional_frames_B[idx] : num_conditional_frames_B[idx] * 2, :, : + ] += 1 + + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + return type(self)(**kwargs) + + def broadcast(self, process_group: torch.distributed.ProcessGroup) -> "CameraConditionedCondition": + if self.is_broadcasted: + return self + # extra efforts + gt_frames = self.gt_frames + condition_video_input_mask_B_C_T_H_W = self.condition_video_input_mask_B_C_T_H_W + camera = self.camera + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = None + kwargs["condition_video_input_mask_B_C_T_H_W"] = None + new_condition = Text2WorldCondition.broadcast( + type(self)(**kwargs), + process_group, + ) + + kwargs = new_condition.to_dict(skip_underscore=False) + _, _, T, _, _ = gt_frames.shape + if process_group is not None: + if T > 1 and process_group.size() > 1: + gt_frames = broadcast_split_tensor(gt_frames, seq_dim=2, process_group=process_group) + condition_video_input_mask_B_C_T_H_W = broadcast_split_tensor( + condition_video_input_mask_B_C_T_H_W, seq_dim=2, process_group=process_group + ) + camera = broadcast_split_tensor(camera, seq_dim=1, process_group=process_group) + kwargs["gt_frames"] = gt_frames + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + kwargs["camera"] = camera + return type(self)(**kwargs) + + +class CameraConditionedConditioner(Video2WorldConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> CameraConditionedCondition: + output = super()._forward(batch, override_dropout_rate) + assert "camera" in output, "CameraConditionedConditioner requires 'camera' in output" + return CameraConditionedCondition(**output) + + +@dataclass(frozen=True) +class CameraConditionedFrameinitCondition(Video2WorldCondition): + camera: Optional[torch.Tensor] = None + + def set_camera_conditioned_video_condition( + self, + gt_frames: torch.Tensor, + num_conditional_frames: Optional[int] = None, + ) -> "CameraConditionedFrameinitCondition": + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = gt_frames + + # condition_video_input_mask_B_C_T_H_W + B, _, T, H, W = gt_frames.shape + condition_video_input_mask_B_C_T_H_W = torch.zeros( + B, 1, T, H, W, dtype=gt_frames.dtype, device=gt_frames.device + ) + if T == 1: # handle image batch + num_conditional_frames_B = torch.zeros(B, dtype=torch.int32) + else: # handle video batch + if isinstance(num_conditional_frames, torch.Tensor): + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames.cpu() + else: + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames + for idx in range(B): + condition_video_input_mask_B_C_T_H_W[idx, :, 0, :, :] += 1 + condition_video_input_mask_B_C_T_H_W[idx, :, (T // 3) : (T // 3 + num_conditional_frames_B[idx]), :, :] += 1 + condition_video_input_mask_B_C_T_H_W[ + idx, :, (T // 3 * 2) : (T // 3 * 2 + num_conditional_frames_B[idx]), :, : + ] += 1 + + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + return type(self)(**kwargs) + + def broadcast(self, process_group: torch.distributed.ProcessGroup) -> "CameraConditionedFrameinitCondition": + if self.is_broadcasted: + return self + # extra efforts + gt_frames = self.gt_frames + condition_video_input_mask_B_C_T_H_W = self.condition_video_input_mask_B_C_T_H_W + camera = self.camera + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = None + kwargs["condition_video_input_mask_B_C_T_H_W"] = None + new_condition = Text2WorldCondition.broadcast( + type(self)(**kwargs), + process_group, + ) + + kwargs = new_condition.to_dict(skip_underscore=False) + _, _, T, _, _ = gt_frames.shape + if process_group is not None: + if T > 1 and process_group.size() > 1: + gt_frames = broadcast_split_tensor(gt_frames, seq_dim=2, process_group=process_group) + condition_video_input_mask_B_C_T_H_W = broadcast_split_tensor( + condition_video_input_mask_B_C_T_H_W, seq_dim=2, process_group=process_group + ) + camera = broadcast_split_tensor(camera, seq_dim=1, process_group=process_group) + kwargs["gt_frames"] = gt_frames + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + kwargs["camera"] = camera + return type(self)(**kwargs) + + +class CameraConditionedFrameinitConditioner(Video2WorldConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> CameraConditionedFrameinitCondition: + output = super()._forward(batch, override_dropout_rate) + assert "camera" in output, "CameraConditionedFrameinitConditioner requires 'camera' in output" + return CameraConditionedFrameinitCondition(**output) + + +@dataclass(frozen=True) +class CameraConditionedARCondition(Video2WorldCondition): + camera: Optional[torch.Tensor] = None + + def set_camera_conditioned_ar_video_condition( + self, + gt_frames: torch.Tensor, + num_conditional_frames: Optional[int] = None, + is_training: Optional[bool] = True, + is_lvg: Optional[bool] = False, + ) -> "CameraConditionedARCondition": + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = gt_frames + + # condition_video_input_mask_B_C_T_H_W + B, _, T, H, W = gt_frames.shape + condition_video_input_mask_B_C_T_H_W = torch.zeros( + B, 1, T, H, W, dtype=gt_frames.dtype, device=gt_frames.device + ) + + if T == 1: # handle image batch + num_conditional_frames_B = torch.zeros(B, dtype=torch.int32) + else: # handle video batch + if isinstance(num_conditional_frames, torch.Tensor): + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames.cpu() + else: + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames + for idx in range(B): + condition_video_input_mask_B_C_T_H_W[idx, :, : num_conditional_frames_B[idx] * 2, :, :] += 1 + condition_video_input_mask_B_C_T_H_W[idx, :, (-num_conditional_frames_B[idx] * 2) :, :, :] += 1 + + if (is_training and random.random() < 0.45) or is_lvg: + condition_video_input_mask_B_C_T_H_W[ + idx, :, (num_conditional_frames_B[idx] * 2) : (num_conditional_frames_B[idx] * 2 + 6), :, : + ] += 1 + + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + + return type(self)(**kwargs) + + def broadcast(self, process_group: torch.distributed.ProcessGroup) -> "CameraConditionedARCondition": + if self.is_broadcasted: + return self + # extra efforts + gt_frames = self.gt_frames + condition_video_input_mask_B_C_T_H_W = self.condition_video_input_mask_B_C_T_H_W + camera = self.camera + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = None + kwargs["condition_video_input_mask_B_C_T_H_W"] = None + new_condition = Text2WorldCondition.broadcast( + type(self)(**kwargs), + process_group, + ) + + kwargs = new_condition.to_dict(skip_underscore=False) + _, _, T, _, _ = gt_frames.shape + if process_group is not None: + if T > 1 and process_group.size() > 1: + gt_frames = broadcast_split_tensor(gt_frames, seq_dim=2, process_group=process_group) + condition_video_input_mask_B_C_T_H_W = broadcast_split_tensor( + condition_video_input_mask_B_C_T_H_W, seq_dim=2, process_group=process_group + ) + camera = broadcast_split_tensor(camera, seq_dim=1, process_group=process_group) + kwargs["gt_frames"] = gt_frames + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + kwargs["camera"] = camera + return type(self)(**kwargs) + + +class CameraConditionedARConditioner(Video2WorldConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> CameraConditionedARCondition: + output = super()._forward(batch, override_dropout_rate) + assert "camera" in output, "CameraConditionedARConditioner requires 'camera' in output" + return CameraConditionedARCondition(**output) + + +CameraConditionedConditionerConfig: LazyDict = L(CameraConditionedConditioner)( + **_SHARED_CONFIG, + camera=L(ReMapkey)( + input_key="camera", + output_key="camera", + dropout_rate=0.0, + dtype=None, + ), +) + +CameraConditionedFrameinitConditionerConfig: LazyDict = L(CameraConditionedFrameinitConditioner)( + **_SHARED_CONFIG, + camera=L(ReMapkey)( + input_key="camera", + output_key="camera", + dropout_rate=0.0, + dtype=None, + ), +) + +CameraConditionedARConditionerConfig: LazyDict = L(CameraConditionedARConditioner)( + **_SHARED_CONFIG, + camera=L(ReMapkey)( + input_key="camera", + output_key="camera", + dropout_rate=0.0, + dtype=None, + ), +) + + +def register_conditioner(): + cs = ConfigStore.instance() + cs.store( + group="conditioner", + package="model.config.conditioner", + name="video_prediction_conditioner", + node=VideoPredictionConditioner, + ) + + cs.store( + group="conditioner", + package="model.config.conditioner", + name="camera_conditioned_video_conditioner", + node=CameraConditionedConditionerConfig, + ) + + cs.store( + group="conditioner", + package="model.config.conditioner", + name="camera_conditioned_frameinit_video_conditioner", + node=CameraConditionedFrameinitConditionerConfig, + ) + + cs.store( + group="conditioner", + package="model.config.conditioner", + name="camera_conditioned_ar_video_conditioner", + node=CameraConditionedARConditionerConfig, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/config.py b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/config.py new file mode 100644 index 0000000000000000000000000000000000000000..4f815f543494ad6047507d078bfadbe31e3b99a0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/config.py @@ -0,0 +1,104 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any, List + +import attrs + +from cosmos_policy._src.imaginaire import config +from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer as Trainer +from cosmos_policy._src.imaginaire.utils.config_helper import import_all_modules_from_package +from cosmos_policy._src.predict2.camera.configs.multiview_camera.conditioner import register_conditioner +from cosmos_policy._src.predict2.camera.configs.multiview_camera.data import register_camera_data +from cosmos_policy._src.predict2.camera.configs.multiview_camera.model import register_model +from cosmos_policy._src.predict2.camera.configs.multiview_camera.net import register_net +from cosmos_policy._src.predict2.configs.common.defaults.checkpoint import register_checkpoint +from cosmos_policy._src.predict2.configs.common.defaults.ckpt_type import register_ckpt_type +from cosmos_policy._src.predict2.configs.common.defaults.dataloader import register_training_and_val_data +from cosmos_policy._src.predict2.configs.common.defaults.ema import register_ema +from cosmos_policy._src.predict2.configs.common.defaults.optimizer import register_optimizer +from cosmos_policy._src.predict2.configs.common.defaults.scheduler import register_scheduler +from cosmos_policy._src.predict2.configs.common.defaults.tokenizer import register_tokenizer +from cosmos_policy._src.predict2.configs.video2world.defaults.callbacks import register_callbacks + + +@attrs.define(slots=False) +class Config(config.Config): + # default config groups that will be used unless overwritten + # see config groups in registry.py + defaults: List[Any] = attrs.field( + factory=lambda: [ + "_self_", + {"data_train": "mock"}, + {"data_val": "mock"}, + {"optimizer": "fusedadamw"}, + {"scheduler": "lambdalinear"}, + {"model": "ddp"}, + {"callbacks": "basic"}, + {"net": None}, + {"conditioner": "video_prediction_conditioner"}, + {"ema": "power"}, + {"tokenizer": "cosmos_tokenizer_causal_cv8x8x8_c16_res720_t121_it121_v1_0"}, + {"checkpoint": "s3"}, + {"ckpt_type": "dummy"}, + # the list is with order, we need global experiment to be the last one + {"experiment": None}, + ] + ) + + +def make_config() -> Config: + c = Config( + model=None, + optimizer=None, + scheduler=None, + dataloader_train=None, + dataloader_val=None, + ) + + # Specifying values through instances of attrs + c.job.project = "cosmos_diffusion_v2" + c.job.group = "debug" + c.job.name = "delete_${now:%Y-%m-%d}_${now:%H-%M-%S}" + + c.trainer.type = Trainer + c.trainer.straggler_detection.enabled = False + c.trainer.max_iter = 400_000 + c.trainer.logging_iter = 10 + c.trainer.validation_iter = 100 + c.trainer.run_validation = False + c.trainer.callbacks = None + + # Call this function to register config groups for advanced overriding. the order follows the default config groups + register_training_and_val_data() + # add camera conditioned data + register_camera_data() + register_optimizer() + register_scheduler() + register_model() + register_callbacks() + register_net() + register_conditioner() + register_ema() + register_tokenizer() + register_checkpoint() + register_ckpt_type() + + # experiment config are defined in the experiment folder + # call import_all_modules_from_package to register them + import_all_modules_from_package( + "cosmos_policy._src.predict2.camera.configs.multiview_camera.experiment", reload=True + ) + return c diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/data.py b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/data.py new file mode 100644 index 0000000000000000000000000000000000000000..a310c3b444f8dd3eef2cdebe6287e5a87f621914 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/data.py @@ -0,0 +1,20 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + + +def register_camera_data(): + cs = ConfigStore.instance() diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/experiment/exp_2b.py b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/experiment/exp_2b.py new file mode 100644 index 0000000000000000000000000000000000000000..ebee584377e35bcb97331f9e65d1345458574324 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/experiment/exp_2b.py @@ -0,0 +1,306 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.configs.video2world.experiment.reason_embeddings.model_2B_reason_1p1 import ( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16, +) +from cosmos_policy._src.predict2.configs.video2world.experiment.reason_embeddings.stage3_2B import ( + I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY, + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_102_SIZE_2B_RES_480_FPS16_HQ_V5_from_26, + build_debug_runs, +) +from cosmos_policy._src.predict2.configs.video2world.experiment.specialized_model.SFT_2B_RF import ( + STAGE_C_PT_4_INDEX_2_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_WITH_EDM_CKPT, +) + +""" +# run local debug & training +""" + + +MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16 = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt", + {"override /data_train": "local_multiview_train"}, + {"override /net": "cosmos_v1_2B_net_camera_conditioned"}, + {"override /conditioner": "camera_conditioned_video_conditioner"}, + {"override /model": "camera_conditioned_rectified_flow_fsdp"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="multicamera_video2video_rectified_flow_2b_res_720_fps16", + ), + model_parallel=dict( + context_parallel_size=2, + ), + checkpoint=dict( + save_iter=500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_GRPO-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-posttrain_data-HQ_V7_RF_MERGE_LOCAL_ag_every2_guidance0_scorekeyoverall_reward_databeta0.01_mincon0/checkpoints/iter_000000288", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=40_000, + logging_iter=200, + straggler_detection=dict( + enabled=False, + max_diff=1.5, + ), + ), + dataloader_train=dict( + batch_size=1, + ), + ), + flags={"allow_objects": True}, +) + + +MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_480_FPS16_S3_MULTICAM_SYNCAM = LazyDict( + dict( + defaults=[ + "/experiment/multicamera_video2video_rectified_flow_2b_res_720_fps16", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="multicamera_video2video_rectified_flow_2b_res_480_fps16_s3_multicam_syncam", + ), + dataloader_train=dict( + batch_size=1, + ), + model_parallel=dict( + context_parallel_size=2, + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/multicamera_video2video_2b_res_720_fps16_s3_multicam_syncam/checkpoints/iter_000011000/", + ), + ), + flags={"allow_objects": True}, +) + + +MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_MULTICAM_SYNCAM = LazyDict( + dict( + defaults=[ + "/experiment/multicamera_video2video_rectified_flow_2b_res_720_fps16", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_multicam_syncam", + ), + dataloader_train=dict( + batch_size=1, + ), + model_parallel=dict( + context_parallel_size=4, + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/multicamera_video2video_2b_res_720_fps16_s3_multicam_syncam/checkpoints/iter_000011000/", + ), + ), + flags={"allow_objects": True}, +) + + +MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_AGIBOT = LazyDict( + dict( + defaults=[ + "/experiment/multicamera_video2video_rectified_flow_2b_res_720_fps16", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_agibot", + ), + dataloader_train=dict( + batch_size=1, + ), + model_parallel=dict( + context_parallel_size=4, + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/multicamera_video2video_2b_res_720_fps16_s3_agibot/checkpoints/iter_000015000/", + ), + ), + flags={"allow_objects": True}, +) + +MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_AGIBOT_FRAMEINIT = LazyDict( + dict( + defaults=[ + "/experiment/multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_agibot", + {"override /model": "camera_conditioned_frameinit_rectified_flow_fsdp"}, + {"override /conditioner": "camera_conditioned_frameinit_video_conditioner"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_agibot_frameinit", + ), + dataloader_train=dict( + batch_size=1, + ), + model_parallel=dict( + context_parallel_size=8, + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_agibot/checkpoints/iter_000003000", + ), + ), + flags={"allow_objects": True}, +) + +MULTICAMERA_AR_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_MULTICAM_SYNCAM = LazyDict( + dict( + defaults=[ + "/experiment/multicamera_video2video_rectified_flow_2b_res_720_fps16", + {"override /data_train": "mock"}, + {"override /conditioner": "camera_conditioned_ar_video_conditioner"}, + {"override /model": "camera_conditioned_ar_rectified_flow_fsdp"}, + {"override /net": "cosmos_v1_2B_net_camera_conditioned_ar"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="multicamera_ar_video2video_rectified_flow_2b_res_720_fps16_s3_multicam_syncam", + ), + dataloader_train=dict( + batch_size=1, + ), + model_parallel=dict( + context_parallel_size=2, + ), + checkpoint=dict( + save_iter=100, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_multicam_syncam/checkpoints/iter_000002000/", + ), + ), + flags={"allow_objects": True}, +) + + +MULTICAMERA_AR_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_480_FPS16_S3_MULTICAM_SYNCAM_IN4OUT1 = LazyDict( + dict( + defaults=[ + "/experiment/multicamera_video2video_rectified_flow_2b_res_720_fps16", + {"override /data_train": "s3_multiview_ar_train_multicam_syncam_480p_in4out1"}, + {"override /conditioner": "camera_conditioned_ar_video_conditioner"}, + {"override /model": "camera_conditioned_ar_rectified_flow_fsdp"}, + {"override /net": "cosmos_v1_2B_net_camera_conditioned_ar"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="multicamera_ar_video2video_rectified_flow_2b_res_480_fps16_s3_multicam_syncam_in4out1", + ), + dataloader_train=dict( + batch_size=1, + ), + model_parallel=dict( + context_parallel_size=2, + ), + checkpoint=dict( + save_iter=200, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/multicamera_video2video_rectified_flow_2b_res_720_fps16_s3_multicam_syncam/checkpoints/iter_000002000/", + ), + ), + flags={"allow_objects": True}, +) + + +""" +# run s3 debug +""" + + +""" +# run webdataset debug +""" + + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_102_SIZE_2B_RES_480_FPS16_HQ_V5_from_26, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_102_SIZE_2B_RES_480_FPS16_HQ_V5_from_26), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16), + ], + [ + STAGE_C_PT_4_INDEX_2_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_WITH_EDM_CKPT, + *build_debug_runs(STAGE_C_PT_4_INDEX_2_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_WITH_EDM_CKPT), + ], + [ + MULTICAMERA_AR_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_MULTICAM_SYNCAM, + *build_debug_runs(MULTICAMERA_AR_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_MULTICAM_SYNCAM), + ], + [ + MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_480_FPS16_S3_MULTICAM_SYNCAM, + *build_debug_runs(MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_480_FPS16_S3_MULTICAM_SYNCAM), + ], + [ + MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_MULTICAM_SYNCAM, + *build_debug_runs(MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_MULTICAM_SYNCAM), + ], + [ + MULTICAMERA_AR_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_480_FPS16_S3_MULTICAM_SYNCAM_IN4OUT1, + *build_debug_runs(MULTICAMERA_AR_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_480_FPS16_S3_MULTICAM_SYNCAM_IN4OUT1), + ], + [ + MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_AGIBOT_FRAMEINIT, + *build_debug_runs(MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_AGIBOT_FRAMEINIT), + ], + [ + MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16, + *build_debug_runs(MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16), + ], + [ + MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_AGIBOT, + *build_debug_runs(MULTICAMERA_VIDEO2VIDEO_RECTIFIED_FLOW_SIZE_2B_RES_720_FPS16_S3_AGIBOT), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/model.py b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/model.py new file mode 100644 index 0000000000000000000000000000000000000000..37715dd07b9bb82fb47dd375bd6099c7bbebab80 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/model.py @@ -0,0 +1,88 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.camera.models.multiview_camera_ar_video2world_model import ( + CameraConditionedARVideo2WorldModelRectifiedFlow, + CameraConditionedARVideo2WorldRectifiedFlowConfig, +) +from cosmos_policy._src.predict2.camera.models.multiview_camera_frameinit_video2world_model import ( + CameraConditionedFrameinitVideo2WorldModelRectifiedFlow, + CameraConditionedFrameinitVideo2WorldRectifiedFlowConfig, +) +from cosmos_policy._src.predict2.camera.models.multiview_camera_video2world_model import ( + CameraConditionedVideo2WorldModelRectifiedFlow, + CameraConditionedVideo2WorldRectifiedFlowConfig, +) + +CAMERA_CONDITIONED_FSDP_RECTIFIED_FLOW_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(CameraConditionedVideo2WorldModelRectifiedFlow)( + config=CameraConditionedVideo2WorldRectifiedFlowConfig( + fsdp_shard_size=8, + ), + _recursive_=False, + ), +) + +CAMERA_CONDITIONED_FRAMEINIT_FSDP_RECTIFIED_FLOW_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(CameraConditionedFrameinitVideo2WorldModelRectifiedFlow)( + config=CameraConditionedFrameinitVideo2WorldRectifiedFlowConfig( + fsdp_shard_size=8, + ), + _recursive_=False, + ), +) + +CAMERA_CONDITIONED_AR_FSDP_RECTIFIED_FLOW_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(CameraConditionedARVideo2WorldModelRectifiedFlow)( + config=CameraConditionedARVideo2WorldRectifiedFlowConfig( + fsdp_shard_size=8, + ), + _recursive_=False, + ), +) + + +def register_model(): + cs = ConfigStore.instance() + cs.store( + group="model", + package="_global_", + name="camera_conditioned_rectified_flow_fsdp", + node=CAMERA_CONDITIONED_FSDP_RECTIFIED_FLOW_CONFIG, + ) + cs.store( + group="model", + package="_global_", + name="camera_conditioned_frameinit_rectified_flow_fsdp", + node=CAMERA_CONDITIONED_FRAMEINIT_FSDP_RECTIFIED_FLOW_CONFIG, + ) + cs.store( + group="model", + package="_global_", + name="camera_conditioned_ar_rectified_flow_fsdp", + node=CAMERA_CONDITIONED_AR_FSDP_RECTIFIED_FLOW_CONFIG, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/net.py b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/net.py new file mode 100644 index 0000000000000000000000000000000000000000..91658952327b9994624e25ff037e2a55f453497f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/configs/multiview_camera/net.py @@ -0,0 +1,113 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import copy + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.camera.networks.dit_multiview_camera import ( + CameraMiniTrainDITwithConditionalMask, + SACConfig, +) +from cosmos_policy._src.predict2.camera.networks.dit_multiview_camera_ar import ( + CameraARMiniTrainDITwithConditionalMask, +) + +# ------------------------------------------------------------ +# Camera Conditioned +# ------------------------------------------------------------ + +CAMERA_COSMOS_V1_7B_NET_MININET: LazyDict = L(CameraMiniTrainDITwithConditionalMask)( + max_img_h=240, + max_img_w=240, + max_frames=128, + in_channels=16, + out_channels=16, + patch_spatial=2, + patch_temporal=1, + model_channels=4096, + num_blocks=28, + num_heads=32, + concat_padding_mask=True, + pos_emb_cls="rope3d", + pos_emb_learnable=True, + pos_emb_interpolation="crop", + use_adaln_lora=True, + adaln_lora_dim=256, + atten_backend="minimal_a2a", + extra_per_block_abs_pos_emb=True, + rope_h_extrapolation_ratio=1.0, + rope_w_extrapolation_ratio=1.0, + rope_t_extrapolation_ratio=2.0, + sac_config=SACConfig(), +) +CAMERA_COSMOS_V1_2B_NET_MININET = copy.deepcopy(CAMERA_COSMOS_V1_7B_NET_MININET) +CAMERA_COSMOS_V1_2B_NET_MININET.model_channels = 2048 +CAMERA_COSMOS_V1_2B_NET_MININET.num_heads = 16 +CAMERA_COSMOS_V1_2B_NET_MININET.num_blocks = 28 +CAMERA_COSMOS_V1_2B_NET_MININET.extra_per_block_abs_pos_emb = False +CAMERA_COSMOS_V1_2B_NET_MININET.rope_t_extrapolation_ratio = 1.0 + + +CAMERA_COSMOS_AR_V1_7B_NET_MININET: LazyDict = L(CameraARMiniTrainDITwithConditionalMask)( + max_img_h=240, + max_img_w=240, + max_frames=128, + in_channels=16, + out_channels=16, + patch_spatial=2, + patch_temporal=1, + model_channels=4096, + num_blocks=28, + num_heads=32, + concat_padding_mask=True, + pos_emb_cls="rope3d", + pos_emb_learnable=True, + pos_emb_interpolation="crop", + use_adaln_lora=True, + adaln_lora_dim=256, + atten_backend="minimal_a2a", + extra_per_block_abs_pos_emb=True, + rope_h_extrapolation_ratio=1.0, + rope_w_extrapolation_ratio=1.0, + rope_t_extrapolation_ratio=2.0, + sac_config=SACConfig(), +) +CAMERA_COSMOS_AR_V1_2B_NET_MININET = copy.deepcopy(CAMERA_COSMOS_AR_V1_7B_NET_MININET) +CAMERA_COSMOS_AR_V1_2B_NET_MININET.model_channels = 2048 +CAMERA_COSMOS_AR_V1_2B_NET_MININET.num_heads = 16 +CAMERA_COSMOS_AR_V1_2B_NET_MININET.num_blocks = 28 +CAMERA_COSMOS_AR_V1_2B_NET_MININET.extra_per_block_abs_pos_emb = False +CAMERA_COSMOS_AR_V1_2B_NET_MININET.rope_t_extrapolation_ratio = 1.0 + + +def register_net(): + cs = ConfigStore.instance() + + # ------------------------------------------------------------ + cs.store( + group="net", + package="model.config.net", + name="cosmos_v1_2B_net_camera_conditioned", + node=CAMERA_COSMOS_V1_2B_NET_MININET, + ) + cs.store( + group="net", + package="model.config.net", + name="cosmos_v1_2B_net_camera_conditioned_ar", + node=CAMERA_COSMOS_AR_V1_2B_NET_MININET, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/models/multiview_camera_ar_video2world_model.py b/REGEN-main/cosmos_policy/_src/predict2/camera/models/multiview_camera_ar_video2world_model.py new file mode 100644 index 0000000000000000000000000000000000000000..602785fe4edd72e3ab311920bca66effd5b900b3 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/models/multiview_camera_ar_video2world_model.py @@ -0,0 +1,307 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Callable, Dict, Tuple + +import attrs +import torch +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor + +from cosmos_policy._src.imaginaire.utils import misc +from cosmos_policy._src.imaginaire.utils.context_parallel import ( + broadcast_split_tensor, + cat_outputs_cp, +) +from cosmos_policy._src.predict2.camera.configs.multiview_camera.conditioner import CameraConditionedCondition +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.models.video2world_model_rectified_flow import ( + NUM_CONDITIONAL_FRAMES_KEY, + Video2WorldModelRectifiedFlow, + Video2WorldModelRectifiedFlowConfig, +) + +IS_PREPROCESSED_KEY = "is_preprocessed" + + +@attrs.define(slots=False) +class CameraConditionedARVideo2WorldRectifiedFlowConfig(Video2WorldModelRectifiedFlowConfig): + pass + + +class CameraConditionedARVideo2WorldModelRectifiedFlow(Video2WorldModelRectifiedFlow): + def get_data_and_condition( + self, data_batch: dict[str, torch.Tensor] + ) -> Tuple[Tensor, Tensor, CameraConditionedCondition]: + self._normalize_multicam_video_databatch_inplace(data_batch) + self._augment_multicam_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + + # Latent cond state + split_size = data_batch["num_frames"].item() + raw_state_cond = data_batch[self.input_data_key + "_cond"] + raw_state_cond_chunks = torch.split(raw_state_cond, split_size_or_sections=split_size, dim=2) + latent_state_cond_list = [] + for raw_state_cond_chunk in raw_state_cond_chunks: + latent_state_cond_chunk = self.encode(raw_state_cond_chunk).contiguous().float() + latent_state_cond_list.append(latent_state_cond_chunk) + + # Latent tgt state + raw_state_src = data_batch[self.input_data_key] + raw_state_src_chunks = torch.split(raw_state_src, split_size_or_sections=split_size, dim=2) + latent_state_src_list = [] + for raw_state_src_chunk in raw_state_src_chunks: + latent_state_src_chunk = self.encode(raw_state_src_chunk).contiguous().float() + latent_state_src_list.append(latent_state_src_chunk) + + raw_state = torch.cat( + ( + raw_state_cond_chunks[0], + raw_state_cond_chunks[1], + raw_state_src_chunks[0], + raw_state_cond_chunks[2], + raw_state_cond_chunks[3], + ), + dim=2, + ) + latent_state = torch.cat( + ( + latent_state_cond_list[0], + latent_state_cond_list[1], + latent_state_src_list[0], + latent_state_cond_list[2], + latent_state_cond_list[3], + ), + dim=2, + ) + + # Condition + camera_list = torch.chunk(data_batch["camera"], len(latent_state_cond_list) + len(latent_state_src_list), dim=1) + camera = torch.cat((camera_list[0], camera_list[1], camera_list[4], camera_list[2], camera_list[3]), dim=1) + data_batch["camera"] = camera + condition = self.conditioner(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + condition = condition.set_camera_conditioned_ar_video_condition( + gt_frames=latent_state.to(**self.tensor_kwargs), + num_conditional_frames=data_batch.get(NUM_CONDITIONAL_FRAMES_KEY, None), + is_training=True, + is_lvg=False, + ) + + # torch.distributed.breakpoint() + return raw_state, latent_state, condition + + def _normalize_multicam_video_databatch_inplace( + self, data_batch: dict[str, torch.Tensor], input_key: str = None + ) -> None: + """ + Normalizes video data in-place on a CUDA device to reduce data loading overhead. + """ + input_key = self.input_data_key if input_key is None else input_key + # only handle video batch + if input_key in data_batch: + # Check if the data has already been normalized and avoid re-normalizing + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert torch.is_floating_point(data_batch[input_key]), "Video data is not in float format." + assert torch.all( + (data_batch[input_key] >= -1.0001) + & (data_batch[input_key] <= 1.0001) + & (data_batch[input_key + "_cond"] >= -1.0001) + & (data_batch[input_key + "_cond"] <= 1.0001) + ), ( + f"Video data is not in the range [-1, 1]. get data range [{data_batch[input_key].min()}, {data_batch[input_key].max()}]" + ) + else: + assert data_batch[input_key].dtype == torch.uint8, "Video data is not in uint8 format." + data_batch[input_key] = data_batch[input_key].to(**self.tensor_kwargs) / 127.5 - 1.0 + data_batch[input_key + "_cond"] = data_batch[input_key + "_cond"].to(**self.tensor_kwargs) / 127.5 - 1.0 + data_batch[IS_PREPROCESSED_KEY] = True + + def _augment_multicam_image_dim_inplace(self, data_batch: dict[str, torch.Tensor], input_key: str = None) -> None: + input_key = self.input_image_key if input_key is None else input_key + if input_key in data_batch: + # Check if the data has already been augmented and avoid re-augmenting + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert data_batch[input_key].shape[2] == 1, ( + f"Image data is claimed be augmented while its shape is {data_batch[input_key].shape}" + ) + return + else: + data_batch[input_key] = rearrange(data_batch[input_key], "b c h w -> b c 1 h w").contiguous() + data_batch[input_key + "_cond"] = rearrange( + data_batch[input_key + "_cond"], "b c h w -> b c 1 h w" + ).contiguous() + data_batch[IS_PREPROCESSED_KEY] = True + + @torch.no_grad() + def get_velocity_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + num_input_video: int = 2, + num_output_video: int = 1, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `x0_fn` based on the provided data batch and guidance factor. + """ + + camera_list = torch.chunk(data_batch["camera"], num_input_video + num_output_video, dim=1) + camera = torch.cat((camera_list[0], camera_list[1], camera_list[4], camera_list[2], camera_list[3]), dim=1) + data_batch["camera"] = camera + + x0_cond_chunks = torch.chunk(data_batch[self.input_data_key], num_input_video, dim=2) + x0_cond_list = [] + for x0_cond_chunk in x0_cond_chunks: + x0_cond = self.encode(x0_cond_chunk).contiguous().float() + x0_cond_list.append(x0_cond) + x0_conds = torch.cat(x0_cond_list, dim=2) + data_batch["video_cond"] = x0_conds + + tgt_video = torch.zeros_like(x0_cond) + if data_batch["tgt_video_cond"] is not None: + tgt_video_cond = self.encode(data_batch["tgt_video_cond"]).contiguous().float() + tgt_video[:, :, : tgt_video_cond.shape[2], :, :] = tgt_video_cond + + x0 = torch.cat([x0_cond_list[0], x0_cond_list[1], tgt_video, x0_cond_list[2], x0_cond_list[3]], dim=2) + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + is_image_batch = self.is_image_batch(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + + if data_batch["tgt_video_cond"] is None: + is_lvg = False + else: + is_lvg = True + + # override condition with inference mode; num_conditional_frames used Here! + condition = condition.set_camera_conditioned_ar_video_condition( + gt_frames=x0, + num_conditional_frames=data_batch[NUM_CONDITIONAL_FRAMES_KEY], + is_training=False, + is_lvg=is_lvg, + ) + uncondition = uncondition.set_camera_conditioned_ar_video_condition( + gt_frames=x0, + num_conditional_frames=data_batch[NUM_CONDITIONAL_FRAMES_KEY], + is_training=False, + is_lvg=is_lvg, + ) + + # torch.distributed.breakpoint() + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def velocity_fn(noise: torch.Tensor, noise_x: torch.Tensor, timestep: torch.Tensor) -> torch.Tensor: + cond_v = self.denoise(noise, noise_x, timestep, condition) + uncond_v = self.denoise(noise, noise_x, timestep, uncondition) + velocity_pred = cond_v + guidance * (cond_v - uncond_v) + return velocity_pred + + return velocity_fn, x0_cond_list + + @torch.no_grad() + def generate_samples_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + num_input_video: int = 3, + num_output_video: int = 1, + is_negative_prompt: bool = False, + num_steps: int = 35, + shift: float = 5.0, + **kwargs, + ) -> torch.Tensor: + """ + Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples. + """ + + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + _T = _T // num_input_video + state_shape = [ + self.config.state_ch, + self.tokenizer.get_latent_num_frames(_T), + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + velocity_fn, x0_cond_list = self.get_velocity_fn_from_batch( + data_batch, guidance, num_input_video, num_output_video, is_negative_prompt=is_negative_prompt + ) + + noise_list = [] + for i in range(num_output_video): + noise = misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + + noise_list.append(noise) + + noise = torch.cat([x0_cond_list[0], x0_cond_list[1], noise_list[0], x0_cond_list[2], x0_cond_list[3]], dim=2) + + seed_g = torch.Generator(device=self.tensor_kwargs["device"]) + seed_g.manual_seed(seed) + + self.sample_scheduler.set_timesteps(num_steps, device=self.tensor_kwargs["device"], shift=shift) + + timesteps = self.sample_scheduler.timesteps + + if self.net.is_context_parallel_enabled: + noise = broadcast_split_tensor(tensor=noise, seq_dim=2, process_group=self.get_context_parallel_group()) + latents = noise + + for _, t in enumerate(timesteps): + latent_model_input = latents + timestep = [t] + + timestep = torch.stack(timestep) + + velocity_pred = velocity_fn(noise, latent_model_input, timestep.unsqueeze(0)) + temp_x0 = self.sample_scheduler.step( + velocity_pred.unsqueeze(0), t, latents[0].unsqueeze(0), return_dict=False, generator=seed_g + )[0] + latents = temp_x0.squeeze(0) + + if self.net.is_context_parallel_enabled: + latents = cat_outputs_cp(latents, seq_dim=2, cp_group=self.get_context_parallel_group()) + + sample_chunks = torch.chunk(latents, num_input_video + num_output_video, dim=2) + sample_list = [sample_chunks[2]] + + return sample_list diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/models/multiview_camera_frameinit_video2world_model.py b/REGEN-main/cosmos_policy/_src/predict2/camera/models/multiview_camera_frameinit_video2world_model.py new file mode 100644 index 0000000000000000000000000000000000000000..b0c7a28a3f261a4d1ea57a8ce3ca7e5e58dd2830 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/models/multiview_camera_frameinit_video2world_model.py @@ -0,0 +1,280 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Callable, Dict, Tuple + +import attrs +import torch +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor + +from cosmos_policy._src.imaginaire.utils import misc +from cosmos_policy._src.imaginaire.utils.context_parallel import ( + broadcast_split_tensor, + cat_outputs_cp, +) +from cosmos_policy._src.predict2.camera.configs.multiview_camera.conditioner import CameraConditionedCondition +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.models.video2world_model_rectified_flow import ( + NUM_CONDITIONAL_FRAMES_KEY, + Video2WorldModelRectifiedFlow, + Video2WorldModelRectifiedFlowConfig, +) + +IS_PREPROCESSED_KEY = "is_preprocessed" + + +@attrs.define(slots=False) +class CameraConditionedFrameinitVideo2WorldRectifiedFlowConfig(Video2WorldModelRectifiedFlowConfig): + pass + + +class CameraConditionedFrameinitVideo2WorldModelRectifiedFlow(Video2WorldModelRectifiedFlow): + def get_data_and_condition( + self, data_batch: dict[str, torch.Tensor] + ) -> Tuple[Tensor, Tensor, CameraConditionedCondition]: + self._normalize_multicam_video_databatch_inplace(data_batch) + self._augment_multicam_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + + # Latent cond state + split_size = data_batch["num_frames"].item() + raw_state_cond = data_batch[self.input_data_key + "_cond"] + raw_state_cond_chunks = torch.split(raw_state_cond, split_size_or_sections=split_size, dim=2) + latent_state_cond_list = [] + for raw_state_cond_chunk in raw_state_cond_chunks: + latent_state_cond_chunk = self.encode(raw_state_cond_chunk).contiguous().float() + latent_state_cond_list.append(latent_state_cond_chunk) + + # Latent tgt state + raw_state_src = data_batch[self.input_data_key] + raw_state_src_chunks = torch.split(raw_state_src, split_size_or_sections=split_size, dim=2) + latent_state_src_list = [] + for raw_state_src_chunk in raw_state_src_chunks: + latent_state_src_chunk = self.encode(raw_state_src_chunk).contiguous().float() + latent_state_src_list.append(latent_state_src_chunk) + + raw_state = torch.cat( + (raw_state_src_chunks[0], raw_state_cond_chunks[0], raw_state_src_chunks[1]), + dim=2, + ) + latent_state = torch.cat( + (latent_state_src_list[0], latent_state_cond_list[0], latent_state_src_list[1]), + dim=2, + ) + + # Condition + camera_list = torch.chunk(data_batch["camera"], len(latent_state_cond_list) + len(latent_state_src_list), dim=1) + camera = torch.cat((camera_list[1], camera_list[0], camera_list[2]), dim=1) + data_batch["camera"] = camera + + condition = self.conditioner(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + condition = condition.set_camera_conditioned_video_condition( + gt_frames=latent_state.to(**self.tensor_kwargs), + num_conditional_frames=data_batch.get(NUM_CONDITIONAL_FRAMES_KEY, None), + ) + + # torch.distributed.breakpoint() + return raw_state, latent_state, condition + + def _normalize_multicam_video_databatch_inplace( + self, data_batch: dict[str, torch.Tensor], input_key: str = None + ) -> None: + """ + Normalizes video data in-place on a CUDA device to reduce data loading overhead. + """ + input_key = self.input_data_key if input_key is None else input_key + # only handle video batch + if input_key in data_batch: + # Check if the data has already been normalized and avoid re-normalizing + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert torch.is_floating_point(data_batch[input_key]), "Video data is not in float format." + assert torch.all( + (data_batch[input_key] >= -1.0001) + & (data_batch[input_key] <= 1.0001) + & (data_batch[input_key + "_cond"] >= -1.0001) + & (data_batch[input_key + "_cond"] <= 1.0001) + ), ( + f"Video data is not in the range [-1, 1]. get data range [{data_batch[input_key].min()}, {data_batch[input_key].max()}]" + ) + else: + assert data_batch[input_key].dtype == torch.uint8, "Video data is not in uint8 format." + data_batch[input_key] = data_batch[input_key].to(**self.tensor_kwargs) / 127.5 - 1.0 + data_batch[input_key + "_cond"] = data_batch[input_key + "_cond"].to(**self.tensor_kwargs) / 127.5 - 1.0 + data_batch[IS_PREPROCESSED_KEY] = True + + def _augment_multicam_image_dim_inplace(self, data_batch: dict[str, torch.Tensor], input_key: str = None) -> None: + input_key = self.input_image_key if input_key is None else input_key + if input_key in data_batch: + # Check if the data has already been augmented and avoid re-augmenting + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert data_batch[input_key].shape[2] == 1, ( + f"Image data is claimed be augmented while its shape is {data_batch[input_key].shape}" + ) + return + else: + data_batch[input_key] = rearrange(data_batch[input_key], "b c h w -> b c 1 h w").contiguous() + data_batch[input_key + "_cond"] = rearrange( + data_batch[input_key + "_cond"], "b c h w -> b c 1 h w" + ).contiguous() + data_batch[IS_PREPROCESSED_KEY] = True + + @torch.no_grad() + def get_velocity_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + num_output_video: int = 3, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `x0_fn` based on the provided data batch and guidance factor. + """ + + if NUM_CONDITIONAL_FRAMES_KEY in data_batch: + num_conditional_frames = data_batch[NUM_CONDITIONAL_FRAMES_KEY] + else: + num_conditional_frames = 1 + + camera_list = torch.chunk(data_batch["camera"], num_output_video, dim=1) + camera = torch.cat((camera_list[1], camera_list[0], camera_list[2]), dim=1) + data_batch["camera"] = camera + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + is_image_batch = self.is_image_batch(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + + x0_cond_chunks = torch.chunk(data_batch[self.input_data_key], num_output_video, dim=2) + x0_cond_list = [] + for x0_cond_chunk in x0_cond_chunks: + x0_cond = self.encode(x0_cond_chunk).contiguous().float() + x0_cond_list.append(x0_cond) + + x0 = torch.cat([x0_cond_list[1], x0_cond_list[0], x0_cond_list[2]], dim=2) + # override condition with inference mode; num_conditional_frames used Here! + condition = condition.set_camera_conditioned_video_condition( + gt_frames=x0, + num_conditional_frames=num_conditional_frames, + ) + uncondition = uncondition.set_camera_conditioned_video_condition( + gt_frames=x0, + num_conditional_frames=num_conditional_frames, + ) + + # torch.distributed.breakpoint() + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def velocity_fn(noise: torch.Tensor, noise_x: torch.Tensor, timestep: torch.Tensor) -> torch.Tensor: + cond_v = self.denoise(noise, noise_x, timestep, condition) + uncond_v = self.denoise(noise, noise_x, timestep, uncondition) + velocity_pred = cond_v + guidance * (cond_v - uncond_v) + return velocity_pred + + return velocity_fn, x0_cond_list + + @torch.no_grad() + def generate_samples_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + num_output_video: int = 3, + is_negative_prompt: bool = False, + num_steps: int = 35, + shift: float = 5.0, + **kwargs, + ) -> torch.Tensor: + """ + Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples. + """ + + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + state_shape = [ + self.config.state_ch, + self.tokenizer.get_latent_num_frames(_T // num_output_video), + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + velocity_fn, x0_cond_list = self.get_velocity_fn_from_batch( + data_batch, guidance, num_output_video, is_negative_prompt=is_negative_prompt + ) + + noise_list = [] + for i in range(num_output_video): + noise = misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + noise[:, :, 0, :, :] = x0_cond_list[i][:, :, 0, :, :] + noise_list.append(noise) + + noise = torch.cat(noise_list, dim=2) + + seed_g = torch.Generator(device=self.tensor_kwargs["device"]) + seed_g.manual_seed(seed) + + self.sample_scheduler.set_timesteps(num_steps, device=self.tensor_kwargs["device"], shift=shift) + + timesteps = self.sample_scheduler.timesteps + + if self.net.is_context_parallel_enabled: + noise = broadcast_split_tensor(tensor=noise, seq_dim=2, process_group=self.get_context_parallel_group()) + latents = noise + + for _, t in enumerate(timesteps): + latent_model_input = latents + timestep = [t] + + timestep = torch.stack(timestep) + + velocity_pred = velocity_fn(noise, latent_model_input, timestep.unsqueeze(0)) + temp_x0 = self.sample_scheduler.step( + velocity_pred.unsqueeze(0), t, latents[0].unsqueeze(0), return_dict=False, generator=seed_g + )[0] + latents = temp_x0.squeeze(0) + + if self.net.is_context_parallel_enabled: + latents = cat_outputs_cp(latents, seq_dim=2, cp_group=self.get_context_parallel_group()) + + sample_chunks = torch.chunk(latents, num_output_video, dim=2) + sample_list = [sample_chunks[1], sample_chunks[0], sample_chunks[2]] + + return sample_list diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/models/multiview_camera_video2world_model.py b/REGEN-main/cosmos_policy/_src/predict2/camera/models/multiview_camera_video2world_model.py new file mode 100644 index 0000000000000000000000000000000000000000..9e8be122ce3fc62453656a08210baa9766c3a3fa --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/models/multiview_camera_video2world_model.py @@ -0,0 +1,299 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Callable, Dict, Tuple + +import attrs +import torch +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor + +from cosmos_policy._src.imaginaire.utils import misc +from cosmos_policy._src.imaginaire.utils.context_parallel import ( + broadcast_split_tensor, + cat_outputs_cp, +) +from cosmos_policy._src.predict2.camera.configs.multiview_camera.conditioner import CameraConditionedCondition +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.models.video2world_model_rectified_flow import ( + NUM_CONDITIONAL_FRAMES_KEY, + Video2WorldModelRectifiedFlow, + Video2WorldModelRectifiedFlowConfig, +) + +IS_PREPROCESSED_KEY = "is_preprocessed" + + +@attrs.define(slots=False) +class CameraConditionedVideo2WorldRectifiedFlowConfig(Video2WorldModelRectifiedFlowConfig): + pass + + +class CameraConditionedVideo2WorldModelRectifiedFlow(Video2WorldModelRectifiedFlow): + def get_data_and_condition( + self, data_batch: dict[str, torch.Tensor] + ) -> Tuple[Tensor, Tensor, CameraConditionedCondition]: + self._normalize_multicam_video_databatch_inplace(data_batch) + self._augment_multicam_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + + # Latent cond state + split_size = data_batch["num_frames"].item() + raw_state_cond = data_batch[self.input_data_key + "_cond"] + raw_state_cond_chunks = torch.split(raw_state_cond, split_size_or_sections=split_size, dim=2) + latent_state_cond_list = [] + for raw_state_cond_chunk in raw_state_cond_chunks: + latent_state_cond_chunk = self.encode(raw_state_cond_chunk).contiguous().float() + latent_state_cond_list.append(latent_state_cond_chunk) + + # Latent tgt state + raw_state_src = data_batch[self.input_data_key] + raw_state_src_chunks = torch.split(raw_state_src, split_size_or_sections=split_size, dim=2) + latent_state_src_list = [] + for raw_state_src_chunk in raw_state_src_chunks: + latent_state_src_chunk = self.encode(raw_state_src_chunk).contiguous().float() + latent_state_src_list.append(latent_state_src_chunk) + + raw_state = torch.cat( + (raw_state_src_chunks[0], raw_state_cond_chunks[0], raw_state_src_chunks[1]), + dim=2, + ) + latent_state = torch.cat( + (latent_state_src_list[0], latent_state_cond_list[0], latent_state_src_list[1]), + dim=2, + ) + + # Condition + camera_list = torch.chunk(data_batch["camera"], len(latent_state_cond_list) + len(latent_state_src_list), dim=1) + camera = torch.cat((camera_list[1], camera_list[0], camera_list[2]), dim=1) + data_batch["camera"] = camera + + condition = self.conditioner(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + condition = condition.set_camera_conditioned_video_condition( + gt_frames=latent_state.to(**self.tensor_kwargs), + num_conditional_frames=data_batch.get(NUM_CONDITIONAL_FRAMES_KEY, None), + ) + + # torch.distributed.breakpoint() + return raw_state, latent_state, condition + + def _normalize_multicam_video_databatch_inplace( + self, data_batch: dict[str, torch.Tensor], input_key: str = None + ) -> None: + """ + Normalizes video data in-place on a CUDA device to reduce data loading overhead. + + This function modifies the video data tensor within the provided data_batch dictionary + in-place, scaling the uint8 data from the range [0, 255] to the normalized range [-1, 1]. + + Warning: + A warning is issued if the data has not been previously normalized. + + Args: + data_batch (dict[str, Tensor]): A dictionary containing the video data under a specific key. + This tensor is expected to be on a CUDA device and have dtype of torch.uint8. + + Side Effects: + Modifies the 'input_data_key' tensor within the 'data_batch' dictionary in-place. + + Note: + This operation is performed directly on the CUDA device to avoid the overhead associated + with moving data to/from the GPU. Ensure that the tensor is already on the appropriate device + and has the correct dtype (torch.uint8) to avoid unexpected behaviors. + """ + input_key = self.input_data_key if input_key is None else input_key + # only handle video batch + if input_key in data_batch: + # Check if the data has already been normalized and avoid re-normalizing + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert torch.is_floating_point(data_batch[input_key]), "Video data is not in float format." + assert torch.all( + (data_batch[input_key] >= -1.0001) + & (data_batch[input_key] <= 1.0001) + & (data_batch[input_key + "_cond"] >= -1.0001) + & (data_batch[input_key + "_cond"] <= 1.0001) + ), ( + f"Video data is not in the range [-1, 1]. get data range [{data_batch[input_key].min()}, {data_batch[input_key].max()}]" + ) + else: + assert data_batch[input_key].dtype == torch.uint8, "Video data is not in uint8 format." + data_batch[input_key] = data_batch[input_key].to(**self.tensor_kwargs) / 127.5 - 1.0 + data_batch[input_key + "_cond"] = data_batch[input_key + "_cond"].to(**self.tensor_kwargs) / 127.5 - 1.0 + data_batch[IS_PREPROCESSED_KEY] = True + + def _augment_multicam_image_dim_inplace(self, data_batch: dict[str, torch.Tensor], input_key: str = None) -> None: + input_key = self.input_image_key if input_key is None else input_key + if input_key in data_batch: + # Check if the data has already been augmented and avoid re-augmenting + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert data_batch[input_key].shape[2] == 1, ( + f"Image data is claimed be augmented while its shape is {data_batch[input_key].shape}" + ) + return + else: + data_batch[input_key] = rearrange(data_batch[input_key], "b c h w -> b c 1 h w").contiguous() + data_batch[input_key + "_cond"] = rearrange( + data_batch[input_key + "_cond"], "b c h w -> b c 1 h w" + ).contiguous() + data_batch[IS_PREPROCESSED_KEY] = True + + @torch.no_grad() + def get_velocity_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + num_input_video: int = 1, + num_output_video: int = 2, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `x0_fn` based on the provided data batch and guidance factor. + """ + + if NUM_CONDITIONAL_FRAMES_KEY in data_batch: + num_conditional_frames = data_batch[NUM_CONDITIONAL_FRAMES_KEY] + else: + num_conditional_frames = 1 + + camera_list = torch.chunk(data_batch["camera"], num_input_video + num_output_video, dim=1) + camera = torch.cat((camera_list[1], camera_list[0], camera_list[2]), dim=1) + data_batch["camera"] = camera + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + is_image_batch = self.is_image_batch(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + + x0_cond_chunks = torch.chunk(data_batch[self.input_data_key], num_input_video, dim=2) + x0_cond_list = [] + for x0_cond_chunk in x0_cond_chunks: + x0_cond = self.encode(x0_cond_chunk).contiguous().float() + x0_cond_list.append(x0_cond) + + x0 = torch.cat([torch.zeros_like(x0_cond), x0_cond_list[0], torch.zeros_like(x0_cond)], dim=2) + # override condition with inference mode; num_conditional_frames used Here! + condition = condition.set_camera_conditioned_video_condition( + gt_frames=x0, + num_conditional_frames=num_conditional_frames, + ) + uncondition = uncondition.set_camera_conditioned_video_condition( + gt_frames=x0, + num_conditional_frames=num_conditional_frames, + ) + + # torch.distributed.breakpoint() + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def velocity_fn(noise: torch.Tensor, noise_x: torch.Tensor, timestep: torch.Tensor) -> torch.Tensor: + cond_v = self.denoise(noise, noise_x, timestep, condition) + uncond_v = self.denoise(noise, noise_x, timestep, uncondition) + velocity_pred = cond_v + guidance * (cond_v - uncond_v) + return velocity_pred + + return velocity_fn, x0_cond_list + + @torch.no_grad() + def generate_samples_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + num_input_video: int = 1, + num_output_video: int = 2, + is_negative_prompt: bool = False, + num_steps: int = 35, + shift: float = 5.0, + **kwargs, + ) -> torch.Tensor: + """ + Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples. + """ + + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + _T = _T // num_input_video + state_shape = [ + self.config.state_ch, + self.tokenizer.get_latent_num_frames(_T), + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + velocity_fn, x0_cond_list = self.get_velocity_fn_from_batch( + data_batch, guidance, num_input_video, num_output_video, is_negative_prompt=is_negative_prompt + ) + + noise_list = [] + for i in range(num_output_video): + noise = misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + noise_list.append(noise) + + noise = torch.cat([noise_list[0], x0_cond_list[0], noise_list[1]], dim=2) + + seed_g = torch.Generator(device=self.tensor_kwargs["device"]) + seed_g.manual_seed(seed) + + self.sample_scheduler.set_timesteps(num_steps, device=self.tensor_kwargs["device"], shift=shift) + + timesteps = self.sample_scheduler.timesteps + + if self.net.is_context_parallel_enabled: + noise = broadcast_split_tensor(tensor=noise, seq_dim=2, process_group=self.get_context_parallel_group()) + latents = noise + + for _, t in enumerate(timesteps): + latent_model_input = latents + timestep = [t] + + timestep = torch.stack(timestep) + + velocity_pred = velocity_fn(noise, latent_model_input, timestep.unsqueeze(0)) + temp_x0 = self.sample_scheduler.step( + velocity_pred.unsqueeze(0), t, latents[0].unsqueeze(0), return_dict=False, generator=seed_g + )[0] + latents = temp_x0.squeeze(0) + + if self.net.is_context_parallel_enabled: + latents = cat_outputs_cp(latents, seq_dim=2, cp_group=self.get_context_parallel_group()) + + sample_chunks = torch.chunk(latents, num_input_video + num_output_video, dim=2) + sample_list = [sample_chunks[0], sample_chunks[2]] + + return sample_list diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/networks/dit_multiview_camera.py b/REGEN-main/cosmos_policy/_src/predict2/camera/networks/dit_multiview_camera.py new file mode 100644 index 0000000000000000000000000000000000000000..134068774363955d47e0240d34a073a892fdcabc --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/networks/dit_multiview_camera.py @@ -0,0 +1,1871 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import collections +import math +from collections import namedtuple +from collections.abc import Sequence +from dataclasses import dataclass +from enum import Enum +from typing import List, Optional, Tuple, Union + +import numpy as np +import torch +import torch.amp as amp +import transformer_engine as te +from einops import rearrange, repeat +from einops.layers.torch import Rearrange +from torch import nn +from torch.distributed import ProcessGroup, get_process_group_ranks +from torch.distributed._composable.fsdp import fully_shard +from torch.distributed.algorithms._checkpoint.checkpoint_wrapper import checkpoint_wrapper as ptd_checkpoint_wrapper + +try: + from torch.utils.checkpoint import CheckpointPolicy, create_selective_checkpoint_contexts +except ImportError: + CheckpointPolicy = None + +from packaging.version import Version +from torchvision import transforms + +if Version(te.__version__) >= Version("2.8.0"): + from transformer_engine.pytorch.attention.rope import apply_rotary_pos_emb +else: + from transformer_engine.pytorch.attention import apply_rotary_pos_emb + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.context_parallel import split_inputs_cp +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.modules.neighborhood_attn import NeighborhoodAttention +from cosmos_policy._src.predict2.networks.a2a_cp import MinimalA2AAttnOp, NattenA2AAttnOp +from cosmos_policy._src.predict2.networks.model_weights_stats import WeightTrainingStat +from cosmos_policy._src.predict2.networks.selective_activation_checkpoint import SACConfig as _SACConfig + + +# selective activation checkpoint; only apply to the minimal v4 model. if there are change in the networks, some policy will not work as we expect. +def predict2_2B_720_context_fn(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + op_count_key = f"{mode}_mm_count" + # from cosmos_policy._src.imaginaire.utils import log + # log.info(f"op_count_key: {op_count_key}, op_count[op_count_key]: {op_count[op_count_key]}, {args[0].shape}, {args[1].shape}") + # there are totally 6 + 4 + 4 + 2 = 16 block + op_count[op_count_key] = (op_count[op_count_key] + 1) % 16 + if op_count[op_count_key] > 8: # recompute self attn first 3 linear layers + return CheckpointPolicy.MUST_SAVE + if "flash_attn" in str(func): + op_count_key = f"{mode}_flash_attn_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 2 + if op_count[op_count_key]: + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_2B_720_context_fn_aggressive(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + # The default policy is to recompute everything. This is the most memory-efficient + # starting point. We then selectively choose what to save. + default_policy = CheckpointPolicy.PREFER_RECOMPUTE + + # Save the output of Flash Attention. This is the most computationally + # expensive part of a transformer block. Saving its output provides a + # good balance between memory savings and computational overhead. + if "flash_attn" in str(func): + return CheckpointPolicy.MUST_SAVE + + # All other operations (e.g., torch.ops.aten.mm.default, layer norms, additions) + # will fall through to the default policy and be recomputed. + return default_policy + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_2B_720_context_fn_aggressive_v2(): + """ + The most memory-aggressive checkpointing policy. Recomputes ALL operations. + """ + + def policy_fn(ctx, func, *args, **kwargs): + # The policy is to always recompute everything. + # This saves the maximum amount of memory but incurs the highest + # computational cost during the backward pass. + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_14B_720_context_fn(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + op_count_key = f"{mode}_mm_count" + # from cosmos_policy._src.imaginaire.utils import log + # log.info(f"op_count_key: {op_count_key}, op_count[op_count_key]: {op_count[op_count_key]}, {args[0].shape}, {args[1].shape}") + # there are totally 6 + 4 + 4 + 2 = 16 block + op_count[op_count_key] = (op_count[op_count_key] + 1) % 16 + if op_count[op_count_key] > 8: # recompute self attn first 1 linear layers + return CheckpointPolicy.MUST_SAVE + if "flash_attn" in str(func): + op_count_key = f"{mode}_flash_attn_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 2 + if op_count[op_count_key]: + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_14B_720_context_fn_aggressive(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + op_count_key = f"{mode}_mm_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 16 + if op_count[op_count_key] > 12: # recompute self attn first 1 linear layers + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def linear_selfattn_context_fn(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + return CheckpointPolicy.MUST_SAVE + if "flash_attn" in str(func): + op_count_key = f"{mode}_flash_attn_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 2 + if op_count[op_count_key]: + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +class CheckpointMode(str, Enum): + NONE = "none" + MM_ONLY = "mm_only" + BLOCK_WISE = "block_wise" + LINEAR_SELFATTN = "linear_selfattn" + PREDICT2_2B_720 = "predict2_2b_720" + PREDICT2_14B_720 = "predict2_14b_720" + PREDICT2_2B_720_AGGRESSIVE = "predict2_2b_720_aggressive" + PREDICT2_2B_720_AGGRESSIVE_V2 = "predict2_2b_720_aggressive_v2" + PREDICT2_14B_720_AGGRESSIVE = "predict2_14b_720_aggressive" + + def __str__(self) -> str: + return self.value + + +@dataclass +class SACConfig(_SACConfig): + def get_context_fn(self): + if self.mode == CheckpointMode.LINEAR_SELFATTN: + return linear_selfattn_context_fn + elif self.mode == CheckpointMode.PREDICT2_2B_720: + return predict2_2B_720_context_fn + elif self.mode == CheckpointMode.PREDICT2_2B_720_AGGRESSIVE: + return predict2_2B_720_context_fn_aggressive + elif self.mode == CheckpointMode.PREDICT2_2B_720_AGGRESSIVE_V2: + return predict2_2B_720_context_fn_aggressive_v2 + elif self.mode == CheckpointMode.PREDICT2_14B_720: + return predict2_14B_720_context_fn + elif self.mode == CheckpointMode.PREDICT2_14B_720_AGGRESSIVE: + return predict2_14B_720_context_fn_aggressive + else: + # Reuse parent class implementation for other modes + return super().get_context_fn() + + +VideoSize = namedtuple("VideoSize", ["T", "H", "W"]) + + +class RMSNorm(torch.nn.Module): + def __init__(self, dim: int, eps: float = 1e-5): + super().__init__() + self.eps = eps + self.weight = nn.Parameter(torch.ones(dim)) + + def reset_parameters(self): + torch.nn.init.ones_(self.weight) + + def _norm(self, x): + return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps) + + def forward(self, x: torch.Tensor) -> torch.Tensor: + output = self._norm(x.float()).type_as(x) + return output * self.weight + + +# ---------------------- Feed Forward Network ----------------------- +class GPT2FeedForward(nn.Module): + def __init__(self, d_model: int, d_ff: int): + super().__init__() + self.activation = nn.GELU() + self.layer1 = nn.Linear(d_model, d_ff, bias=False) + self.layer2 = nn.Linear(d_ff, d_model, bias=False) + + self._layer_id = None + self._dim = d_model + self._hidden_dim = d_ff + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self._dim) + torch.nn.init.trunc_normal_(self.layer1.weight, std=std, a=-3 * std, b=3 * std) + + # scale init by depth as in https://arxiv.org/abs/1908.11365 -- worked slightly better. + std = 1.0 / math.sqrt(self._hidden_dim) + if self._layer_id is not None: + std = std / math.sqrt(2 * (self._layer_id + 1)) + torch.nn.init.trunc_normal_(self.layer2.weight, std=std, a=-3 * std, b=3 * std) + + def forward(self, x: torch.Tensor): + x = self.layer1(x) + + x = self.activation(x) + x = self.layer2(x) + return x + + +def torch_attention_op(q_B_S_H_D, k_B_S_H_D, v_B_S_H_D): + """Computes multi-head attention using PyTorch's native implementation. + + This function provides a PyTorch backend alternative to Transformer Engine's attention operation. + It rearranges the input tensors to match PyTorch's expected format, computes scaled dot-product + attention, and rearranges the output back to the original format. + + The input tensor names use the following dimension conventions: + + - B: batch size + - S: sequence length + - H: number of attention heads + - D: head dimension + + Args: + q_B_S_H_D: Query tensor with shape (batch, seq_len, n_heads, head_dim) + k_B_S_H_D: Key tensor with shape (batch, seq_len, n_heads, head_dim) + v_B_S_H_D: Value tensor with shape (batch, seq_len, n_heads, head_dim) + + Returns: + Attention output tensor with shape (batch, seq_len, n_heads * head_dim) + """ + in_q_shape = q_B_S_H_D.shape + in_k_shape = k_B_S_H_D.shape + q_B_H_S_D = rearrange(q_B_S_H_D, "b ... h k -> b h ... k").view(in_q_shape[0], in_q_shape[-2], -1, in_q_shape[-1]) + k_B_H_S_D = rearrange(k_B_S_H_D, "b ... h v -> b h ... v").view(in_k_shape[0], in_k_shape[-2], -1, in_k_shape[-1]) + v_B_H_S_D = rearrange(v_B_S_H_D, "b ... h v -> b h ... v").view(in_k_shape[0], in_k_shape[-2], -1, in_k_shape[-1]) + result_B_S_HD = rearrange( + torch.nn.functional.scaled_dot_product_attention(q_B_H_S_D, k_B_H_S_D, v_B_H_S_D), "b h ... l -> b ... (h l)" + ) + + return result_B_S_HD + + +class Attention(nn.Module): + """ + A flexible attention module supporting both self-attention and cross-attention mechanisms. + + This module implements a multi-head attention layer that can operate in either self-attention + or cross-attention mode. The mode is determined by whether a context dimension is provided. + The implementation uses scaled dot-product attention and supports optional bias terms and + dropout regularization. + + Args: + query_dim (int): The dimensionality of the query vectors. + context_dim (int, optional): The dimensionality of the context (key/value) vectors. + If None, the module operates in self-attention mode using query_dim. Default: None + n_heads (int, optional): Number of attention heads for multi-head attention. Default: 8 + head_dim (int, optional): The dimension of each attention head. Default: 64 + dropout (float, optional): Dropout probability applied to the output. Default: 0.0 + qkv_format (str, optional): Format specification for QKV tensors. Default: "bshd" + backend (str, optional): Backend to use for the attention operation. Default: "transformer_engine" + + Examples: + >>> # Self-attention with 512 dimensions and 8 heads + >>> self_attn = Attention(query_dim=512) + >>> x = torch.randn(32, 16, 512) # (batch_size, seq_len, dim) + >>> out = self_attn(x) # (32, 16, 512) + + >>> # Cross-attention + >>> cross_attn = Attention(query_dim=512, context_dim=256) + >>> query = torch.randn(32, 16, 512) + >>> context = torch.randn(32, 8, 256) + >>> out = cross_attn(query, context) # (32, 16, 512) + """ + + def __init__( + self, + query_dim: int, + context_dim=None, + n_heads=8, + head_dim=64, + dropout=0.0, + qkv_format: str = "bshd", + backend: str = "transformer_engine", + use_wan_fp32_strategy: bool = False, + ) -> None: + super().__init__() + log.debug( + f"Setting up {self.__class__.__name__}. Query dim is {query_dim}, context_dim is {context_dim} and using " + f"{n_heads} heads with a dimension of {head_dim}." + ) + self.is_selfattn = context_dim is None # self attention + + assert backend in ["transformer_engine", "torch", "minimal_a2a"], f"Invalid backend: {backend}" + self.backend = backend + + context_dim = query_dim if context_dim is None else context_dim + inner_dim = head_dim * n_heads + + self.n_heads = n_heads + self.head_dim = head_dim + self.qkv_format = qkv_format + self.query_dim = query_dim + self.context_dim = context_dim + self.use_wan_fp32_strategy = use_wan_fp32_strategy + + self.q_proj = nn.Linear(query_dim, inner_dim, bias=False) + self.q_norm = te.pytorch.RMSNorm(self.head_dim, eps=1e-6) + + self.k_proj = nn.Linear(context_dim, inner_dim, bias=False) + self.k_norm = te.pytorch.RMSNorm(self.head_dim, eps=1e-6) + + self.v_proj = nn.Linear(context_dim, inner_dim, bias=False) + self.v_norm = nn.Identity() + + self.output_proj = nn.Linear(inner_dim, query_dim, bias=False) + self.output_dropout = nn.Dropout(dropout) if dropout > 1e-4 else nn.Identity() + + if self.backend == "transformer_engine": + from transformer_engine.pytorch.attention import DotProductAttention + + self.attn_op = DotProductAttention( + self.n_heads, + self.head_dim, + num_gqa_groups=self.n_heads, + attention_dropout=0, + qkv_format=qkv_format, + attn_mask_type="no_mask", + ) + elif self.backend == "minimal_a2a": + self.attn_op = MinimalA2AAttnOp() + elif self.backend == "torch": + self.attn_op = torch_attention_op + + self._query_dim = query_dim + self._context_dim = context_dim + self._inner_dim = inner_dim + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self._query_dim) + torch.nn.init.trunc_normal_(self.q_proj.weight, std=std, a=-3 * std, b=3 * std) + std = 1.0 / math.sqrt(self._context_dim) + torch.nn.init.trunc_normal_(self.k_proj.weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.v_proj.weight, std=std, a=-3 * std, b=3 * std) + + std = 1.0 / math.sqrt(self._inner_dim) + torch.nn.init.trunc_normal_(self.output_proj.weight, std=std, a=-3 * std, b=3 * std) + + for layer in self.q_norm, self.k_norm, self.v_norm: + if hasattr(layer, "reset_parameters"): + layer.reset_parameters() + + def compute_qkv(self, x, context=None, rope_emb=None) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor]: + q = self.q_proj(x) + context = x if context is None else context + k = self.k_proj(context) + v = self.v_proj(context) + q, k, v = map( + lambda t: rearrange(t, "b ... (h d) -> b ... h d", h=self.n_heads, d=self.head_dim), + (q, k, v), + ) + + def apply_norm_and_rotary_pos_emb(q, k, v, rope_emb): + q = self.q_norm(q) + k = self.k_norm(k) + v = self.v_norm(v) + if self.is_selfattn and rope_emb is not None: # only apply to self-attention! + if self.use_wan_fp32_strategy: # wan will force q and k to fp32 before rotary pos emb + q = q.to(torch.float32) + k = k.to(torch.float32) + q = apply_rotary_pos_emb(q, rope_emb, tensor_format=self.qkv_format, fused=True) + k = apply_rotary_pos_emb(k, rope_emb, tensor_format=self.qkv_format, fused=True) + return q, k, v + + q, k, v = apply_norm_and_rotary_pos_emb(q, k, v, rope_emb) + + return q, k, v + + def compute_attention(self, q, k, v, video_size: Optional[VideoSize] = None): + additional_args = {} + if isinstance(self.attn_op, (NattenA2AAttnOp, NeighborhoodAttention)): + additional_args["video_size"] = video_size + + result = self.attn_op(q, k, v, **additional_args) # [B, S, H, D] + return self.output_dropout(self.output_proj(result)) + + def forward( + self, + x, + context: Optional[torch.Tensor] = None, + rope_emb: Optional[torch.Tensor] = None, + video_size: Optional[VideoSize] = None, + ): + """ + Args: + x (Tensor): The query tensor of shape [B, Mq, K] + context (Optional[Tensor]): The key tensor of shape [B, Mk, K] or use x as context [self attention] if None + rope_emb (Optional[Tensor]): RoPE embedding tensor, or no RoPE embeddings (i.e. in cross attention) + video_size(VideoSize): Shape [T, H, W] + """ + q, k, v = self.compute_qkv(x, context, rope_emb=rope_emb) + return self.compute_attention(q, k, v, video_size=video_size) + + def set_context_parallel_group(self, process_group, ranks, stream): + # self.attn_op.set_context_parallel_group(process_group, ranks, stream, cp_comm_type="a2a") + self.attn_op.set_context_parallel_group(process_group, ranks, stream) + + +class I2VCrossAttention(Attention): + def __init__(self, *args, img_latent_dim: int = 1024, **kwargs): + super().__init__(*args, **kwargs) + inner_dim = self.head_dim * self.n_heads + self.k_img = nn.Linear(img_latent_dim, inner_dim, bias=False) + self.v_img = nn.Linear(img_latent_dim, inner_dim, bias=False) + self.k_img_norm = te.pytorch.RMSNorm(self.head_dim, eps=1e-6) + + def init_weights(self) -> None: + super().init_weights() + torch.nn.init.trunc_normal_(self.k_img.weight, std=1.0 / math.sqrt(self._inner_dim)) + torch.nn.init.trunc_normal_(self.v_img.weight, std=1.0 / math.sqrt(self._inner_dim)) + self.k_img_norm.reset_parameters() + + def compute_qkv( + self, x, context, rope_emb=None + ) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]: + text_context, img_context = context + q, k, v = super().compute_qkv(x, text_context, rope_emb) + k_img = self.k_img(img_context) + v_img = self.v_img(img_context) + # Rearrange k_img, v_img + k_img, v_img = map( + lambda t: rearrange(t, "b ... (h d) -> b ... h d", h=self.n_heads, d=self.head_dim), + (k_img, v_img), + ) + + return q, k, v, self.k_img_norm(k_img), v_img + + def compute_attention(self, q, k, v, k_img, v_img): + result = self.attn_op(q, k, v) # [B, S, H, D] + result_img = self.attn_op(q, k_img, v_img) + return self.output_dropout(self.output_proj(result + result_img)) + + def forward( + self, + x, + context=None, + rope_emb=None, + ): + q, k, v, k_img, v_img = self.compute_qkv(x, context, rope_emb) + return self.compute_attention(q, k, v, k_img, v_img) + + +class VideoPositionEmb(nn.Module): + def __init__(self): + super().__init__() + self._cp_group = None + + def enable_context_parallel(self, process_group: ProcessGroup): + self._cp_group = process_group + + def disable_context_parallel(self): + self._cp_group = None + + @property + def seq_dim(self): + return 1 + + def forward(self, x_B_T_H_W_C: torch.Tensor, fps=Optional[torch.Tensor]) -> torch.Tensor: + """ + With CP, the function assume that the input tensor is already split. + It delegates the embedding generation to generate_embeddings function. + """ + B_T_H_W_C = x_B_T_H_W_C.shape + if self._cp_group is not None: + cp_ranks = get_process_group_ranks(self._cp_group) + cp_size = len(cp_ranks) + B, T, H, W, C = B_T_H_W_C + B_T_H_W_C = (B, T * cp_size, H, W, C) + embeddings = self.generate_embeddings(B_T_H_W_C, fps=fps) + + return self._split_for_context_parallel(embeddings) + + def generate_embeddings(self, B_T_H_W_C: torch.Size, fps=Optional[torch.Tensor]): + raise NotImplementedError + + def _split_for_context_parallel(self, embeddings): + if self._cp_group is not None: + embeddings = split_inputs_cp(x=embeddings, seq_dim=self.seq_dim, cp_group=self._cp_group) + return embeddings + + +class VideoRopePosition3DEmb(VideoPositionEmb): + def __init__( + self, + *, # enforce keyword arguments + head_dim: int, + len_h: int, + len_w: int, + len_t: int, + base_fps: int = 24, + h_extrapolation_ratio: float = 1.0, + w_extrapolation_ratio: float = 1.0, + t_extrapolation_ratio: float = 1.0, + enable_fps_modulation: bool = True, + **kwargs, # used for compatibility with other positional embeddings; unused in this class + ): + del kwargs + super().__init__() + self.register_buffer("seq", torch.arange(max(len_h, len_w, len_t), dtype=torch.float)) + self.base_fps = base_fps + self.max_h = len_h + self.max_w = len_w + self.max_t = len_t + self.enable_fps_modulation = enable_fps_modulation + dim = head_dim + dim_h = dim // 6 * 2 + dim_w = dim_h + dim_t = dim - 2 * dim_h + assert dim == dim_h + dim_w + dim_t, f"bad dim: {dim} != {dim_h} + {dim_w} + {dim_t}" + + self.register_buffer( + "dim_spatial_range", + torch.arange(0, dim_h, 2)[: (dim_h // 2)].float() / dim_h, + persistent=True, + ) + self.register_buffer( + "dim_temporal_range", + torch.arange(0, dim_t, 2)[: (dim_t // 2)].float() / dim_t, + persistent=True, + ) + self._dim_h = dim_h + self._dim_t = dim_t + + self.h_ntk_factor = h_extrapolation_ratio ** (dim_h / (dim_h - 2)) + self.w_ntk_factor = w_extrapolation_ratio ** (dim_w / (dim_w - 2)) + self.t_ntk_factor = t_extrapolation_ratio ** (dim_t / (dim_t - 2)) + self.reset_parameters() + + def reset_parameters(self) -> None: + dim_h = self._dim_h + dim_t = self._dim_t + + self.seq = torch.arange(max(self.max_h, self.max_w, self.max_t)).float().to(self.dim_spatial_range.device) + self.dim_spatial_range = ( + torch.arange(0, dim_h, 2)[: (dim_h // 2)].float().to(self.dim_spatial_range.device) / dim_h + ) + self.dim_temporal_range = ( + torch.arange(0, dim_t, 2)[: (dim_t // 2)].float().to(self.dim_spatial_range.device) / dim_t + ) + + def generate_embeddings( + self, + B_T_H_W_C: torch.Size, + fps: Optional[torch.Tensor] = None, + h_ntk_factor: Optional[float] = None, + w_ntk_factor: Optional[float] = None, + t_ntk_factor: Optional[float] = None, + ): + """ + Generate embeddings for the given input size. + + Args: + B_T_H_W_C (torch.Size): Input tensor size (Batch, Time, Height, Width, Channels). + fps (Optional[torch.Tensor], optional): Frames per second. Defaults to None. + h_ntk_factor (Optional[float], optional): Height NTK factor. If None, uses self.h_ntk_factor. + w_ntk_factor (Optional[float], optional): Width NTK factor. If None, uses self.w_ntk_factor. + t_ntk_factor (Optional[float], optional): Time NTK factor. If None, uses self.t_ntk_factor. + + Returns: + Not specified in the original code snippet. + """ + h_ntk_factor = h_ntk_factor if h_ntk_factor is not None else self.h_ntk_factor + w_ntk_factor = w_ntk_factor if w_ntk_factor is not None else self.w_ntk_factor + t_ntk_factor = t_ntk_factor if t_ntk_factor is not None else self.t_ntk_factor + + h_theta = 10000.0 * h_ntk_factor + w_theta = 10000.0 * w_ntk_factor + t_theta = 10000.0 * t_ntk_factor + + h_spatial_freqs = 1.0 / (h_theta ** self.dim_spatial_range.float()) + w_spatial_freqs = 1.0 / (w_theta ** self.dim_spatial_range.float()) + temporal_freqs = 1.0 / (t_theta ** self.dim_temporal_range.float()) + + B, T, H, W, _ = B_T_H_W_C + assert H <= self.max_h and W <= self.max_w, ( + f"Input dimensions (H={H}, W={W}) exceed the maximum dimensions (max_h={self.max_h}, max_w={self.max_w})" + ) + half_emb_h = torch.outer(self.seq[:H], h_spatial_freqs) + half_emb_w = torch.outer(self.seq[:W], w_spatial_freqs) + + if self.enable_fps_modulation: + uniform_fps = (fps is None) or (fps.min() == fps.max()) + assert uniform_fps or B == 1 or T == 1, ( + "For video batch, batch size should be 1 for non-uniform fps. For image batch, T should be 1" + ) + + # apply sequence scaling in temporal dimension + if fps is None: # image case + assert T == 1, "T should be 1 for image batch." + half_emb_t = torch.outer(self.seq[:T], temporal_freqs) + else: + half_emb_t = torch.outer(self.seq[:T] / fps[:1] * self.base_fps, temporal_freqs) + else: + half_emb_t = torch.outer(self.seq[:T], temporal_freqs) + + em_T_H_W_D = torch.cat( + [ + repeat(half_emb_t, "t d -> t h w d", h=H, w=W), + repeat(half_emb_h, "h d -> t h w d", t=T, w=W), + repeat(half_emb_w, "w d -> t h w d", t=T, h=H), + ] + * 2, + dim=-1, + ) + + return rearrange(em_T_H_W_D, "t h w d -> (t h w) 1 1 d").float() + + @property + def seq_dim(self): + return 0 + + +class LearnablePosEmbAxis(VideoPositionEmb): + def __init__( + self, + *, # enforce keyword arguments + interpolation: str, + model_channels: int, + len_h: int, + len_w: int, + len_t: int, + **kwargs, + ): + """ + Args: + interpolation (str): we curretly only support "crop", ideally when we need extrapolation capacity, we should adjust frequency or other more advanced methods. they are not implemented yet. + """ + del kwargs # unused + super().__init__() + self.interpolation = interpolation + assert self.interpolation in ["crop"], f"Unknown interpolation method {self.interpolation}" + self.model_channels = model_channels + + self.pos_emb_h = nn.Parameter(torch.zeros(len_h, model_channels)) + self.pos_emb_w = nn.Parameter(torch.zeros(len_w, model_channels)) + self.pos_emb_t = nn.Parameter(torch.zeros(len_t, model_channels)) + + self.reset_parameters() + + def reset_parameters(self): + std = 1.0 / math.sqrt(self.model_channels) + torch.nn.init.trunc_normal_(self.pos_emb_h, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.pos_emb_w, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.pos_emb_t, std=std, a=-3 * std, b=3 * std) + + def generate_embeddings(self, B_T_H_W_C: torch.Size, fps=Optional[torch.Tensor]) -> torch.Tensor: + B, T, H, W, _ = B_T_H_W_C + if self.interpolation == "crop": + emb_h_H = self.pos_emb_h[:H] + emb_w_W = self.pos_emb_w[:W] + emb_t_T = self.pos_emb_t[:T] + emb = ( + repeat(emb_t_T, "t d-> b t h w d", b=B, h=H, w=W) + + repeat(emb_h_H, "h d-> b t h w d", b=B, t=T, w=W) + + repeat(emb_w_W, "w d-> b t h w d", b=B, t=T, h=H) + ) + assert list(emb.shape)[:4] == [B, T, H, W], f"bad shape: {list(emb.shape)[:4]} != {B, T, H, W}" + else: + raise ValueError(f"Unknown interpolation method {self.interpolation}") + + norm = torch.linalg.vector_norm(emb, dim=-1, keepdim=True, dtype=torch.float32) + norm = torch.add(1e-6, norm, alpha=np.sqrt(norm.numel() / emb.numel())) + return emb / norm.to(emb.dtype) + + +def modulate(x, shift, scale): + return x * (1 + scale) + shift + + +class Timesteps(nn.Module): + def __init__(self, num_channels): + super().__init__() + self.num_channels = num_channels + + def forward(self, timesteps_B_T): + assert timesteps_B_T.ndim == 2, f"Expected 2D input, got {timesteps_B_T.ndim}" + # wan need emb to be in fp32 + in_dype = timesteps_B_T.dtype + timesteps = timesteps_B_T.flatten().float() + half_dim = self.num_channels // 2 + exponent = -math.log(10000) * torch.arange(half_dim, dtype=torch.float32, device=timesteps.device) + exponent = exponent / (half_dim - 0.0) + + emb = torch.exp(exponent) + emb = timesteps[:, None].float() * emb[None, :] + + sin_emb = torch.sin(emb) + cos_emb = torch.cos(emb) + emb = torch.cat([cos_emb, sin_emb], dim=-1) + + return rearrange(emb.to(dtype=in_dype), "(b t) d -> b t d", b=timesteps_B_T.shape[0], t=timesteps_B_T.shape[1]) + + +class TimestepEmbedding(nn.Module): + def __init__(self, in_features: int, out_features: int, use_adaln_lora: bool = False): + super().__init__() + log.debug( + f"Using AdaLN LoRA Flag: {use_adaln_lora}. We enable bias if no AdaLN LoRA for backward compatibility." + ) + self.in_dim = in_features + self.out_dim = out_features + self.linear_1 = nn.Linear(in_features, out_features, bias=not use_adaln_lora) + self.activation = nn.SiLU() + self.use_adaln_lora = use_adaln_lora + if use_adaln_lora: + self.linear_2 = nn.Linear(out_features, 3 * out_features, bias=False) + else: + self.linear_2 = nn.Linear(out_features, out_features, bias=False) + + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self.in_dim) + torch.nn.init.trunc_normal_(self.linear_1.weight, std=std, a=-3 * std, b=3 * std) + + std = 1.0 / math.sqrt(self.out_dim) + torch.nn.init.trunc_normal_(self.linear_2.weight, std=std, a=-3 * std, b=3 * std) + + def forward(self, sample: torch.Tensor) -> torch.Tensor: + emb = self.linear_1(sample) + emb = self.activation(emb) + emb = self.linear_2(emb) + + if self.use_adaln_lora: + adaln_lora_B_T_3D = emb + emb_B_T_D = sample + else: + emb_B_T_D = emb + adaln_lora_B_T_3D = None + + return emb_B_T_D, adaln_lora_B_T_3D + + +class FourierFeatures(nn.Module): + """ + Implements a layer that generates Fourier features from input tensors, based on randomly sampled + frequencies and phases. This can help in learning high-frequency functions in low-dimensional problems. + + [B] -> [B, D] + + Parameters: + num_channels (int): The number of Fourier features to generate. + bandwidth (float, optional): The scaling factor for the frequency of the Fourier features. Defaults to 1. + normalize (bool, optional): If set to True, the outputs are scaled by sqrt(2), usually to normalize + the variance of the features. Defaults to False. + + Example: + >>> layer = FourierFeatures(num_channels=256, bandwidth=0.5, normalize=True) + >>> x = torch.randn(10, 256) # Example input tensor + >>> output = layer(x) + >>> print(output.shape) # Expected shape: (10, 256) + """ + + def __init__(self, num_channels, bandwidth=1, normalize=False): + super().__init__() + self.register_buffer("freqs", 2 * np.pi * bandwidth * torch.randn(num_channels), persistent=True) + self.register_buffer("phases", 2 * np.pi * torch.rand(num_channels), persistent=True) + self.gain = np.sqrt(2) if normalize else 1 + self.bandwidth = bandwidth + self.num_channels = num_channels + + self.reset_parameters() + + def reset_parameters(self) -> None: + generator = torch.Generator() + generator.manual_seed(0) + self.freqs = ( + 2 * np.pi * self.bandwidth * torch.randn(self.num_channels, generator=generator).to(self.freqs.device) + ) + self.phases = 2 * np.pi * torch.rand(self.num_channels, generator=generator).to(self.freqs.device) + + def forward(self, x, gain: float = 1.0): + """ + Apply the Fourier feature transformation to the input tensor. + + Args: + x (torch.Tensor): The input tensor. + gain (float, optional): An additional gain factor applied during the forward pass. Defaults to 1. + + Returns: + torch.Tensor: The transformed tensor, with Fourier features applied. + """ + in_dtype = x.dtype + x = x.to(torch.float32).ger(self.freqs.to(torch.float32)).add(self.phases.to(torch.float32)) + x = x.cos().mul(self.gain * gain).to(in_dtype) + return x + + +class PatchEmbed(nn.Module): + """ + PatchEmbed is a module for embedding patches from an input tensor by applying either 3D or 2D convolutional layers, + depending on the . This module can process inputs with temporal (video) and spatial (image) dimensions, + making it suitable for video and image processing tasks. It supports dividing the input into patches + and embedding each patch into a vector of size `out_channels`. + + Parameters: + - spatial_patch_size (int): The size of each spatial patch. + - temporal_patch_size (int): The size of each temporal patch. + - in_channels (int): Number of input channels. Default: 3. + - out_channels (int): The dimension of the embedding vector for each patch. Default: 768. + - bias (bool): If True, adds a learnable bias to the output of the convolutional layers. Default: True. + """ + + def __init__( + self, + spatial_patch_size, + temporal_patch_size, + in_channels=3, + out_channels=768, + ): + super().__init__() + self.spatial_patch_size = spatial_patch_size + self.temporal_patch_size = temporal_patch_size + + self.proj = nn.Sequential( + Rearrange( + "b c (t r) (h m) (w n) -> b t h w (c r m n)", + r=temporal_patch_size, + m=spatial_patch_size, + n=spatial_patch_size, + ), + nn.Linear( + in_channels * spatial_patch_size * spatial_patch_size * temporal_patch_size, out_channels, bias=False + ), + ) + self.dim = in_channels * spatial_patch_size * spatial_patch_size * temporal_patch_size + + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self.dim) + torch.nn.init.trunc_normal_(self.proj[1].weight, std=std, a=-3 * std, b=3 * std) + + def forward(self, x): + """ + Forward pass of the PatchEmbed module. + + Parameters: + - x (torch.Tensor): The input tensor of shape (B, C, T, H, W) where + B is the batch size, + C is the number of channels, + T is the temporal dimension, + H is the height, and + W is the width of the input. + + Returns: + - torch.Tensor: The embedded patches as a tensor, with shape b t h w c. + """ + assert x.dim() == 5 + _, _, T, H, W = x.shape + assert H % self.spatial_patch_size == 0 and W % self.spatial_patch_size == 0, ( + f"H,W {(H, W)} should be divisible by spatial_patch_size {self.spatial_patch_size}" + ) + assert T % self.temporal_patch_size == 0 + x = self.proj(x) + return x + + +class FinalLayer(nn.Module): + """ + The final layer of video DiT. + """ + + def __init__( + self, + hidden_size, + spatial_patch_size, + temporal_patch_size, + out_channels, + use_adaln_lora: bool = False, + adaln_lora_dim: int = 256, + use_wan_fp32_strategy: bool = False, + ): + super().__init__() + self.use_wan_fp32_strategy = use_wan_fp32_strategy + self.layer_norm = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6) + self.linear = nn.Linear( + hidden_size, spatial_patch_size * spatial_patch_size * temporal_patch_size * out_channels, bias=False + ) + self.hidden_size = hidden_size + self.n_adaln_chunks = 2 + self.use_adaln_lora = use_adaln_lora + self.adaln_lora_dim = adaln_lora_dim + if use_adaln_lora: + self.adaln_modulation = nn.Sequential( + nn.SiLU(), + nn.Linear(hidden_size, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, self.n_adaln_chunks * hidden_size, bias=False), + ) + else: + self.adaln_modulation = nn.Sequential( + nn.SiLU(), nn.Linear(hidden_size, self.n_adaln_chunks * hidden_size, bias=False) + ) + + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self.hidden_size) + torch.nn.init.trunc_normal_(self.linear.weight, std=std, a=-3 * std, b=3 * std) + if self.use_adaln_lora: + torch.nn.init.trunc_normal_(self.adaln_modulation[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.zeros_(self.adaln_modulation[2].weight) + else: + torch.nn.init.zeros_(self.adaln_modulation[1].weight) + + self.layer_norm.reset_parameters() + + def forward( + self, + # x_BT_HW_D, + x_B_T_H_W_D, + emb_B_T_D, + adaln_lora_B_T_3D: Optional[torch.Tensor] = None, + ): + if self.use_wan_fp32_strategy: + assert emb_B_T_D.dtype == torch.float32 + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if self.use_adaln_lora: + assert adaln_lora_B_T_3D is not None + shift_B_T_D, scale_B_T_D = ( + self.adaln_modulation(emb_B_T_D) + adaln_lora_B_T_3D[:, :, : 2 * self.hidden_size] + ).chunk(2, dim=-1) + else: + shift_B_T_D, scale_B_T_D = self.adaln_modulation(emb_B_T_D).chunk(2, dim=-1) + + shift_B_T_1_1_D, scale_B_T_1_1_D = ( + rearrange(shift_B_T_D, "b t d -> b t 1 1 d"), + rearrange(scale_B_T_D, "b t d -> b t 1 1 d"), + ) + + def _fn(_x_B_T_H_W_D, _norm_layer, _scale_B_T_1_1_D, _shift_B_T_1_1_D): + return _norm_layer(_x_B_T_H_W_D) * (1 + _scale_B_T_1_1_D) + _shift_B_T_1_1_D + + x_B_T_H_W_D = _fn(x_B_T_H_W_D, self.layer_norm, scale_B_T_1_1_D, shift_B_T_1_1_D) + x_B_T_H_W_O = self.linear( + x_B_T_H_W_D + ) # O = spatial_patch_size * spatial_patch_size * temporal_patch_size * out_channels + return x_B_T_H_W_O + + +class Block(nn.Module): + """ + A transformer block that combines self-attention, cross-attention and MLP layers with AdaLN modulation. + Each component (self-attention, cross-attention, MLP) has its own layer normalization and AdaLN modulation. + + Parameters: + x_dim (int): Dimension of input features + context_dim (int): Dimension of context features for cross-attention + num_heads (int): Number of attention heads + mlp_ratio (float): Multiplier for MLP hidden dimension. Default: 4.0 + use_adaln_lora (bool): Whether to use AdaLN-LoRA modulation. Default: False + adaln_lora_dim (int): Hidden dimension for AdaLN-LoRA layers. Default: 256 + use_wan_fp32_strategy (bool): Whether to use Wan's FP32 strategy. Default: False + If True, in Attention layer, if do self-attention, q and k will be forced to fp32 before rotary pos emb + also, in modulation computation, force entire computation in fp32 + + The block applies the following sequence: + 1. Self-attention with AdaLN modulation + 2. Cross-attention with AdaLN modulation + 3. MLP with AdaLN modulation + + Each component uses skip connections and layer normalization. + """ + + def __init__( + self, + x_dim: int, + context_dim: int, + num_heads: int, + mlp_ratio: float = 4.0, + use_adaln_lora: bool = False, + adaln_lora_dim: int = 256, + cam_dim: int = 1536, + backend: str = "transformer_engine", + image_context_dim: Optional[int] = None, + use_wan_fp32_strategy: bool = False, + ): + super().__init__() + self.x_dim = x_dim + self.layer_norm_self_attn = nn.LayerNorm(x_dim, elementwise_affine=False, eps=1e-6) + self.self_attn = Attention( + x_dim, + None, + num_heads, + x_dim // num_heads, + qkv_format="bshd", + backend=backend, + use_wan_fp32_strategy=use_wan_fp32_strategy, + ) + + self.layer_norm_cross_attn = nn.LayerNorm(x_dim, elementwise_affine=False, eps=1e-6) + self.cross_attn = Attention( + x_dim, context_dim, num_heads, x_dim // num_heads, qkv_format="bshd", backend=backend + ) + if image_context_dim is None: + self.cross_attn = Attention(x_dim, context_dim, num_heads, x_dim // num_heads, qkv_format="bshd") + else: + self.cross_attn = I2VCrossAttention( + x_dim, context_dim, num_heads, x_dim // num_heads, img_latent_dim=image_context_dim, qkv_format="bshd" + ) + + self.layer_norm_mlp = nn.LayerNorm(x_dim, elementwise_affine=False, eps=1e-6) + self.mlp = GPT2FeedForward(x_dim, int(x_dim * mlp_ratio)) + + self.use_adaln_lora = use_adaln_lora + if self.use_adaln_lora: + self.adaln_modulation_self_attn = nn.Sequential( + nn.SiLU(), + nn.Linear(x_dim, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, 3 * x_dim, bias=False), + ) + self.adaln_modulation_cross_attn = nn.Sequential( + nn.SiLU(), + nn.Linear(x_dim, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, 3 * x_dim, bias=False), + ) + self.adaln_modulation_mlp = nn.Sequential( + nn.SiLU(), + nn.Linear(x_dim, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, 3 * x_dim, bias=False), + ) + else: + self.adaln_modulation_self_attn = nn.Sequential(nn.SiLU(), nn.Linear(x_dim, 3 * x_dim, bias=False)) + self.adaln_modulation_cross_attn = nn.Sequential(nn.SiLU(), nn.Linear(x_dim, 3 * x_dim, bias=False)) + self.adaln_modulation_mlp = nn.Sequential(nn.SiLU(), nn.Linear(x_dim, 3 * x_dim, bias=False)) + + self.cam_dim = cam_dim + self.cam_encoder = nn.Linear(self.cam_dim, x_dim, bias=False) + + self.cp_size = None + self.use_wan_fp32_strategy = use_wan_fp32_strategy + + def set_context_parallel_group(self, process_group, ranks, stream): + self.cp_size = None if ranks is None else len(ranks) + self.self_attn.set_context_parallel_group( + process_group=process_group, + ranks=ranks, + stream=stream, + ) + + def reset_parameters(self) -> None: + self.layer_norm_self_attn.reset_parameters() + self.layer_norm_cross_attn.reset_parameters() + self.layer_norm_mlp.reset_parameters() + + if self.use_adaln_lora: + std = 1.0 / math.sqrt(self.x_dim) + torch.nn.init.trunc_normal_(self.adaln_modulation_self_attn[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.adaln_modulation_cross_attn[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.adaln_modulation_mlp[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.zeros_(self.adaln_modulation_self_attn[2].weight) + torch.nn.init.zeros_(self.adaln_modulation_cross_attn[2].weight) + torch.nn.init.zeros_(self.adaln_modulation_mlp[2].weight) + else: + torch.nn.init.zeros_(self.adaln_modulation_self_attn[1].weight) + torch.nn.init.zeros_(self.adaln_modulation_cross_attn[1].weight) + torch.nn.init.zeros_(self.adaln_modulation_mlp[1].weight) + + def init_weights(self) -> None: + self.reset_parameters() + self.self_attn.init_weights() + self.cross_attn.init_weights() + self.mlp.init_weights() + + # init camera weights + std = 1.0 / math.sqrt(self.x_dim) + torch.nn.init.trunc_normal_(self.cam_encoder.weight, std=std, a=-3 * std, b=3 * std) + + def forward( + self, + x_B_T_H_W_D: torch.Tensor, + emb_B_T_D: torch.Tensor, + crossattn_emb: torch.Tensor, + rope_emb_L_1_1_D: Optional[torch.Tensor] = None, + adaln_lora_B_T_3D: Optional[torch.Tensor] = None, + extra_per_block_pos_emb: Optional[torch.Tensor] = None, + camera: Optional[torch.Tensor] = None, + ) -> torch.Tensor: + if extra_per_block_pos_emb is not None: + x_B_T_H_W_D = x_B_T_H_W_D + extra_per_block_pos_emb + + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if self.use_adaln_lora: + shift_self_attn_B_T_D, scale_self_attn_B_T_D, gate_self_attn_B_T_D = ( + self.adaln_modulation_self_attn(emb_B_T_D) + adaln_lora_B_T_3D + ).chunk(3, dim=-1) + shift_cross_attn_B_T_D, scale_cross_attn_B_T_D, gate_cross_attn_B_T_D = ( + self.adaln_modulation_cross_attn(emb_B_T_D) + adaln_lora_B_T_3D + ).chunk(3, dim=-1) + shift_mlp_B_T_D, scale_mlp_B_T_D, gate_mlp_B_T_D = ( + self.adaln_modulation_mlp(emb_B_T_D) + adaln_lora_B_T_3D + ).chunk(3, dim=-1) + else: + shift_self_attn_B_T_D, scale_self_attn_B_T_D, gate_self_attn_B_T_D = self.adaln_modulation_self_attn( + emb_B_T_D + ).chunk(3, dim=-1) + shift_cross_attn_B_T_D, scale_cross_attn_B_T_D, gate_cross_attn_B_T_D = ( + self.adaln_modulation_cross_attn(emb_B_T_D).chunk(3, dim=-1) + ) + shift_mlp_B_T_D, scale_mlp_B_T_D, gate_mlp_B_T_D = self.adaln_modulation_mlp(emb_B_T_D).chunk(3, dim=-1) + + # Reshape tensors from (B, T, D) to (B, T, 1, 1, D) for broadcasting + shift_self_attn_B_T_1_1_D = rearrange(shift_self_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + scale_self_attn_B_T_1_1_D = rearrange(scale_self_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + gate_self_attn_B_T_1_1_D = rearrange(gate_self_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + + shift_cross_attn_B_T_1_1_D = rearrange(shift_cross_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + scale_cross_attn_B_T_1_1_D = rearrange(scale_cross_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + gate_cross_attn_B_T_1_1_D = rearrange(gate_cross_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + + shift_mlp_B_T_1_1_D = rearrange(shift_mlp_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + scale_mlp_B_T_1_1_D = rearrange(scale_mlp_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + gate_mlp_B_T_1_1_D = rearrange(gate_mlp_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + + B, T, H, W, D = x_B_T_H_W_D.shape + + def _fn(_x_B_T_H_W_D, _norm_layer, _scale_B_T_1_1_D, _shift_B_T_1_1_D): + return _norm_layer(_x_B_T_H_W_D) * (1 + _scale_B_T_1_1_D) + _shift_B_T_1_1_D + + normalized_x_B_T_H_W_D = _fn( + x_B_T_H_W_D, + self.layer_norm_self_attn, + scale_self_attn_B_T_1_1_D, + shift_self_attn_B_T_1_1_D, + ) + + video_size = VideoSize(T=T, H=H, W=W) + + # (ahassani): Hack to correct `video_size` when CP is enabled. + # I really don't like this, but there doesn't seem to be any central + # piece of code that's responsible for handling CP/TP that also defines the + # layout of shardings. Other parts of the code (i.e. RoPE) seem to make this + # assumption that CP sharding is always done along T. + if self.cp_size is not None and self.cp_size > 1: + video_size = VideoSize(T=T * self.cp_size, H=H, W=W) + + cam_emb = self.cam_encoder(camera) + + result_B_T_H_W_D = rearrange( + self.self_attn( + # normalized_x_B_T_HW_D, + rearrange(normalized_x_B_T_H_W_D + cam_emb, "b t h w d -> b (t h w) d"), + None, + rope_emb=rope_emb_L_1_1_D, + video_size=video_size, + ), + "b (t h w) d -> b t h w d", + t=T, + h=H, + w=W, + ) + x_B_T_H_W_D = x_B_T_H_W_D + gate_self_attn_B_T_1_1_D * result_B_T_H_W_D + + def _x_fn( + _x_B_T_H_W_D, + layer_norm_cross_attn, + _scale_cross_attn_B_T_1_1_D, + _shift_cross_attn_B_T_1_1_D, + _gate_cross_attn_B_T_1_1_D, + ): + _normalized_x_B_T_H_W_D = _fn( + _x_B_T_H_W_D, layer_norm_cross_attn, _scale_cross_attn_B_T_1_1_D, _shift_cross_attn_B_T_1_1_D + ) + _result_B_T_H_W_D = rearrange( + self.cross_attn( + rearrange(_normalized_x_B_T_H_W_D, "b t h w d -> b (t h w) d"), + crossattn_emb, + rope_emb=rope_emb_L_1_1_D, + ), + "b (t h w) d -> b t h w d", + t=T, + h=H, + w=W, + ) + # _x_B_T_H_W_D = _x_B_T_H_W_D + _gate_cross_attn_B_T_1_1_D * _result_B_T_H_W_D + return _result_B_T_H_W_D + + result_B_T_H_W_D = _x_fn( + x_B_T_H_W_D, + self.layer_norm_cross_attn, + scale_cross_attn_B_T_1_1_D, + shift_cross_attn_B_T_1_1_D, + gate_cross_attn_B_T_1_1_D, + ) + x_B_T_H_W_D = result_B_T_H_W_D * gate_cross_attn_B_T_1_1_D + x_B_T_H_W_D + + normalized_x_B_T_H_W_D = _fn( + x_B_T_H_W_D, + self.layer_norm_mlp, + scale_mlp_B_T_1_1_D, + shift_mlp_B_T_1_1_D, + ) + result_B_T_H_W_D = self.mlp(normalized_x_B_T_H_W_D) + x_B_T_H_W_D = x_B_T_H_W_D + gate_mlp_B_T_1_1_D * result_B_T_H_W_D + return x_B_T_H_W_D + + +class CameraMiniTrainDIT(WeightTrainingStat): + """ + A clean impl of DIT that can load and reproduce the training results of the original DIT model in edify_video/v4~(cosmos 1) + A general implementation of adaln-modulated VIT-like~(DiT) transformer for video processing. + + Args: + max_img_h (int): Maximum height of the input images. + max_img_w (int): Maximum width of the input images. + max_frames (int): Maximum number of frames in the video sequence. + in_channels (int): Number of input channels (e.g., RGB channels for color images). + out_channels (int): Number of output channels. + patch_spatial (tuple): Spatial resolution of patches for input processing. + patch_temporal (int): Temporal resolution of patches for input processing. + concat_padding_mask (bool): If True, includes a mask channel in the input to handle padding. + model_channels (int): Base number of channels used throughout the model. + num_blocks (int): Number of transformer blocks. + num_heads (int): Number of heads in the multi-head attention layers. + mlp_ratio (float): Expansion ratio for MLP blocks. + crossattn_emb_channels (int): Number of embedding channels for cross-attention. + extra_image_context_dim (int): Number of embedding channels for extra image context. + pos_emb_cls (str): Type of positional embeddings. + pos_emb_learnable (bool): Whether positional embeddings are learnable. + pos_emb_interpolation (str): Method for interpolating positional embeddings. + min_fps (int): Minimum frames per second. + max_fps (int): Maximum frames per second. + use_adaln_lora (bool): Whether to use AdaLN-LoRA. + adaln_lora_dim (int): Dimension for AdaLN-LoRA. + rope_h_extrapolation_ratio (float): Height extrapolation ratio for RoPE. + rope_w_extrapolation_ratio (float): Width extrapolation ratio for RoPE. + rope_t_extrapolation_ratio (float): Temporal extrapolation ratio for RoPE. + extra_per_block_abs_pos_emb (bool): Whether to use extra per-block absolute positional embeddings. + extra_h_extrapolation_ratio (float): Height extrapolation ratio for extra embeddings. + extra_w_extrapolation_ratio (float): Width extrapolation ratio for extra embeddings. + extra_t_extrapolation_ratio (float): Temporal extrapolation ratio for extra embeddings. + n_dense_blocks (`int`, *optional*, defaults to -1): + Number of blocks that will remain dense (not replaced with sparse attention) + If -1, no blocks are replaced with sparse attention + If 0, all blocks use sparse attention + Otherwise, n_dense_blocks blocks will remain dense, distributed evenly across the network + natten_parameters (`dict`, *optional*, defaults to None): + NATTEN (Sparse attention) parameter list. + The list length must be the same as the number of layers, with each list element + indicating NATTEN parameters for that layer. If None, NATTEN will not be used in that + layer and it would remain a full dense self attention. If not None, it must be a + dictionary/mapping with at least the following key: + - window_size: `tuple` of size 3 indicating neighborhood attention window size. + window size of -1 along any dimension means self attention. + Other optional parameters and their keys: + - stride: `tuple` of size 3 indicating neighborhood attention stride value. + stride = 1 is standard neighborhood attention, stride = window size means + blocked/window self attention (WSA) along that dimension. Any other values are + strided neighborhood attention. Refer to the GNA paper for more information. + + - dilation: `tuple` of size 3 indicating neighborhood attention dilation value. + dilation = 1 is standard neighborhood attention. Refer to the DiNAT paper for more + information. + + - is_causal: `tuple` of 3 booleans indicating whether causal masking is enabled for + any of the T, H, W dimensions. + """ + + def __init__( + self, + max_img_h: int, + max_img_w: int, + max_frames: int, + in_channels: int, + out_channels: int, + patch_spatial: tuple, + patch_temporal: int, + concat_padding_mask: bool = True, + # attention settings + model_channels: int = 768, + num_blocks: int = 10, + num_heads: int = 16, + mlp_ratio: float = 4.0, + atten_backend: str = "transformer_engine", + # cross attention settings + crossattn_emb_channels: int = 1024, + use_crossattn_projection: bool = False, + crossattn_proj_in_channels: int = 1024, + extra_image_context_dim: Optional[int] = None, + # positional embedding settings + pos_emb_cls: str = "sincos", + pos_emb_learnable: bool = False, + pos_emb_interpolation: str = "crop", + min_fps: int = 1, # 1 for getty video + max_fps: int = 30, # 120 for getty video but let's use 30 + use_adaln_lora: bool = False, + adaln_lora_dim: int = 256, + rope_h_extrapolation_ratio: float = 1.0, + rope_w_extrapolation_ratio: float = 1.0, + rope_t_extrapolation_ratio: float = 1.0, + extra_per_block_abs_pos_emb: bool = False, + extra_h_extrapolation_ratio: float = 1.0, + extra_w_extrapolation_ratio: float = 1.0, + extra_t_extrapolation_ratio: float = 1.0, + rope_enable_fps_modulation: bool = True, + sac_config: SACConfig = SACConfig(), + n_dense_blocks: int = -1, + natten_parameters: Union[dict, list] = None, + # if True, will closely match wan's strategy to use fp32 in certain layers/operations + use_wan_fp32_strategy: bool = False, + ) -> None: + super().__init__() + self.max_img_h = max_img_h + self.max_img_w = max_img_w + self.max_frames = max_frames + self.in_channels = in_channels + self.out_channels = out_channels + self.patch_spatial = patch_spatial + self.patch_temporal = patch_temporal + self.num_heads = num_heads + self.num_blocks = num_blocks + self.model_channels = model_channels + self.concat_padding_mask = concat_padding_mask + self.atten_backend = atten_backend + # positional embedding settings + self.pos_emb_cls = pos_emb_cls + self.pos_emb_learnable = pos_emb_learnable + self.pos_emb_interpolation = pos_emb_interpolation + self.min_fps = min_fps + self.max_fps = max_fps + self.rope_h_extrapolation_ratio = rope_h_extrapolation_ratio + self.rope_w_extrapolation_ratio = rope_w_extrapolation_ratio + self.rope_t_extrapolation_ratio = rope_t_extrapolation_ratio + self.extra_per_block_abs_pos_emb = extra_per_block_abs_pos_emb + self.extra_h_extrapolation_ratio = extra_h_extrapolation_ratio + self.extra_w_extrapolation_ratio = extra_w_extrapolation_ratio + self.extra_t_extrapolation_ratio = extra_t_extrapolation_ratio + self.rope_enable_fps_modulation = rope_enable_fps_modulation + self.extra_image_context_dim = extra_image_context_dim + self.build_patch_embed() + self.build_pos_embed() + self.use_adaln_lora = use_adaln_lora + self.adaln_lora_dim = adaln_lora_dim + self.t_embedder = nn.Sequential( + Timesteps(model_channels), + TimestepEmbedding(model_channels, model_channels, use_adaln_lora=use_adaln_lora), + ) + self.use_crossattn_projection = use_crossattn_projection + self.crossattn_proj_in_channels = crossattn_proj_in_channels + self.use_wan_fp32_strategy = use_wan_fp32_strategy + + self.blocks = nn.ModuleList( + [ + Block( + x_dim=model_channels, + context_dim=crossattn_emb_channels, + num_heads=num_heads, + mlp_ratio=mlp_ratio, + use_adaln_lora=use_adaln_lora, + adaln_lora_dim=adaln_lora_dim, + backend=atten_backend, + image_context_dim=None if extra_image_context_dim is None else model_channels, + use_wan_fp32_strategy=use_wan_fp32_strategy, + ) + for _ in range(num_blocks) + ] + ) + + self.final_layer = FinalLayer( + hidden_size=self.model_channels, + spatial_patch_size=self.patch_spatial, + temporal_patch_size=self.patch_temporal, + out_channels=self.out_channels, + use_adaln_lora=self.use_adaln_lora, + adaln_lora_dim=self.adaln_lora_dim, + use_wan_fp32_strategy=self.use_wan_fp32_strategy, + ) + + self.t_embedding_norm = te.pytorch.RMSNorm(model_channels, eps=1e-6) + if extra_image_context_dim is not None: + self.img_context_proj = nn.Sequential( + nn.Linear( + extra_image_context_dim, model_channels, bias=True + ), # help distinguish between image and video context + nn.GELU(), + ) + + if use_crossattn_projection: + self.crossattn_proj = nn.Sequential( + nn.Linear(crossattn_proj_in_channels, crossattn_emb_channels, bias=True), + nn.GELU(), + ) + + self.init_weights() + self.enable_selective_checkpoint(sac_config, self.blocks) + + # Replace self-attention with sparse attention if specified + if n_dense_blocks != -1: + self = replace_selfattn_op_with_sparse_attn_op(self, n_dense_blocks, natten_parameters=natten_parameters) + + self._is_context_parallel_enabled = False + + # self.freeze_parameters() + + def freeze_parameters(self): + for name, module in self.named_modules(): + if any(keyword in name for keyword in ["cam_encoder", "self_attn"]): + for param in module.parameters(): + param.requires_grad = True + else: + for param in module.parameters(): + param.requires_grad = False + + def init_weights(self): + self.x_embedder.init_weights() + self.pos_embedder.reset_parameters() + if self.extra_per_block_abs_pos_emb: + self.extra_pos_embedder.reset_parameters() + + self.t_embedder[1].init_weights() + for block in self.blocks: + block.init_weights() + + self.final_layer.init_weights() + self.t_embedding_norm.reset_parameters() + + if self.extra_image_context_dim is not None: + self.img_context_proj[0].reset_parameters() + + def build_patch_embed(self): + ( + concat_padding_mask, + in_channels, + patch_spatial, + patch_temporal, + model_channels, + ) = ( + self.concat_padding_mask, + self.in_channels, + self.patch_spatial, + self.patch_temporal, + self.model_channels, + ) + in_channels = in_channels + 1 if concat_padding_mask else in_channels + self.x_embedder = PatchEmbed( + spatial_patch_size=patch_spatial, + temporal_patch_size=patch_temporal, + in_channels=in_channels, + out_channels=model_channels, + ) + + def build_pos_embed(self): + if self.pos_emb_cls == "rope3d": + cls_type = VideoRopePosition3DEmb + else: + raise ValueError(f"Unknown pos_emb_cls {self.pos_emb_cls}") + + log.debug(f"Building positional embedding with {self.pos_emb_cls} class, impl {cls_type}") + kwargs = dict( + model_channels=self.model_channels, + len_h=self.max_img_h // self.patch_spatial, + len_w=self.max_img_w // self.patch_spatial, + len_t=self.max_frames // self.patch_temporal, + max_fps=self.max_fps, + min_fps=self.min_fps, + is_learnable=self.pos_emb_learnable, + interpolation=self.pos_emb_interpolation, + head_dim=self.model_channels // self.num_heads, + h_extrapolation_ratio=self.rope_h_extrapolation_ratio, + w_extrapolation_ratio=self.rope_w_extrapolation_ratio, + t_extrapolation_ratio=self.rope_t_extrapolation_ratio, + enable_fps_modulation=self.rope_enable_fps_modulation, + ) + self.pos_embedder = cls_type( + **kwargs, + ) + + if self.extra_per_block_abs_pos_emb: + kwargs["h_extrapolation_ratio"] = self.extra_h_extrapolation_ratio + kwargs["w_extrapolation_ratio"] = self.extra_w_extrapolation_ratio + kwargs["t_extrapolation_ratio"] = self.extra_t_extrapolation_ratio + self.extra_pos_embedder = LearnablePosEmbAxis( + **kwargs, + ) + + def prepare_embedded_sequence( + self, + x_B_C_T_H_W: torch.Tensor, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[torch.Tensor]]: + """ + Prepares an embedded sequence tensor by applying positional embeddings and handling padding masks. + + Args: + x_B_C_T_H_W (torch.Tensor): video + fps (Optional[torch.Tensor]): Frames per second tensor to be used for positional embedding when required. + If None, a default value (`self.base_fps`) will be used. + padding_mask (Optional[torch.Tensor]): current it is not used + + Returns: + Tuple[torch.Tensor, Optional[torch.Tensor]]: + - A tensor of shape (B, T, H, W, D) with the embedded sequence. + - An optional positional embedding tensor, returned only if the positional embedding class + (`self.pos_emb_cls`) includes 'rope'. Otherwise, None. + + Notes: + - If `self.concat_padding_mask` is True, a padding mask channel is concatenated to the input tensor. + - The method of applying positional embeddings depends on the value of `self.pos_emb_cls`. + - If 'rope' is in `self.pos_emb_cls` (case insensitive), the positional embeddings are generated using + the `self.pos_embedder` with the shape [T, H, W]. + - If "fps_aware" is in `self.pos_emb_cls`, the positional embeddings are generated using the + `self.pos_embedder` with the fps tensor. + - Otherwise, the positional embeddings are generated without considering fps. + """ + if self.concat_padding_mask: + padding_mask = transforms.functional.resize( + padding_mask, list(x_B_C_T_H_W.shape[-2:]), interpolation=transforms.InterpolationMode.NEAREST + ) + x_B_C_T_H_W = torch.cat( + [x_B_C_T_H_W, padding_mask.unsqueeze(1).repeat(1, 1, x_B_C_T_H_W.shape[2], 1, 1)], dim=1 + ) + x_B_T_H_W_D = self.x_embedder(x_B_C_T_H_W) + + if self.extra_per_block_abs_pos_emb: + extra_pos_emb = self.extra_pos_embedder(x_B_T_H_W_D, fps=fps) + else: + extra_pos_emb = None + + if "rope" in self.pos_emb_cls.lower(): + return x_B_T_H_W_D, self.pos_embedder(x_B_T_H_W_D, fps=fps), extra_pos_emb + x_B_T_H_W_D = x_B_T_H_W_D + self.pos_embedder(x_B_T_H_W_D) # [B, T, H, W, D] + + return x_B_T_H_W_D, None, extra_pos_emb + + def unpatchify(self, x_B_T_H_W_M): + x_B_C_Tt_Hp_Wp = rearrange( + x_B_T_H_W_M, + "B T H W (p1 p2 t C) -> B C (T t) (H p1) (W p2)", + p1=self.patch_spatial, + p2=self.patch_spatial, + t=self.patch_temporal, + ) + return x_B_C_Tt_Hp_Wp + + def forward( + self, + x_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + crossattn_emb: torch.Tensor, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + data_type: Optional[DataType] = DataType.VIDEO, + intermediate_feature_ids: Optional[List[int]] = None, + img_context_emb: Optional[torch.Tensor] = None, + camera: Optional[torch.Tensor] = None, + ) -> torch.Tensor | List[torch.Tensor] | Tuple[torch.Tensor, List[torch.Tensor]]: + """ + Args: + x: (B, C, T, H, W) tensor of spatial-temp inputs + timesteps: (B, ) tensor of timesteps + crossattn_emb: (B, N, D) tensor of cross-attention embeddings + """ + assert isinstance(data_type, DataType), ( + f"Expected DataType, got {type(data_type)}. We need discuss this flag later." + ) + x_B_T_H_W_D, rope_emb_L_1_1_D, extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D = self.prepare_embedded_sequence( + x_B_C_T_H_W, + fps=fps, + padding_mask=padding_mask, + ) + + if self.use_crossattn_projection: + crossattn_emb = self.crossattn_proj(crossattn_emb) + + if img_context_emb is not None: + assert self.extra_image_context_dim is not None, ( + "extra_image_context_dim must be set if img_context_emb is provided" + ) + img_context_emb = self.img_context_proj(img_context_emb) + context_input = (crossattn_emb, img_context_emb) + else: + context_input = crossattn_emb + + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if timesteps_B_T.ndim == 1: + timesteps_B_T = timesteps_B_T.unsqueeze(1) + t_embedding_B_T_D, adaln_lora_B_T_3D = self.t_embedder(timesteps_B_T) + t_embedding_B_T_D = self.t_embedding_norm(t_embedding_B_T_D) + + # for logging purpose + affline_scale_log_info = {} + affline_scale_log_info["t_embedding_B_T_D"] = t_embedding_B_T_D.detach() + self.affline_scale_log_info = affline_scale_log_info + self.affline_emb = t_embedding_B_T_D + self.crossattn_emb = crossattn_emb + + if extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D is not None: + assert x_B_T_H_W_D.shape == extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D.shape, ( + f"{x_B_T_H_W_D.shape} != {extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D.shape}" + ) + + B, T, H, W, D = x_B_T_H_W_D.shape + # x_B_THW_D = rearrange(x_B_T_H_W_D, "b t h w d -> b (t h w) d") + + intermediate_features_outputs = [] + for i, block in enumerate(self.blocks): + x_B_T_H_W_D = block( + x_B_T_H_W_D, + t_embedding_B_T_D, + context_input, + rope_emb_L_1_1_D=rope_emb_L_1_1_D, + adaln_lora_B_T_3D=adaln_lora_B_T_3D, + extra_per_block_pos_emb=extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D, + camera=camera, + ) + if intermediate_feature_ids and i in intermediate_feature_ids: + x_reshaped_for_disc = rearrange(x_B_T_H_W_D, "b tp hp wp d -> b (tp hp wp) d") + intermediate_features_outputs.append(x_reshaped_for_disc) + + # x_B_T_H_W_D = rearrange(x_B_THW_D, "b (t h w) d -> b t h w d", t=T, h=H, w=W) + # O = out_channels * spatial_patch_size * spatial_patch_size * temporal_patch_size + x_B_T_H_W_O = self.final_layer(x_B_T_H_W_D, t_embedding_B_T_D, adaln_lora_B_T_3D=adaln_lora_B_T_3D) + x_B_C_Tt_Hp_Wp = self.unpatchify(x_B_T_H_W_O) + if intermediate_feature_ids: + if len(intermediate_features_outputs) != len(intermediate_feature_ids): + log.warning( + f"Collected {len(intermediate_features_outputs)} intermediate features, " + f"but expected {len(intermediate_feature_ids)}. " + f"Requested IDs: {intermediate_feature_ids}" + ) + return x_B_C_Tt_Hp_Wp, intermediate_features_outputs + + return x_B_C_Tt_Hp_Wp + + def enable_selective_checkpoint(self, sac_config: SACConfig, blocks: nn.ModuleList): + if sac_config.mode == CheckpointMode.NONE: + return self + + log.info( + f"Enable selective checkpoint with {sac_config.mode}, for every {sac_config.every_n_blocks} blocks. Total blocks: {len(blocks)}" + ) + _context_fn = sac_config.get_context_fn() + for block_id, block in blocks.named_children(): + if int(block_id) % sac_config.every_n_blocks == 0: + log.info(f"Enable selective checkpoint for block {block_id}") + block = ptd_checkpoint_wrapper( + block, + context_fn=_context_fn, + preserve_rng_state=False, + ) + blocks.register_module(block_id, block) + self.register_module( + "final_layer", + ptd_checkpoint_wrapper( + self.final_layer, + context_fn=_context_fn, + preserve_rng_state=False, + ), + ) + + return self + + def fully_shard(self, mesh): + for i, block in enumerate(self.blocks): + reshard_after_forward = i < len(self.blocks) - 1 + fully_shard(block, mesh=mesh, reshard_after_forward=reshard_after_forward) + + fully_shard(self.final_layer, mesh=mesh, reshard_after_forward=True) + if self.extra_per_block_abs_pos_emb: + fully_shard(self.extra_pos_embedder, mesh=mesh, reshard_after_forward=True) + fully_shard(self.t_embedder, mesh=mesh, reshard_after_forward=False) + if self.extra_image_context_dim is not None: + fully_shard(self.img_context_proj, mesh=mesh, reshard_after_forward=False) + + def disable_context_parallel(self): + # pos_embedder + self.pos_embedder.disable_context_parallel() + if self.extra_per_block_abs_pos_emb: + self.extra_pos_embedder.disable_context_parallel() + + # attention + for block in self.blocks: + block.set_context_parallel_group( + process_group=None, + ranks=None, + stream=torch.cuda.Stream(), + ) + + self._is_context_parallel_enabled = False + + def enable_context_parallel(self, process_group: Optional[ProcessGroup] = None): + # pos_embedder + self.pos_embedder.enable_context_parallel(process_group=process_group) + if self.extra_per_block_abs_pos_emb: + self.extra_pos_embedder.enable_context_parallel(process_group=process_group) + + # attention + cp_ranks = get_process_group_ranks(process_group) + for block in self.blocks: + block.set_context_parallel_group( + process_group=process_group, + ranks=cp_ranks, + stream=torch.cuda.Stream(), + ) + + self._is_context_parallel_enabled = True + + @property + def is_context_parallel_enabled(self): + return self._is_context_parallel_enabled + + +class CameraMiniTrainDITwithConditionalMask(CameraMiniTrainDIT): + def __init__(self, *args, timestep_scale: float = 1.0, **kwargs): + assert "in_channels" in kwargs, "in_channels must be provided" + kwargs["in_channels"] += 1 + self.timestep_scale = timestep_scale + super().__init__(*args, **kwargs) + + def forward( + self, + x_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + crossattn_emb: torch.Tensor, + condition_video_input_mask_B_C_T_H_W: Optional[torch.Tensor] = None, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + data_type: Optional[DataType] = DataType.VIDEO, + img_context_emb: Optional[torch.Tensor] = None, + camera: Optional[torch.Tensor] = None, + **kwargs, + ) -> torch.Tensor | List[torch.Tensor] | Tuple[torch.Tensor, List[torch.Tensor]]: + del kwargs + if data_type == DataType.VIDEO: + x_B_C_T_H_W = torch.cat([x_B_C_T_H_W, condition_video_input_mask_B_C_T_H_W.type_as(x_B_C_T_H_W)], dim=1) + else: + B, _, T, H, W = x_B_C_T_H_W.shape + x_B_C_T_H_W = torch.cat( + [x_B_C_T_H_W, torch.zeros((B, 1, T, H, W), dtype=x_B_C_T_H_W.dtype, device=x_B_C_T_H_W.device)], dim=1 + ) + return super().forward( + x_B_C_T_H_W=x_B_C_T_H_W, + timesteps_B_T=timesteps_B_T * self.timestep_scale, + crossattn_emb=crossattn_emb, + fps=fps, + padding_mask=padding_mask, + data_type=data_type, + img_context_emb=img_context_emb, + camera=camera, + ) + + +def replace_selfattn_op_with_sparse_attn_op( + model: CameraMiniTrainDIT, n_dense_blocks: int = 0, natten_parameters: Union[dict, list] = None +) -> CameraMiniTrainDIT: + """ + Replace the self-attention operator with a sparse self-attention operator. + + Args: + model: CameraMiniTrainDIT instance + n_dense_blocks: Number of blocks that will remain dense (not replaced with NeighborhoodAttention) + If 0, all blocks use NeighborhoodAttention. + If -1, return model directly without any modifications. + Otherwise, n_dense_blocks blocks will remain dense, distributed evenly across the network. + + Returns: + Modified instance + """ + # Special case: return model directly without modifications + if n_dense_blocks == -1: + return model + + num_blocks = len(model.blocks) + + if natten_parameters is None: + raise ValueError("Please specify natten_parameters when n_dense_blocks > -1.") + + if isinstance(natten_parameters, Sequence) and len(natten_parameters) != num_blocks: + raise ValueError( + "List of NATTEN parameters must be the same length as the number of blocks, " + f"got {len(natten_parameters)=} != {num_blocks=}." + ) + + if isinstance(natten_parameters, Sequence) and n_dense_blocks > 0: + log.warning(f"NATTEN parameters was a list; ignoring {n_dense_blocks=}.") + + if isinstance(natten_parameters, Sequence): + natten_parameters_list = natten_parameters + else: + if n_dense_blocks >= num_blocks: + raise ValueError(f"n_dense_blocks ({n_dense_blocks}) must be less than the number of blocks ({num_blocks})") + + # Determine which blocks should remain dense + dense_indices = set() + + if n_dense_blocks > 0: + # General rule: distribute n_dense_blocks blocks evenly across the network + if n_dense_blocks == 1: + # Special case: just the middle block + dense_indices.add(num_blocks // 2) + else: + # For multiple blocks, distribute them evenly from start to end + indices = np.linspace(0, num_blocks - 1, n_dense_blocks, dtype=int) + dense_indices.update(indices.tolist()) + + natten_parameters_list = [None if i in dense_indices else natten_parameters for i in range(num_blocks)] + + # Replace self-attention with NeighborhoodAttention for non-dense blocks + for i, block in enumerate(model.blocks): + natten_params = natten_parameters_list[i] + if natten_params is not None: + natten_parameters_layer = {k: v for k, v in natten_params.items()} + natten_parameters_layer["layer_id"] = i + if block.self_attn.backend == "minimal_a2a": + sparse_attn_op = NattenA2AAttnOp(natten_parameters=natten_parameters_layer) + else: + raise NotImplementedError( + f"Using sparsity with attention backend {block.self_attn.backend} is not supported." + ) + + block.self_attn.register_module("attn_op", sparse_attn_op) + + return model diff --git a/REGEN-main/cosmos_policy/_src/predict2/camera/networks/dit_multiview_camera_ar.py b/REGEN-main/cosmos_policy/_src/predict2/camera/networks/dit_multiview_camera_ar.py new file mode 100644 index 0000000000000000000000000000000000000000..5f5b25c8f3e95782efe4e48a3f756f0ac3204552 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/camera/networks/dit_multiview_camera_ar.py @@ -0,0 +1,1872 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import collections +import math +from collections import namedtuple +from collections.abc import Sequence +from dataclasses import dataclass +from enum import Enum +from typing import List, Optional, Tuple, Union + +import numpy as np +import torch +import torch.amp as amp +import transformer_engine as te +from einops import rearrange, repeat +from einops.layers.torch import Rearrange +from torch import nn +from torch.distributed import ProcessGroup, get_process_group_ranks +from torch.distributed._composable.fsdp import fully_shard +from torch.distributed.algorithms._checkpoint.checkpoint_wrapper import checkpoint_wrapper as ptd_checkpoint_wrapper + +try: + from torch.utils.checkpoint import CheckpointPolicy, create_selective_checkpoint_contexts +except ImportError: + CheckpointPolicy = None + +from packaging.version import Version +from torchvision import transforms + +if Version(te.__version__) >= Version("2.8.0"): + from transformer_engine.pytorch.attention.rope import apply_rotary_pos_emb +else: + from transformer_engine.pytorch.attention import apply_rotary_pos_emb + + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.context_parallel import split_inputs_cp +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.modules.neighborhood_attn import NeighborhoodAttention +from cosmos_policy._src.predict2.networks.a2a_cp import MinimalA2AAttnOp, NattenA2AAttnOp +from cosmos_policy._src.predict2.networks.model_weights_stats import WeightTrainingStat +from cosmos_policy._src.predict2.networks.selective_activation_checkpoint import SACConfig as _SACConfig + + +# selective activation checkpoint; only apply to the minimal v4 model. if there are change in the networks, some policy will not work as we expect. +def predict2_2B_720_context_fn(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + op_count_key = f"{mode}_mm_count" + # from cosmos_policy._src.imaginaire.utils import log + # log.info(f"op_count_key: {op_count_key}, op_count[op_count_key]: {op_count[op_count_key]}, {args[0].shape}, {args[1].shape}") + # there are totally 6 + 4 + 4 + 2 = 16 block + op_count[op_count_key] = (op_count[op_count_key] + 1) % 16 + if op_count[op_count_key] > 8: # recompute self attn first 3 linear layers + return CheckpointPolicy.MUST_SAVE + if "flash_attn" in str(func): + op_count_key = f"{mode}_flash_attn_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 2 + if op_count[op_count_key]: + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_2B_720_context_fn_aggressive(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + # The default policy is to recompute everything. This is the most memory-efficient + # starting point. We then selectively choose what to save. + default_policy = CheckpointPolicy.PREFER_RECOMPUTE + + # Save the output of Flash Attention. This is the most computationally + # expensive part of a transformer block. Saving its output provides a + # good balance between memory savings and computational overhead. + if "flash_attn" in str(func): + return CheckpointPolicy.MUST_SAVE + + # All other operations (e.g., torch.ops.aten.mm.default, layer norms, additions) + # will fall through to the default policy and be recomputed. + return default_policy + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_2B_720_context_fn_aggressive_v2(): + """ + The most memory-aggressive checkpointing policy. Recomputes ALL operations. + """ + + def policy_fn(ctx, func, *args, **kwargs): + # The policy is to always recompute everything. + # This saves the maximum amount of memory but incurs the highest + # computational cost during the backward pass. + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_14B_720_context_fn(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + op_count_key = f"{mode}_mm_count" + # from cosmos_policy._src.imaginaire.utils import log + # log.info(f"op_count_key: {op_count_key}, op_count[op_count_key]: {op_count[op_count_key]}, {args[0].shape}, {args[1].shape}") + # there are totally 6 + 4 + 4 + 2 = 16 block + op_count[op_count_key] = (op_count[op_count_key] + 1) % 16 + if op_count[op_count_key] > 8: # recompute self attn first 1 linear layers + return CheckpointPolicy.MUST_SAVE + if "flash_attn" in str(func): + op_count_key = f"{mode}_flash_attn_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 2 + if op_count[op_count_key]: + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_14B_720_context_fn_aggressive(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + op_count_key = f"{mode}_mm_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 16 + if op_count[op_count_key] > 12: # recompute self attn first 1 linear layers + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def linear_selfattn_context_fn(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + return CheckpointPolicy.MUST_SAVE + if "flash_attn" in str(func): + op_count_key = f"{mode}_flash_attn_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 2 + if op_count[op_count_key]: + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +class CheckpointMode(str, Enum): + NONE = "none" + MM_ONLY = "mm_only" + BLOCK_WISE = "block_wise" + LINEAR_SELFATTN = "linear_selfattn" + PREDICT2_2B_720 = "predict2_2b_720" + PREDICT2_14B_720 = "predict2_14b_720" + PREDICT2_2B_720_AGGRESSIVE = "predict2_2b_720_aggressive" + PREDICT2_2B_720_AGGRESSIVE_V2 = "predict2_2b_720_aggressive_v2" + PREDICT2_14B_720_AGGRESSIVE = "predict2_14b_720_aggressive" + + def __str__(self) -> str: + return self.value + + +@dataclass +class SACConfig(_SACConfig): + def get_context_fn(self): + if self.mode == CheckpointMode.LINEAR_SELFATTN: + return linear_selfattn_context_fn + elif self.mode == CheckpointMode.PREDICT2_2B_720: + return predict2_2B_720_context_fn + elif self.mode == CheckpointMode.PREDICT2_2B_720_AGGRESSIVE: + return predict2_2B_720_context_fn_aggressive + elif self.mode == CheckpointMode.PREDICT2_2B_720_AGGRESSIVE_V2: + return predict2_2B_720_context_fn_aggressive_v2 + elif self.mode == CheckpointMode.PREDICT2_14B_720: + return predict2_14B_720_context_fn + elif self.mode == CheckpointMode.PREDICT2_14B_720_AGGRESSIVE: + return predict2_14B_720_context_fn_aggressive + else: + # Reuse parent class implementation for other modes + return super().get_context_fn() + + +VideoSize = namedtuple("VideoSize", ["T", "H", "W"]) + + +class RMSNorm(torch.nn.Module): + def __init__(self, dim: int, eps: float = 1e-5): + super().__init__() + self.eps = eps + self.weight = nn.Parameter(torch.ones(dim)) + + def reset_parameters(self): + torch.nn.init.ones_(self.weight) + + def _norm(self, x): + return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps) + + def forward(self, x: torch.Tensor) -> torch.Tensor: + output = self._norm(x.float()).type_as(x) + return output * self.weight + + +# ---------------------- Feed Forward Network ----------------------- +class GPT2FeedForward(nn.Module): + def __init__(self, d_model: int, d_ff: int): + super().__init__() + self.activation = nn.GELU() + self.layer1 = nn.Linear(d_model, d_ff, bias=False) + self.layer2 = nn.Linear(d_ff, d_model, bias=False) + + self._layer_id = None + self._dim = d_model + self._hidden_dim = d_ff + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self._dim) + torch.nn.init.trunc_normal_(self.layer1.weight, std=std, a=-3 * std, b=3 * std) + + # scale init by depth as in https://arxiv.org/abs/1908.11365 -- worked slightly better. + std = 1.0 / math.sqrt(self._hidden_dim) + if self._layer_id is not None: + std = std / math.sqrt(2 * (self._layer_id + 1)) + torch.nn.init.trunc_normal_(self.layer2.weight, std=std, a=-3 * std, b=3 * std) + + def forward(self, x: torch.Tensor): + x = self.layer1(x) + + x = self.activation(x) + x = self.layer2(x) + return x + + +def torch_attention_op(q_B_S_H_D, k_B_S_H_D, v_B_S_H_D): + """Computes multi-head attention using PyTorch's native implementation. + + This function provides a PyTorch backend alternative to Transformer Engine's attention operation. + It rearranges the input tensors to match PyTorch's expected format, computes scaled dot-product + attention, and rearranges the output back to the original format. + + The input tensor names use the following dimension conventions: + + - B: batch size + - S: sequence length + - H: number of attention heads + - D: head dimension + + Args: + q_B_S_H_D: Query tensor with shape (batch, seq_len, n_heads, head_dim) + k_B_S_H_D: Key tensor with shape (batch, seq_len, n_heads, head_dim) + v_B_S_H_D: Value tensor with shape (batch, seq_len, n_heads, head_dim) + + Returns: + Attention output tensor with shape (batch, seq_len, n_heads * head_dim) + """ + in_q_shape = q_B_S_H_D.shape + in_k_shape = k_B_S_H_D.shape + q_B_H_S_D = rearrange(q_B_S_H_D, "b ... h k -> b h ... k").view(in_q_shape[0], in_q_shape[-2], -1, in_q_shape[-1]) + k_B_H_S_D = rearrange(k_B_S_H_D, "b ... h v -> b h ... v").view(in_k_shape[0], in_k_shape[-2], -1, in_k_shape[-1]) + v_B_H_S_D = rearrange(v_B_S_H_D, "b ... h v -> b h ... v").view(in_k_shape[0], in_k_shape[-2], -1, in_k_shape[-1]) + result_B_S_HD = rearrange( + torch.nn.functional.scaled_dot_product_attention(q_B_H_S_D, k_B_H_S_D, v_B_H_S_D), "b h ... l -> b ... (h l)" + ) + + return result_B_S_HD + + +class Attention(nn.Module): + """ + A flexible attention module supporting both self-attention and cross-attention mechanisms. + + This module implements a multi-head attention layer that can operate in either self-attention + or cross-attention mode. The mode is determined by whether a context dimension is provided. + The implementation uses scaled dot-product attention and supports optional bias terms and + dropout regularization. + + Args: + query_dim (int): The dimensionality of the query vectors. + context_dim (int, optional): The dimensionality of the context (key/value) vectors. + If None, the module operates in self-attention mode using query_dim. Default: None + n_heads (int, optional): Number of attention heads for multi-head attention. Default: 8 + head_dim (int, optional): The dimension of each attention head. Default: 64 + dropout (float, optional): Dropout probability applied to the output. Default: 0.0 + qkv_format (str, optional): Format specification for QKV tensors. Default: "bshd" + backend (str, optional): Backend to use for the attention operation. Default: "transformer_engine" + + Examples: + >>> # Self-attention with 512 dimensions and 8 heads + >>> self_attn = Attention(query_dim=512) + >>> x = torch.randn(32, 16, 512) # (batch_size, seq_len, dim) + >>> out = self_attn(x) # (32, 16, 512) + + >>> # Cross-attention + >>> cross_attn = Attention(query_dim=512, context_dim=256) + >>> query = torch.randn(32, 16, 512) + >>> context = torch.randn(32, 8, 256) + >>> out = cross_attn(query, context) # (32, 16, 512) + """ + + def __init__( + self, + query_dim: int, + context_dim=None, + n_heads=8, + head_dim=64, + dropout=0.0, + qkv_format: str = "bshd", + backend: str = "transformer_engine", + use_wan_fp32_strategy: bool = False, + ) -> None: + super().__init__() + log.debug( + f"Setting up {self.__class__.__name__}. Query dim is {query_dim}, context_dim is {context_dim} and using " + f"{n_heads} heads with a dimension of {head_dim}." + ) + self.is_selfattn = context_dim is None # self attention + + assert backend in ["transformer_engine", "torch", "minimal_a2a"], f"Invalid backend: {backend}" + self.backend = backend + + context_dim = query_dim if context_dim is None else context_dim + inner_dim = head_dim * n_heads + + self.n_heads = n_heads + self.head_dim = head_dim + self.qkv_format = qkv_format + self.query_dim = query_dim + self.context_dim = context_dim + self.use_wan_fp32_strategy = use_wan_fp32_strategy + + self.q_proj = nn.Linear(query_dim, inner_dim, bias=False) + self.q_norm = te.pytorch.RMSNorm(self.head_dim, eps=1e-6) + + self.k_proj = nn.Linear(context_dim, inner_dim, bias=False) + self.k_norm = te.pytorch.RMSNorm(self.head_dim, eps=1e-6) + + self.v_proj = nn.Linear(context_dim, inner_dim, bias=False) + self.v_norm = nn.Identity() + + self.output_proj = nn.Linear(inner_dim, query_dim, bias=False) + self.output_dropout = nn.Dropout(dropout) if dropout > 1e-4 else nn.Identity() + + if self.backend == "transformer_engine": + from transformer_engine.pytorch.attention import DotProductAttention + + self.attn_op = DotProductAttention( + self.n_heads, + self.head_dim, + num_gqa_groups=self.n_heads, + attention_dropout=0, + qkv_format=qkv_format, + attn_mask_type="no_mask", + ) + elif self.backend == "minimal_a2a": + self.attn_op = MinimalA2AAttnOp() + elif self.backend == "torch": + self.attn_op = torch_attention_op + + self._query_dim = query_dim + self._context_dim = context_dim + self._inner_dim = inner_dim + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self._query_dim) + torch.nn.init.trunc_normal_(self.q_proj.weight, std=std, a=-3 * std, b=3 * std) + std = 1.0 / math.sqrt(self._context_dim) + torch.nn.init.trunc_normal_(self.k_proj.weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.v_proj.weight, std=std, a=-3 * std, b=3 * std) + + std = 1.0 / math.sqrt(self._inner_dim) + torch.nn.init.trunc_normal_(self.output_proj.weight, std=std, a=-3 * std, b=3 * std) + + for layer in self.q_norm, self.k_norm, self.v_norm: + if hasattr(layer, "reset_parameters"): + layer.reset_parameters() + + def compute_qkv(self, x, context=None, rope_emb=None) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor]: + q = self.q_proj(x) + context = x if context is None else context + k = self.k_proj(context) + v = self.v_proj(context) + q, k, v = map( + lambda t: rearrange(t, "b ... (h d) -> b ... h d", h=self.n_heads, d=self.head_dim), + (q, k, v), + ) + + def apply_norm_and_rotary_pos_emb(q, k, v, rope_emb): + q = self.q_norm(q) + k = self.k_norm(k) + v = self.v_norm(v) + if self.is_selfattn and rope_emb is not None: # only apply to self-attention! + if self.use_wan_fp32_strategy: # wan will force q and k to fp32 before rotary pos emb + q = q.to(torch.float32) + k = k.to(torch.float32) + q = apply_rotary_pos_emb(q, rope_emb, tensor_format=self.qkv_format, fused=True) + k = apply_rotary_pos_emb(k, rope_emb, tensor_format=self.qkv_format, fused=True) + return q, k, v + + q, k, v = apply_norm_and_rotary_pos_emb(q, k, v, rope_emb) + + return q, k, v + + def compute_attention(self, q, k, v, video_size: Optional[VideoSize] = None): + additional_args = {} + if isinstance(self.attn_op, (NattenA2AAttnOp, NeighborhoodAttention)): + additional_args["video_size"] = video_size + + result = self.attn_op(q, k, v, **additional_args) # [B, S, H, D] + return self.output_dropout(self.output_proj(result)) + + def forward( + self, + x, + context: Optional[torch.Tensor] = None, + rope_emb: Optional[torch.Tensor] = None, + video_size: Optional[VideoSize] = None, + ): + """ + Args: + x (Tensor): The query tensor of shape [B, Mq, K] + context (Optional[Tensor]): The key tensor of shape [B, Mk, K] or use x as context [self attention] if None + rope_emb (Optional[Tensor]): RoPE embedding tensor, or no RoPE embeddings (i.e. in cross attention) + video_size(VideoSize): Shape [T, H, W] + """ + q, k, v = self.compute_qkv(x, context, rope_emb=rope_emb) + return self.compute_attention(q, k, v, video_size=video_size) + + def set_context_parallel_group(self, process_group, ranks, stream): + # self.attn_op.set_context_parallel_group(process_group, ranks, stream, cp_comm_type="a2a") + self.attn_op.set_context_parallel_group(process_group, ranks, stream) + + +class I2VCrossAttention(Attention): + def __init__(self, *args, img_latent_dim: int = 1024, **kwargs): + super().__init__(*args, **kwargs) + inner_dim = self.head_dim * self.n_heads + self.k_img = nn.Linear(img_latent_dim, inner_dim, bias=False) + self.v_img = nn.Linear(img_latent_dim, inner_dim, bias=False) + self.k_img_norm = te.pytorch.RMSNorm(self.head_dim, eps=1e-6) + + def init_weights(self) -> None: + super().init_weights() + torch.nn.init.trunc_normal_(self.k_img.weight, std=1.0 / math.sqrt(self._inner_dim)) + torch.nn.init.trunc_normal_(self.v_img.weight, std=1.0 / math.sqrt(self._inner_dim)) + self.k_img_norm.reset_parameters() + + def compute_qkv( + self, x, context, rope_emb=None + ) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]: + text_context, img_context = context + q, k, v = super().compute_qkv(x, text_context, rope_emb) + k_img = self.k_img(img_context) + v_img = self.v_img(img_context) + # Rearrange k_img, v_img + k_img, v_img = map( + lambda t: rearrange(t, "b ... (h d) -> b ... h d", h=self.n_heads, d=self.head_dim), + (k_img, v_img), + ) + + return q, k, v, self.k_img_norm(k_img), v_img + + def compute_attention(self, q, k, v, k_img, v_img): + result = self.attn_op(q, k, v) # [B, S, H, D] + result_img = self.attn_op(q, k_img, v_img) + return self.output_dropout(self.output_proj(result + result_img)) + + def forward( + self, + x, + context=None, + rope_emb=None, + ): + q, k, v, k_img, v_img = self.compute_qkv(x, context, rope_emb) + return self.compute_attention(q, k, v, k_img, v_img) + + +class VideoPositionEmb(nn.Module): + def __init__(self): + super().__init__() + self._cp_group = None + + def enable_context_parallel(self, process_group: ProcessGroup): + self._cp_group = process_group + + def disable_context_parallel(self): + self._cp_group = None + + @property + def seq_dim(self): + return 1 + + def forward(self, x_B_T_H_W_C: torch.Tensor, fps=Optional[torch.Tensor]) -> torch.Tensor: + """ + With CP, the function assume that the input tensor is already split. + It delegates the embedding generation to generate_embeddings function. + """ + B_T_H_W_C = x_B_T_H_W_C.shape + if self._cp_group is not None: + cp_ranks = get_process_group_ranks(self._cp_group) + cp_size = len(cp_ranks) + B, T, H, W, C = B_T_H_W_C + B_T_H_W_C = (B, T * cp_size, H, W, C) + embeddings = self.generate_embeddings(B_T_H_W_C, fps=fps) + + return self._split_for_context_parallel(embeddings) + + def generate_embeddings(self, B_T_H_W_C: torch.Size, fps=Optional[torch.Tensor]): + raise NotImplementedError + + def _split_for_context_parallel(self, embeddings): + if self._cp_group is not None: + embeddings = split_inputs_cp(x=embeddings, seq_dim=self.seq_dim, cp_group=self._cp_group) + return embeddings + + +class VideoRopePosition3DEmb(VideoPositionEmb): + def __init__( + self, + *, # enforce keyword arguments + head_dim: int, + len_h: int, + len_w: int, + len_t: int, + base_fps: int = 24, + h_extrapolation_ratio: float = 1.0, + w_extrapolation_ratio: float = 1.0, + t_extrapolation_ratio: float = 1.0, + enable_fps_modulation: bool = True, + **kwargs, # used for compatibility with other positional embeddings; unused in this class + ): + del kwargs + super().__init__() + self.register_buffer("seq", torch.arange(max(len_h, len_w, len_t), dtype=torch.float)) + self.base_fps = base_fps + self.max_h = len_h + self.max_w = len_w + self.max_t = len_t + self.enable_fps_modulation = enable_fps_modulation + dim = head_dim + dim_h = dim // 6 * 2 + dim_w = dim_h + dim_t = dim - 2 * dim_h + assert dim == dim_h + dim_w + dim_t, f"bad dim: {dim} != {dim_h} + {dim_w} + {dim_t}" + + self.register_buffer( + "dim_spatial_range", + torch.arange(0, dim_h, 2)[: (dim_h // 2)].float() / dim_h, + persistent=True, + ) + self.register_buffer( + "dim_temporal_range", + torch.arange(0, dim_t, 2)[: (dim_t // 2)].float() / dim_t, + persistent=True, + ) + self._dim_h = dim_h + self._dim_t = dim_t + + self.h_ntk_factor = h_extrapolation_ratio ** (dim_h / (dim_h - 2)) + self.w_ntk_factor = w_extrapolation_ratio ** (dim_w / (dim_w - 2)) + self.t_ntk_factor = t_extrapolation_ratio ** (dim_t / (dim_t - 2)) + self.reset_parameters() + + def reset_parameters(self) -> None: + dim_h = self._dim_h + dim_t = self._dim_t + + self.seq = torch.arange(max(self.max_h, self.max_w, self.max_t)).float().to(self.dim_spatial_range.device) + self.dim_spatial_range = ( + torch.arange(0, dim_h, 2)[: (dim_h // 2)].float().to(self.dim_spatial_range.device) / dim_h + ) + self.dim_temporal_range = ( + torch.arange(0, dim_t, 2)[: (dim_t // 2)].float().to(self.dim_spatial_range.device) / dim_t + ) + + def generate_embeddings( + self, + B_T_H_W_C: torch.Size, + fps: Optional[torch.Tensor] = None, + h_ntk_factor: Optional[float] = None, + w_ntk_factor: Optional[float] = None, + t_ntk_factor: Optional[float] = None, + ): + """ + Generate embeddings for the given input size. + + Args: + B_T_H_W_C (torch.Size): Input tensor size (Batch, Time, Height, Width, Channels). + fps (Optional[torch.Tensor], optional): Frames per second. Defaults to None. + h_ntk_factor (Optional[float], optional): Height NTK factor. If None, uses self.h_ntk_factor. + w_ntk_factor (Optional[float], optional): Width NTK factor. If None, uses self.w_ntk_factor. + t_ntk_factor (Optional[float], optional): Time NTK factor. If None, uses self.t_ntk_factor. + + Returns: + Not specified in the original code snippet. + """ + h_ntk_factor = h_ntk_factor if h_ntk_factor is not None else self.h_ntk_factor + w_ntk_factor = w_ntk_factor if w_ntk_factor is not None else self.w_ntk_factor + t_ntk_factor = t_ntk_factor if t_ntk_factor is not None else self.t_ntk_factor + + h_theta = 10000.0 * h_ntk_factor + w_theta = 10000.0 * w_ntk_factor + t_theta = 10000.0 * t_ntk_factor + + h_spatial_freqs = 1.0 / (h_theta ** self.dim_spatial_range.float()) + w_spatial_freqs = 1.0 / (w_theta ** self.dim_spatial_range.float()) + temporal_freqs = 1.0 / (t_theta ** self.dim_temporal_range.float()) + + B, T, H, W, _ = B_T_H_W_C + assert H <= self.max_h and W <= self.max_w, ( + f"Input dimensions (H={H}, W={W}) exceed the maximum dimensions (max_h={self.max_h}, max_w={self.max_w})" + ) + half_emb_h = torch.outer(self.seq[:H], h_spatial_freqs) + half_emb_w = torch.outer(self.seq[:W], w_spatial_freqs) + + if self.enable_fps_modulation: + uniform_fps = (fps is None) or (fps.min() == fps.max()) + assert uniform_fps or B == 1 or T == 1, ( + "For video batch, batch size should be 1 for non-uniform fps. For image batch, T should be 1" + ) + + # apply sequence scaling in temporal dimension + if fps is None: # image case + assert T == 1, "T should be 1 for image batch." + half_emb_t = torch.outer(self.seq[:T], temporal_freqs) + else: + half_emb_t = torch.outer(self.seq[:T] / fps[:1] * self.base_fps, temporal_freqs) + else: + half_emb_t = torch.outer(self.seq[:T], temporal_freqs) + + em_T_H_W_D = torch.cat( + [ + repeat(half_emb_t, "t d -> t h w d", h=H, w=W), + repeat(half_emb_h, "h d -> t h w d", t=T, w=W), + repeat(half_emb_w, "w d -> t h w d", t=T, h=H), + ] + * 2, + dim=-1, + ) + + return rearrange(em_T_H_W_D, "t h w d -> (t h w) 1 1 d").float() + + @property + def seq_dim(self): + return 0 + + +class LearnablePosEmbAxis(VideoPositionEmb): + def __init__( + self, + *, # enforce keyword arguments + interpolation: str, + model_channels: int, + len_h: int, + len_w: int, + len_t: int, + **kwargs, + ): + """ + Args: + interpolation (str): we curretly only support "crop", ideally when we need extrapolation capacity, we should adjust frequency or other more advanced methods. they are not implemented yet. + """ + del kwargs # unused + super().__init__() + self.interpolation = interpolation + assert self.interpolation in ["crop"], f"Unknown interpolation method {self.interpolation}" + self.model_channels = model_channels + + self.pos_emb_h = nn.Parameter(torch.zeros(len_h, model_channels)) + self.pos_emb_w = nn.Parameter(torch.zeros(len_w, model_channels)) + self.pos_emb_t = nn.Parameter(torch.zeros(len_t, model_channels)) + + self.reset_parameters() + + def reset_parameters(self): + std = 1.0 / math.sqrt(self.model_channels) + torch.nn.init.trunc_normal_(self.pos_emb_h, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.pos_emb_w, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.pos_emb_t, std=std, a=-3 * std, b=3 * std) + + def generate_embeddings(self, B_T_H_W_C: torch.Size, fps=Optional[torch.Tensor]) -> torch.Tensor: + B, T, H, W, _ = B_T_H_W_C + if self.interpolation == "crop": + emb_h_H = self.pos_emb_h[:H] + emb_w_W = self.pos_emb_w[:W] + emb_t_T = self.pos_emb_t[:T] + emb = ( + repeat(emb_t_T, "t d-> b t h w d", b=B, h=H, w=W) + + repeat(emb_h_H, "h d-> b t h w d", b=B, t=T, w=W) + + repeat(emb_w_W, "w d-> b t h w d", b=B, t=T, h=H) + ) + assert list(emb.shape)[:4] == [B, T, H, W], f"bad shape: {list(emb.shape)[:4]} != {B, T, H, W}" + else: + raise ValueError(f"Unknown interpolation method {self.interpolation}") + + norm = torch.linalg.vector_norm(emb, dim=-1, keepdim=True, dtype=torch.float32) + norm = torch.add(1e-6, norm, alpha=np.sqrt(norm.numel() / emb.numel())) + return emb / norm.to(emb.dtype) + + +def modulate(x, shift, scale): + return x * (1 + scale) + shift + + +class Timesteps(nn.Module): + def __init__(self, num_channels): + super().__init__() + self.num_channels = num_channels + + def forward(self, timesteps_B_T): + assert timesteps_B_T.ndim == 2, f"Expected 2D input, got {timesteps_B_T.ndim}" + # wan need emb to be in fp32 + in_dype = timesteps_B_T.dtype + timesteps = timesteps_B_T.flatten().float() + half_dim = self.num_channels // 2 + exponent = -math.log(10000) * torch.arange(half_dim, dtype=torch.float32, device=timesteps.device) + exponent = exponent / (half_dim - 0.0) + + emb = torch.exp(exponent) + emb = timesteps[:, None].float() * emb[None, :] + + sin_emb = torch.sin(emb) + cos_emb = torch.cos(emb) + emb = torch.cat([cos_emb, sin_emb], dim=-1) + + return rearrange(emb.to(dtype=in_dype), "(b t) d -> b t d", b=timesteps_B_T.shape[0], t=timesteps_B_T.shape[1]) + + +class TimestepEmbedding(nn.Module): + def __init__(self, in_features: int, out_features: int, use_adaln_lora: bool = False): + super().__init__() + log.debug( + f"Using AdaLN LoRA Flag: {use_adaln_lora}. We enable bias if no AdaLN LoRA for backward compatibility." + ) + self.in_dim = in_features + self.out_dim = out_features + self.linear_1 = nn.Linear(in_features, out_features, bias=not use_adaln_lora) + self.activation = nn.SiLU() + self.use_adaln_lora = use_adaln_lora + if use_adaln_lora: + self.linear_2 = nn.Linear(out_features, 3 * out_features, bias=False) + else: + self.linear_2 = nn.Linear(out_features, out_features, bias=False) + + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self.in_dim) + torch.nn.init.trunc_normal_(self.linear_1.weight, std=std, a=-3 * std, b=3 * std) + + std = 1.0 / math.sqrt(self.out_dim) + torch.nn.init.trunc_normal_(self.linear_2.weight, std=std, a=-3 * std, b=3 * std) + + def forward(self, sample: torch.Tensor) -> torch.Tensor: + emb = self.linear_1(sample) + emb = self.activation(emb) + emb = self.linear_2(emb) + + if self.use_adaln_lora: + adaln_lora_B_T_3D = emb + emb_B_T_D = sample + else: + emb_B_T_D = emb + adaln_lora_B_T_3D = None + + return emb_B_T_D, adaln_lora_B_T_3D + + +class FourierFeatures(nn.Module): + """ + Implements a layer that generates Fourier features from input tensors, based on randomly sampled + frequencies and phases. This can help in learning high-frequency functions in low-dimensional problems. + + [B] -> [B, D] + + Parameters: + num_channels (int): The number of Fourier features to generate. + bandwidth (float, optional): The scaling factor for the frequency of the Fourier features. Defaults to 1. + normalize (bool, optional): If set to True, the outputs are scaled by sqrt(2), usually to normalize + the variance of the features. Defaults to False. + + Example: + >>> layer = FourierFeatures(num_channels=256, bandwidth=0.5, normalize=True) + >>> x = torch.randn(10, 256) # Example input tensor + >>> output = layer(x) + >>> print(output.shape) # Expected shape: (10, 256) + """ + + def __init__(self, num_channels, bandwidth=1, normalize=False): + super().__init__() + self.register_buffer("freqs", 2 * np.pi * bandwidth * torch.randn(num_channels), persistent=True) + self.register_buffer("phases", 2 * np.pi * torch.rand(num_channels), persistent=True) + self.gain = np.sqrt(2) if normalize else 1 + self.bandwidth = bandwidth + self.num_channels = num_channels + + self.reset_parameters() + + def reset_parameters(self) -> None: + generator = torch.Generator() + generator.manual_seed(0) + self.freqs = ( + 2 * np.pi * self.bandwidth * torch.randn(self.num_channels, generator=generator).to(self.freqs.device) + ) + self.phases = 2 * np.pi * torch.rand(self.num_channels, generator=generator).to(self.freqs.device) + + def forward(self, x, gain: float = 1.0): + """ + Apply the Fourier feature transformation to the input tensor. + + Args: + x (torch.Tensor): The input tensor. + gain (float, optional): An additional gain factor applied during the forward pass. Defaults to 1. + + Returns: + torch.Tensor: The transformed tensor, with Fourier features applied. + """ + in_dtype = x.dtype + x = x.to(torch.float32).ger(self.freqs.to(torch.float32)).add(self.phases.to(torch.float32)) + x = x.cos().mul(self.gain * gain).to(in_dtype) + return x + + +class PatchEmbed(nn.Module): + """ + PatchEmbed is a module for embedding patches from an input tensor by applying either 3D or 2D convolutional layers, + depending on the . This module can process inputs with temporal (video) and spatial (image) dimensions, + making it suitable for video and image processing tasks. It supports dividing the input into patches + and embedding each patch into a vector of size `out_channels`. + + Parameters: + - spatial_patch_size (int): The size of each spatial patch. + - temporal_patch_size (int): The size of each temporal patch. + - in_channels (int): Number of input channels. Default: 3. + - out_channels (int): The dimension of the embedding vector for each patch. Default: 768. + - bias (bool): If True, adds a learnable bias to the output of the convolutional layers. Default: True. + """ + + def __init__( + self, + spatial_patch_size, + temporal_patch_size, + in_channels=3, + out_channels=768, + ): + super().__init__() + self.spatial_patch_size = spatial_patch_size + self.temporal_patch_size = temporal_patch_size + + self.proj = nn.Sequential( + Rearrange( + "b c (t r) (h m) (w n) -> b t h w (c r m n)", + r=temporal_patch_size, + m=spatial_patch_size, + n=spatial_patch_size, + ), + nn.Linear( + in_channels * spatial_patch_size * spatial_patch_size * temporal_patch_size, out_channels, bias=False + ), + ) + self.dim = in_channels * spatial_patch_size * spatial_patch_size * temporal_patch_size + + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self.dim) + torch.nn.init.trunc_normal_(self.proj[1].weight, std=std, a=-3 * std, b=3 * std) + + def forward(self, x): + """ + Forward pass of the PatchEmbed module. + + Parameters: + - x (torch.Tensor): The input tensor of shape (B, C, T, H, W) where + B is the batch size, + C is the number of channels, + T is the temporal dimension, + H is the height, and + W is the width of the input. + + Returns: + - torch.Tensor: The embedded patches as a tensor, with shape b t h w c. + """ + assert x.dim() == 5 + _, _, T, H, W = x.shape + assert H % self.spatial_patch_size == 0 and W % self.spatial_patch_size == 0, ( + f"H,W {(H, W)} should be divisible by spatial_patch_size {self.spatial_patch_size}" + ) + assert T % self.temporal_patch_size == 0 + x = self.proj(x) + return x + + +class FinalLayer(nn.Module): + """ + The final layer of video DiT. + """ + + def __init__( + self, + hidden_size, + spatial_patch_size, + temporal_patch_size, + out_channels, + use_adaln_lora: bool = False, + adaln_lora_dim: int = 256, + use_wan_fp32_strategy: bool = False, + ): + super().__init__() + self.use_wan_fp32_strategy = use_wan_fp32_strategy + self.layer_norm = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6) + self.linear = nn.Linear( + hidden_size, spatial_patch_size * spatial_patch_size * temporal_patch_size * out_channels, bias=False + ) + self.hidden_size = hidden_size + self.n_adaln_chunks = 2 + self.use_adaln_lora = use_adaln_lora + self.adaln_lora_dim = adaln_lora_dim + if use_adaln_lora: + self.adaln_modulation = nn.Sequential( + nn.SiLU(), + nn.Linear(hidden_size, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, self.n_adaln_chunks * hidden_size, bias=False), + ) + else: + self.adaln_modulation = nn.Sequential( + nn.SiLU(), nn.Linear(hidden_size, self.n_adaln_chunks * hidden_size, bias=False) + ) + + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self.hidden_size) + torch.nn.init.trunc_normal_(self.linear.weight, std=std, a=-3 * std, b=3 * std) + if self.use_adaln_lora: + torch.nn.init.trunc_normal_(self.adaln_modulation[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.zeros_(self.adaln_modulation[2].weight) + else: + torch.nn.init.zeros_(self.adaln_modulation[1].weight) + + self.layer_norm.reset_parameters() + + def forward( + self, + # x_BT_HW_D, + x_B_T_H_W_D, + emb_B_T_D, + adaln_lora_B_T_3D: Optional[torch.Tensor] = None, + ): + if self.use_wan_fp32_strategy: + assert emb_B_T_D.dtype == torch.float32 + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if self.use_adaln_lora: + assert adaln_lora_B_T_3D is not None + shift_B_T_D, scale_B_T_D = ( + self.adaln_modulation(emb_B_T_D) + adaln_lora_B_T_3D[:, :, : 2 * self.hidden_size] + ).chunk(2, dim=-1) + else: + shift_B_T_D, scale_B_T_D = self.adaln_modulation(emb_B_T_D).chunk(2, dim=-1) + + shift_B_T_1_1_D, scale_B_T_1_1_D = ( + rearrange(shift_B_T_D, "b t d -> b t 1 1 d"), + rearrange(scale_B_T_D, "b t d -> b t 1 1 d"), + ) + + def _fn(_x_B_T_H_W_D, _norm_layer, _scale_B_T_1_1_D, _shift_B_T_1_1_D): + return _norm_layer(_x_B_T_H_W_D) * (1 + _scale_B_T_1_1_D) + _shift_B_T_1_1_D + + x_B_T_H_W_D = _fn(x_B_T_H_W_D, self.layer_norm, scale_B_T_1_1_D, shift_B_T_1_1_D) + x_B_T_H_W_O = self.linear( + x_B_T_H_W_D + ) # O = spatial_patch_size * spatial_patch_size * temporal_patch_size * out_channels + return x_B_T_H_W_O + + +class Block(nn.Module): + """ + A transformer block that combines self-attention, cross-attention and MLP layers with AdaLN modulation. + Each component (self-attention, cross-attention, MLP) has its own layer normalization and AdaLN modulation. + + Parameters: + x_dim (int): Dimension of input features + context_dim (int): Dimension of context features for cross-attention + num_heads (int): Number of attention heads + mlp_ratio (float): Multiplier for MLP hidden dimension. Default: 4.0 + use_adaln_lora (bool): Whether to use AdaLN-LoRA modulation. Default: False + adaln_lora_dim (int): Hidden dimension for AdaLN-LoRA layers. Default: 256 + use_wan_fp32_strategy (bool): Whether to use Wan's FP32 strategy. Default: False + If True, in Attention layer, if do self-attention, q and k will be forced to fp32 before rotary pos emb + also, in modulation computation, force entire computation in fp32 + + The block applies the following sequence: + 1. Self-attention with AdaLN modulation + 2. Cross-attention with AdaLN modulation + 3. MLP with AdaLN modulation + + Each component uses skip connections and layer normalization. + """ + + def __init__( + self, + x_dim: int, + context_dim: int, + num_heads: int, + mlp_ratio: float = 4.0, + use_adaln_lora: bool = False, + adaln_lora_dim: int = 256, + cam_dim: int = 1536, + backend: str = "transformer_engine", + image_context_dim: Optional[int] = None, + use_wan_fp32_strategy: bool = False, + ): + super().__init__() + self.x_dim = x_dim + self.layer_norm_self_attn = nn.LayerNorm(x_dim, elementwise_affine=False, eps=1e-6) + self.self_attn = Attention( + x_dim, + None, + num_heads, + x_dim // num_heads, + qkv_format="bshd", + backend=backend, + use_wan_fp32_strategy=use_wan_fp32_strategy, + ) + + self.layer_norm_cross_attn = nn.LayerNorm(x_dim, elementwise_affine=False, eps=1e-6) + self.cross_attn = Attention( + x_dim, context_dim, num_heads, x_dim // num_heads, qkv_format="bshd", backend=backend + ) + if image_context_dim is None: + self.cross_attn = Attention(x_dim, context_dim, num_heads, x_dim // num_heads, qkv_format="bshd") + else: + self.cross_attn = I2VCrossAttention( + x_dim, context_dim, num_heads, x_dim // num_heads, img_latent_dim=image_context_dim, qkv_format="bshd" + ) + + self.layer_norm_mlp = nn.LayerNorm(x_dim, elementwise_affine=False, eps=1e-6) + self.mlp = GPT2FeedForward(x_dim, int(x_dim * mlp_ratio)) + + self.use_adaln_lora = use_adaln_lora + if self.use_adaln_lora: + self.adaln_modulation_self_attn = nn.Sequential( + nn.SiLU(), + nn.Linear(x_dim, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, 3 * x_dim, bias=False), + ) + self.adaln_modulation_cross_attn = nn.Sequential( + nn.SiLU(), + nn.Linear(x_dim, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, 3 * x_dim, bias=False), + ) + self.adaln_modulation_mlp = nn.Sequential( + nn.SiLU(), + nn.Linear(x_dim, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, 3 * x_dim, bias=False), + ) + else: + self.adaln_modulation_self_attn = nn.Sequential(nn.SiLU(), nn.Linear(x_dim, 3 * x_dim, bias=False)) + self.adaln_modulation_cross_attn = nn.Sequential(nn.SiLU(), nn.Linear(x_dim, 3 * x_dim, bias=False)) + self.adaln_modulation_mlp = nn.Sequential(nn.SiLU(), nn.Linear(x_dim, 3 * x_dim, bias=False)) + + self.cam_dim = cam_dim + self.cam_encoder = nn.Linear(self.cam_dim, x_dim, bias=False) + + self.cp_size = None + self.use_wan_fp32_strategy = use_wan_fp32_strategy + + def set_context_parallel_group(self, process_group, ranks, stream): + self.cp_size = None if ranks is None else len(ranks) + self.self_attn.set_context_parallel_group( + process_group=process_group, + ranks=ranks, + stream=stream, + ) + + def reset_parameters(self) -> None: + self.layer_norm_self_attn.reset_parameters() + self.layer_norm_cross_attn.reset_parameters() + self.layer_norm_mlp.reset_parameters() + + if self.use_adaln_lora: + std = 1.0 / math.sqrt(self.x_dim) + torch.nn.init.trunc_normal_(self.adaln_modulation_self_attn[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.adaln_modulation_cross_attn[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.adaln_modulation_mlp[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.zeros_(self.adaln_modulation_self_attn[2].weight) + torch.nn.init.zeros_(self.adaln_modulation_cross_attn[2].weight) + torch.nn.init.zeros_(self.adaln_modulation_mlp[2].weight) + else: + torch.nn.init.zeros_(self.adaln_modulation_self_attn[1].weight) + torch.nn.init.zeros_(self.adaln_modulation_cross_attn[1].weight) + torch.nn.init.zeros_(self.adaln_modulation_mlp[1].weight) + + def init_weights(self) -> None: + self.reset_parameters() + self.self_attn.init_weights() + self.cross_attn.init_weights() + self.mlp.init_weights() + + # init camera weights + std = 1.0 / math.sqrt(self.x_dim) + torch.nn.init.trunc_normal_(self.cam_encoder.weight, std=std, a=-3 * std, b=3 * std) + + def forward( + self, + x_B_T_H_W_D: torch.Tensor, + emb_B_T_D: torch.Tensor, + crossattn_emb: torch.Tensor, + rope_emb_L_1_1_D: Optional[torch.Tensor] = None, + adaln_lora_B_T_3D: Optional[torch.Tensor] = None, + extra_per_block_pos_emb: Optional[torch.Tensor] = None, + camera: Optional[torch.Tensor] = None, + ) -> torch.Tensor: + if extra_per_block_pos_emb is not None: + x_B_T_H_W_D = x_B_T_H_W_D + extra_per_block_pos_emb + + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if self.use_adaln_lora: + shift_self_attn_B_T_D, scale_self_attn_B_T_D, gate_self_attn_B_T_D = ( + self.adaln_modulation_self_attn(emb_B_T_D) + adaln_lora_B_T_3D + ).chunk(3, dim=-1) + shift_cross_attn_B_T_D, scale_cross_attn_B_T_D, gate_cross_attn_B_T_D = ( + self.adaln_modulation_cross_attn(emb_B_T_D) + adaln_lora_B_T_3D + ).chunk(3, dim=-1) + shift_mlp_B_T_D, scale_mlp_B_T_D, gate_mlp_B_T_D = ( + self.adaln_modulation_mlp(emb_B_T_D) + adaln_lora_B_T_3D + ).chunk(3, dim=-1) + else: + shift_self_attn_B_T_D, scale_self_attn_B_T_D, gate_self_attn_B_T_D = self.adaln_modulation_self_attn( + emb_B_T_D + ).chunk(3, dim=-1) + shift_cross_attn_B_T_D, scale_cross_attn_B_T_D, gate_cross_attn_B_T_D = ( + self.adaln_modulation_cross_attn(emb_B_T_D).chunk(3, dim=-1) + ) + shift_mlp_B_T_D, scale_mlp_B_T_D, gate_mlp_B_T_D = self.adaln_modulation_mlp(emb_B_T_D).chunk(3, dim=-1) + + # Reshape tensors from (B, T, D) to (B, T, 1, 1, D) for broadcasting + shift_self_attn_B_T_1_1_D = rearrange(shift_self_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + scale_self_attn_B_T_1_1_D = rearrange(scale_self_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + gate_self_attn_B_T_1_1_D = rearrange(gate_self_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + + shift_cross_attn_B_T_1_1_D = rearrange(shift_cross_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + scale_cross_attn_B_T_1_1_D = rearrange(scale_cross_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + gate_cross_attn_B_T_1_1_D = rearrange(gate_cross_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + + shift_mlp_B_T_1_1_D = rearrange(shift_mlp_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + scale_mlp_B_T_1_1_D = rearrange(scale_mlp_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + gate_mlp_B_T_1_1_D = rearrange(gate_mlp_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + + B, T, H, W, D = x_B_T_H_W_D.shape + + def _fn(_x_B_T_H_W_D, _norm_layer, _scale_B_T_1_1_D, _shift_B_T_1_1_D): + return _norm_layer(_x_B_T_H_W_D) * (1 + _scale_B_T_1_1_D) + _shift_B_T_1_1_D + + normalized_x_B_T_H_W_D = _fn( + x_B_T_H_W_D, + self.layer_norm_self_attn, + scale_self_attn_B_T_1_1_D, + shift_self_attn_B_T_1_1_D, + ) + + video_size = VideoSize(T=T, H=H, W=W) + + # (ahassani): Hack to correct `video_size` when CP is enabled. + # I really don't like this, but there doesn't seem to be any central + # piece of code that's responsible for handling CP/TP that also defines the + # layout of shardings. Other parts of the code (i.e. RoPE) seem to make this + # assumption that CP sharding is always done along T. + if self.cp_size is not None and self.cp_size > 1: + video_size = VideoSize(T=T * self.cp_size, H=H, W=W) + + cam_emb = self.cam_encoder(camera) + + result_B_T_H_W_D = rearrange( + self.self_attn( + # normalized_x_B_T_HW_D, + rearrange(normalized_x_B_T_H_W_D + cam_emb, "b t h w d -> b (t h w) d"), + None, + rope_emb=rope_emb_L_1_1_D, + video_size=video_size, + ), + "b (t h w) d -> b t h w d", + t=T, + h=H, + w=W, + ) + x_B_T_H_W_D = x_B_T_H_W_D + gate_self_attn_B_T_1_1_D * result_B_T_H_W_D + + def _x_fn( + _x_B_T_H_W_D, + layer_norm_cross_attn, + _scale_cross_attn_B_T_1_1_D, + _shift_cross_attn_B_T_1_1_D, + _gate_cross_attn_B_T_1_1_D, + ): + _normalized_x_B_T_H_W_D = _fn( + _x_B_T_H_W_D, layer_norm_cross_attn, _scale_cross_attn_B_T_1_1_D, _shift_cross_attn_B_T_1_1_D + ) + _result_B_T_H_W_D = rearrange( + self.cross_attn( + rearrange(_normalized_x_B_T_H_W_D, "b t h w d -> b (t h w) d"), + crossattn_emb, + rope_emb=rope_emb_L_1_1_D, + ), + "b (t h w) d -> b t h w d", + t=T, + h=H, + w=W, + ) + # _x_B_T_H_W_D = _x_B_T_H_W_D + _gate_cross_attn_B_T_1_1_D * _result_B_T_H_W_D + return _result_B_T_H_W_D + + result_B_T_H_W_D = _x_fn( + x_B_T_H_W_D, + self.layer_norm_cross_attn, + scale_cross_attn_B_T_1_1_D, + shift_cross_attn_B_T_1_1_D, + gate_cross_attn_B_T_1_1_D, + ) + x_B_T_H_W_D = result_B_T_H_W_D * gate_cross_attn_B_T_1_1_D + x_B_T_H_W_D + + normalized_x_B_T_H_W_D = _fn( + x_B_T_H_W_D, + self.layer_norm_mlp, + scale_mlp_B_T_1_1_D, + shift_mlp_B_T_1_1_D, + ) + result_B_T_H_W_D = self.mlp(normalized_x_B_T_H_W_D) + x_B_T_H_W_D = x_B_T_H_W_D + gate_mlp_B_T_1_1_D * result_B_T_H_W_D + return x_B_T_H_W_D + + +class CameraARMiniTrainDIT(WeightTrainingStat): + """ + A clean impl of DIT that can load and reproduce the training results of the original DIT model in edify_video/v4~(cosmos 1) + A general implementation of adaln-modulated VIT-like~(DiT) transformer for video processing. + + Args: + max_img_h (int): Maximum height of the input images. + max_img_w (int): Maximum width of the input images. + max_frames (int): Maximum number of frames in the video sequence. + in_channels (int): Number of input channels (e.g., RGB channels for color images). + out_channels (int): Number of output channels. + patch_spatial (tuple): Spatial resolution of patches for input processing. + patch_temporal (int): Temporal resolution of patches for input processing. + concat_padding_mask (bool): If True, includes a mask channel in the input to handle padding. + model_channels (int): Base number of channels used throughout the model. + num_blocks (int): Number of transformer blocks. + num_heads (int): Number of heads in the multi-head attention layers. + mlp_ratio (float): Expansion ratio for MLP blocks. + crossattn_emb_channels (int): Number of embedding channels for cross-attention. + extra_image_context_dim (int): Number of embedding channels for extra image context. + pos_emb_cls (str): Type of positional embeddings. + pos_emb_learnable (bool): Whether positional embeddings are learnable. + pos_emb_interpolation (str): Method for interpolating positional embeddings. + min_fps (int): Minimum frames per second. + max_fps (int): Maximum frames per second. + use_adaln_lora (bool): Whether to use AdaLN-LoRA. + adaln_lora_dim (int): Dimension for AdaLN-LoRA. + rope_h_extrapolation_ratio (float): Height extrapolation ratio for RoPE. + rope_w_extrapolation_ratio (float): Width extrapolation ratio for RoPE. + rope_t_extrapolation_ratio (float): Temporal extrapolation ratio for RoPE. + extra_per_block_abs_pos_emb (bool): Whether to use extra per-block absolute positional embeddings. + extra_h_extrapolation_ratio (float): Height extrapolation ratio for extra embeddings. + extra_w_extrapolation_ratio (float): Width extrapolation ratio for extra embeddings. + extra_t_extrapolation_ratio (float): Temporal extrapolation ratio for extra embeddings. + n_dense_blocks (`int`, *optional*, defaults to -1): + Number of blocks that will remain dense (not replaced with sparse attention) + If -1, no blocks are replaced with sparse attention + If 0, all blocks use sparse attention + Otherwise, n_dense_blocks blocks will remain dense, distributed evenly across the network + natten_parameters (`dict`, *optional*, defaults to None): + NATTEN (Sparse attention) parameter list. + The list length must be the same as the number of layers, with each list element + indicating NATTEN parameters for that layer. If None, NATTEN will not be used in that + layer and it would remain a full dense self attention. If not None, it must be a + dictionary/mapping with at least the following key: + - window_size: `tuple` of size 3 indicating neighborhood attention window size. + window size of -1 along any dimension means self attention. + Other optional parameters and their keys: + - stride: `tuple` of size 3 indicating neighborhood attention stride value. + stride = 1 is standard neighborhood attention, stride = window size means + blocked/window self attention (WSA) along that dimension. Any other values are + strided neighborhood attention. Refer to the GNA paper for more information. + + - dilation: `tuple` of size 3 indicating neighborhood attention dilation value. + dilation = 1 is standard neighborhood attention. Refer to the DiNAT paper for more + information. + + - is_causal: `tuple` of 3 booleans indicating whether causal masking is enabled for + any of the T, H, W dimensions. + """ + + def __init__( + self, + max_img_h: int, + max_img_w: int, + max_frames: int, + in_channels: int, + out_channels: int, + patch_spatial: tuple, + patch_temporal: int, + concat_padding_mask: bool = True, + # attention settings + model_channels: int = 768, + num_blocks: int = 10, + num_heads: int = 16, + mlp_ratio: float = 4.0, + atten_backend: str = "transformer_engine", + # cross attention settings + crossattn_emb_channels: int = 1024, + use_crossattn_projection: bool = False, + crossattn_proj_in_channels: int = 1024, + extra_image_context_dim: Optional[int] = None, + # positional embedding settings + pos_emb_cls: str = "sincos", + pos_emb_learnable: bool = False, + pos_emb_interpolation: str = "crop", + min_fps: int = 1, # 1 for getty video + max_fps: int = 30, # 120 for getty video but let's use 30 + use_adaln_lora: bool = False, + adaln_lora_dim: int = 256, + rope_h_extrapolation_ratio: float = 1.0, + rope_w_extrapolation_ratio: float = 1.0, + rope_t_extrapolation_ratio: float = 1.0, + extra_per_block_abs_pos_emb: bool = False, + extra_h_extrapolation_ratio: float = 1.0, + extra_w_extrapolation_ratio: float = 1.0, + extra_t_extrapolation_ratio: float = 1.0, + rope_enable_fps_modulation: bool = True, + sac_config: SACConfig = SACConfig(), + n_dense_blocks: int = -1, + natten_parameters: Union[dict, list] = None, + # if True, will closely match wan's strategy to use fp32 in certain layers/operations + use_wan_fp32_strategy: bool = False, + ) -> None: + super().__init__() + self.max_img_h = max_img_h + self.max_img_w = max_img_w + self.max_frames = max_frames + self.in_channels = in_channels + self.out_channels = out_channels + self.patch_spatial = patch_spatial + self.patch_temporal = patch_temporal + self.num_heads = num_heads + self.num_blocks = num_blocks + self.model_channels = model_channels + self.concat_padding_mask = concat_padding_mask + self.atten_backend = atten_backend + # positional embedding settings + self.pos_emb_cls = pos_emb_cls + self.pos_emb_learnable = pos_emb_learnable + self.pos_emb_interpolation = pos_emb_interpolation + self.min_fps = min_fps + self.max_fps = max_fps + self.rope_h_extrapolation_ratio = rope_h_extrapolation_ratio + self.rope_w_extrapolation_ratio = rope_w_extrapolation_ratio + self.rope_t_extrapolation_ratio = rope_t_extrapolation_ratio + self.extra_per_block_abs_pos_emb = extra_per_block_abs_pos_emb + self.extra_h_extrapolation_ratio = extra_h_extrapolation_ratio + self.extra_w_extrapolation_ratio = extra_w_extrapolation_ratio + self.extra_t_extrapolation_ratio = extra_t_extrapolation_ratio + self.rope_enable_fps_modulation = rope_enable_fps_modulation + self.extra_image_context_dim = extra_image_context_dim + self.build_patch_embed() + self.build_pos_embed() + self.use_adaln_lora = use_adaln_lora + self.adaln_lora_dim = adaln_lora_dim + self.t_embedder = nn.Sequential( + Timesteps(model_channels), + TimestepEmbedding(model_channels, model_channels, use_adaln_lora=use_adaln_lora), + ) + self.use_crossattn_projection = use_crossattn_projection + self.crossattn_proj_in_channels = crossattn_proj_in_channels + self.use_wan_fp32_strategy = use_wan_fp32_strategy + + self.blocks = nn.ModuleList( + [ + Block( + x_dim=model_channels, + context_dim=crossattn_emb_channels, + num_heads=num_heads, + mlp_ratio=mlp_ratio, + use_adaln_lora=use_adaln_lora, + adaln_lora_dim=adaln_lora_dim, + backend=atten_backend, + image_context_dim=None if extra_image_context_dim is None else model_channels, + use_wan_fp32_strategy=use_wan_fp32_strategy, + ) + for _ in range(num_blocks) + ] + ) + + self.final_layer = FinalLayer( + hidden_size=self.model_channels, + spatial_patch_size=self.patch_spatial, + temporal_patch_size=self.patch_temporal, + out_channels=self.out_channels, + use_adaln_lora=self.use_adaln_lora, + adaln_lora_dim=self.adaln_lora_dim, + use_wan_fp32_strategy=self.use_wan_fp32_strategy, + ) + + self.t_embedding_norm = te.pytorch.RMSNorm(model_channels, eps=1e-6) + if extra_image_context_dim is not None: + self.img_context_proj = nn.Sequential( + nn.Linear( + extra_image_context_dim, model_channels, bias=True + ), # help distinguish between image and video context + nn.GELU(), + ) + + if use_crossattn_projection: + self.crossattn_proj = nn.Sequential( + nn.Linear(crossattn_proj_in_channels, crossattn_emb_channels, bias=True), + nn.GELU(), + ) + + self.init_weights() + self.enable_selective_checkpoint(sac_config, self.blocks) + + # Replace self-attention with sparse attention if specified + if n_dense_blocks != -1: + self = replace_selfattn_op_with_sparse_attn_op(self, n_dense_blocks, natten_parameters=natten_parameters) + + self._is_context_parallel_enabled = False + + self.freeze_parameters() + + def freeze_parameters(self): + for name, module in self.named_modules(): + if any(keyword in name for keyword in ["cam_encoder", "self_attn"]): + for param in module.parameters(): + param.requires_grad = True + else: + for param in module.parameters(): + param.requires_grad = False + + def init_weights(self): + self.x_embedder.init_weights() + self.pos_embedder.reset_parameters() + if self.extra_per_block_abs_pos_emb: + self.extra_pos_embedder.reset_parameters() + + self.t_embedder[1].init_weights() + for block in self.blocks: + block.init_weights() + + self.final_layer.init_weights() + self.t_embedding_norm.reset_parameters() + + if self.extra_image_context_dim is not None: + self.img_context_proj[0].reset_parameters() + + def build_patch_embed(self): + ( + concat_padding_mask, + in_channels, + patch_spatial, + patch_temporal, + model_channels, + ) = ( + self.concat_padding_mask, + self.in_channels, + self.patch_spatial, + self.patch_temporal, + self.model_channels, + ) + in_channels = in_channels + 1 if concat_padding_mask else in_channels + self.x_embedder = PatchEmbed( + spatial_patch_size=patch_spatial, + temporal_patch_size=patch_temporal, + in_channels=in_channels, + out_channels=model_channels, + ) + + def build_pos_embed(self): + if self.pos_emb_cls == "rope3d": + cls_type = VideoRopePosition3DEmb + else: + raise ValueError(f"Unknown pos_emb_cls {self.pos_emb_cls}") + + log.debug(f"Building positional embedding with {self.pos_emb_cls} class, impl {cls_type}") + kwargs = dict( + model_channels=self.model_channels, + len_h=self.max_img_h // self.patch_spatial, + len_w=self.max_img_w // self.patch_spatial, + len_t=self.max_frames // self.patch_temporal, + max_fps=self.max_fps, + min_fps=self.min_fps, + is_learnable=self.pos_emb_learnable, + interpolation=self.pos_emb_interpolation, + head_dim=self.model_channels // self.num_heads, + h_extrapolation_ratio=self.rope_h_extrapolation_ratio, + w_extrapolation_ratio=self.rope_w_extrapolation_ratio, + t_extrapolation_ratio=self.rope_t_extrapolation_ratio, + enable_fps_modulation=self.rope_enable_fps_modulation, + ) + self.pos_embedder = cls_type( + **kwargs, + ) + + if self.extra_per_block_abs_pos_emb: + kwargs["h_extrapolation_ratio"] = self.extra_h_extrapolation_ratio + kwargs["w_extrapolation_ratio"] = self.extra_w_extrapolation_ratio + kwargs["t_extrapolation_ratio"] = self.extra_t_extrapolation_ratio + self.extra_pos_embedder = LearnablePosEmbAxis( + **kwargs, + ) + + def prepare_embedded_sequence( + self, + x_B_C_T_H_W: torch.Tensor, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[torch.Tensor]]: + """ + Prepares an embedded sequence tensor by applying positional embeddings and handling padding masks. + + Args: + x_B_C_T_H_W (torch.Tensor): video + fps (Optional[torch.Tensor]): Frames per second tensor to be used for positional embedding when required. + If None, a default value (`self.base_fps`) will be used. + padding_mask (Optional[torch.Tensor]): current it is not used + + Returns: + Tuple[torch.Tensor, Optional[torch.Tensor]]: + - A tensor of shape (B, T, H, W, D) with the embedded sequence. + - An optional positional embedding tensor, returned only if the positional embedding class + (`self.pos_emb_cls`) includes 'rope'. Otherwise, None. + + Notes: + - If `self.concat_padding_mask` is True, a padding mask channel is concatenated to the input tensor. + - The method of applying positional embeddings depends on the value of `self.pos_emb_cls`. + - If 'rope' is in `self.pos_emb_cls` (case insensitive), the positional embeddings are generated using + the `self.pos_embedder` with the shape [T, H, W]. + - If "fps_aware" is in `self.pos_emb_cls`, the positional embeddings are generated using the + `self.pos_embedder` with the fps tensor. + - Otherwise, the positional embeddings are generated without considering fps. + """ + if self.concat_padding_mask: + padding_mask = transforms.functional.resize( + padding_mask, list(x_B_C_T_H_W.shape[-2:]), interpolation=transforms.InterpolationMode.NEAREST + ) + x_B_C_T_H_W = torch.cat( + [x_B_C_T_H_W, padding_mask.unsqueeze(1).repeat(1, 1, x_B_C_T_H_W.shape[2], 1, 1)], dim=1 + ) + x_B_T_H_W_D = self.x_embedder(x_B_C_T_H_W) + + if self.extra_per_block_abs_pos_emb: + extra_pos_emb = self.extra_pos_embedder(x_B_T_H_W_D, fps=fps) + else: + extra_pos_emb = None + + if "rope" in self.pos_emb_cls.lower(): + return x_B_T_H_W_D, self.pos_embedder(x_B_T_H_W_D, fps=fps), extra_pos_emb + x_B_T_H_W_D = x_B_T_H_W_D + self.pos_embedder(x_B_T_H_W_D) # [B, T, H, W, D] + + return x_B_T_H_W_D, None, extra_pos_emb + + def unpatchify(self, x_B_T_H_W_M): + x_B_C_Tt_Hp_Wp = rearrange( + x_B_T_H_W_M, + "B T H W (p1 p2 t C) -> B C (T t) (H p1) (W p2)", + p1=self.patch_spatial, + p2=self.patch_spatial, + t=self.patch_temporal, + ) + return x_B_C_Tt_Hp_Wp + + def forward( + self, + x_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + crossattn_emb: torch.Tensor, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + data_type: Optional[DataType] = DataType.VIDEO, + intermediate_feature_ids: Optional[List[int]] = None, + img_context_emb: Optional[torch.Tensor] = None, + camera: Optional[torch.Tensor] = None, + ) -> torch.Tensor | List[torch.Tensor] | Tuple[torch.Tensor, List[torch.Tensor]]: + """ + Args: + x: (B, C, T, H, W) tensor of spatial-temp inputs + timesteps: (B, ) tensor of timesteps + crossattn_emb: (B, N, D) tensor of cross-attention embeddings + """ + assert isinstance(data_type, DataType), ( + f"Expected DataType, got {type(data_type)}. We need discuss this flag later." + ) + x_B_T_H_W_D, rope_emb_L_1_1_D, extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D = self.prepare_embedded_sequence( + x_B_C_T_H_W, + fps=fps, + padding_mask=padding_mask, + ) + + if self.use_crossattn_projection: + crossattn_emb = self.crossattn_proj(crossattn_emb) + + if img_context_emb is not None: + assert self.extra_image_context_dim is not None, ( + "extra_image_context_dim must be set if img_context_emb is provided" + ) + img_context_emb = self.img_context_proj(img_context_emb) + context_input = (crossattn_emb, img_context_emb) + else: + context_input = crossattn_emb + + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if timesteps_B_T.ndim == 1: + timesteps_B_T = timesteps_B_T.unsqueeze(1) + t_embedding_B_T_D, adaln_lora_B_T_3D = self.t_embedder(timesteps_B_T) + t_embedding_B_T_D = self.t_embedding_norm(t_embedding_B_T_D) + + # for logging purpose + affline_scale_log_info = {} + affline_scale_log_info["t_embedding_B_T_D"] = t_embedding_B_T_D.detach() + self.affline_scale_log_info = affline_scale_log_info + self.affline_emb = t_embedding_B_T_D + self.crossattn_emb = crossattn_emb + + if extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D is not None: + assert x_B_T_H_W_D.shape == extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D.shape, ( + f"{x_B_T_H_W_D.shape} != {extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D.shape}" + ) + + B, T, H, W, D = x_B_T_H_W_D.shape + # x_B_THW_D = rearrange(x_B_T_H_W_D, "b t h w d -> b (t h w) d") + + intermediate_features_outputs = [] + for i, block in enumerate(self.blocks): + x_B_T_H_W_D = block( + x_B_T_H_W_D, + t_embedding_B_T_D, + context_input, + rope_emb_L_1_1_D=rope_emb_L_1_1_D, + adaln_lora_B_T_3D=adaln_lora_B_T_3D, + extra_per_block_pos_emb=extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D, + camera=camera, + ) + if intermediate_feature_ids and i in intermediate_feature_ids: + x_reshaped_for_disc = rearrange(x_B_T_H_W_D, "b tp hp wp d -> b (tp hp wp) d") + intermediate_features_outputs.append(x_reshaped_for_disc) + + # x_B_T_H_W_D = rearrange(x_B_THW_D, "b (t h w) d -> b t h w d", t=T, h=H, w=W) + # O = out_channels * spatial_patch_size * spatial_patch_size * temporal_patch_size + x_B_T_H_W_O = self.final_layer(x_B_T_H_W_D, t_embedding_B_T_D, adaln_lora_B_T_3D=adaln_lora_B_T_3D) + x_B_C_Tt_Hp_Wp = self.unpatchify(x_B_T_H_W_O) + if intermediate_feature_ids: + if len(intermediate_features_outputs) != len(intermediate_feature_ids): + log.warning( + f"Collected {len(intermediate_features_outputs)} intermediate features, " + f"but expected {len(intermediate_feature_ids)}. " + f"Requested IDs: {intermediate_feature_ids}" + ) + return x_B_C_Tt_Hp_Wp, intermediate_features_outputs + + return x_B_C_Tt_Hp_Wp + + def enable_selective_checkpoint(self, sac_config: SACConfig, blocks: nn.ModuleList): + if sac_config.mode == CheckpointMode.NONE: + return self + + log.info( + f"Enable selective checkpoint with {sac_config.mode}, for every {sac_config.every_n_blocks} blocks. Total blocks: {len(blocks)}" + ) + _context_fn = sac_config.get_context_fn() + for block_id, block in blocks.named_children(): + if int(block_id) % sac_config.every_n_blocks == 0: + log.info(f"Enable selective checkpoint for block {block_id}") + block = ptd_checkpoint_wrapper( + block, + context_fn=_context_fn, + preserve_rng_state=False, + ) + blocks.register_module(block_id, block) + self.register_module( + "final_layer", + ptd_checkpoint_wrapper( + self.final_layer, + context_fn=_context_fn, + preserve_rng_state=False, + ), + ) + + return self + + def fully_shard(self, mesh): + for i, block in enumerate(self.blocks): + reshard_after_forward = i < len(self.blocks) - 1 + fully_shard(block, mesh=mesh, reshard_after_forward=reshard_after_forward) + + fully_shard(self.final_layer, mesh=mesh, reshard_after_forward=True) + if self.extra_per_block_abs_pos_emb: + fully_shard(self.extra_pos_embedder, mesh=mesh, reshard_after_forward=True) + fully_shard(self.t_embedder, mesh=mesh, reshard_after_forward=False) + if self.extra_image_context_dim is not None: + fully_shard(self.img_context_proj, mesh=mesh, reshard_after_forward=False) + + def disable_context_parallel(self): + # pos_embedder + self.pos_embedder.disable_context_parallel() + if self.extra_per_block_abs_pos_emb: + self.extra_pos_embedder.disable_context_parallel() + + # attention + for block in self.blocks: + block.set_context_parallel_group( + process_group=None, + ranks=None, + stream=torch.cuda.Stream(), + ) + + self._is_context_parallel_enabled = False + + def enable_context_parallel(self, process_group: Optional[ProcessGroup] = None): + # pos_embedder + self.pos_embedder.enable_context_parallel(process_group=process_group) + if self.extra_per_block_abs_pos_emb: + self.extra_pos_embedder.enable_context_parallel(process_group=process_group) + + # attention + cp_ranks = get_process_group_ranks(process_group) + for block in self.blocks: + block.set_context_parallel_group( + process_group=process_group, + ranks=cp_ranks, + stream=torch.cuda.Stream(), + ) + + self._is_context_parallel_enabled = True + + @property + def is_context_parallel_enabled(self): + return self._is_context_parallel_enabled + + +class CameraARMiniTrainDITwithConditionalMask(CameraARMiniTrainDIT): + def __init__(self, *args, timestep_scale: float = 1.0, **kwargs): + assert "in_channels" in kwargs, "in_channels must be provided" + kwargs["in_channels"] += 1 + self.timestep_scale = timestep_scale + super().__init__(*args, **kwargs) + + def forward( + self, + x_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + crossattn_emb: torch.Tensor, + condition_video_input_mask_B_C_T_H_W: Optional[torch.Tensor] = None, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + data_type: Optional[DataType] = DataType.VIDEO, + img_context_emb: Optional[torch.Tensor] = None, + camera: Optional[torch.Tensor] = None, + **kwargs, + ) -> torch.Tensor | List[torch.Tensor] | Tuple[torch.Tensor, List[torch.Tensor]]: + del kwargs + if data_type == DataType.VIDEO: + x_B_C_T_H_W = torch.cat([x_B_C_T_H_W, condition_video_input_mask_B_C_T_H_W.type_as(x_B_C_T_H_W)], dim=1) + else: + B, _, T, H, W = x_B_C_T_H_W.shape + x_B_C_T_H_W = torch.cat( + [x_B_C_T_H_W, torch.zeros((B, 1, T, H, W), dtype=x_B_C_T_H_W.dtype, device=x_B_C_T_H_W.device)], dim=1 + ) + return super().forward( + x_B_C_T_H_W=x_B_C_T_H_W, + timesteps_B_T=timesteps_B_T * self.timestep_scale, + crossattn_emb=crossattn_emb, + fps=fps, + padding_mask=padding_mask, + data_type=data_type, + img_context_emb=img_context_emb, + camera=camera, + ) + + +def replace_selfattn_op_with_sparse_attn_op( + model: CameraARMiniTrainDIT, n_dense_blocks: int = 0, natten_parameters: Union[dict, list] = None +) -> CameraARMiniTrainDIT: + """ + Replace the self-attention operator with a sparse self-attention operator. + + Args: + model: CameraARMiniTrainDIT instance + n_dense_blocks: Number of blocks that will remain dense (not replaced with NeighborhoodAttention) + If 0, all blocks use NeighborhoodAttention. + If -1, return model directly without any modifications. + Otherwise, n_dense_blocks blocks will remain dense, distributed evenly across the network. + + Returns: + Modified instance + """ + # Special case: return model directly without modifications + if n_dense_blocks == -1: + return model + + num_blocks = len(model.blocks) + + if natten_parameters is None: + raise ValueError("Please specify natten_parameters when n_dense_blocks > -1.") + + if isinstance(natten_parameters, Sequence) and len(natten_parameters) != num_blocks: + raise ValueError( + "List of NATTEN parameters must be the same length as the number of blocks, " + f"got {len(natten_parameters)=} != {num_blocks=}." + ) + + if isinstance(natten_parameters, Sequence) and n_dense_blocks > 0: + log.warning(f"NATTEN parameters was a list; ignoring {n_dense_blocks=}.") + + if isinstance(natten_parameters, Sequence): + natten_parameters_list = natten_parameters + else: + if n_dense_blocks >= num_blocks: + raise ValueError(f"n_dense_blocks ({n_dense_blocks}) must be less than the number of blocks ({num_blocks})") + + # Determine which blocks should remain dense + dense_indices = set() + + if n_dense_blocks > 0: + # General rule: distribute n_dense_blocks blocks evenly across the network + if n_dense_blocks == 1: + # Special case: just the middle block + dense_indices.add(num_blocks // 2) + else: + # For multiple blocks, distribute them evenly from start to end + indices = np.linspace(0, num_blocks - 1, n_dense_blocks, dtype=int) + dense_indices.update(indices.tolist()) + + natten_parameters_list = [None if i in dense_indices else natten_parameters for i in range(num_blocks)] + + # Replace self-attention with NeighborhoodAttention for non-dense blocks + for i, block in enumerate(model.blocks): + natten_params = natten_parameters_list[i] + if natten_params is not None: + natten_parameters_layer = {k: v for k, v in natten_params.items()} + natten_parameters_layer["layer_id"] = i + if block.self_attn.backend == "minimal_a2a": + sparse_attn_op = NattenA2AAttnOp(natten_parameters=natten_parameters_layer) + else: + raise NotImplementedError( + f"Using sparsity with attention backend {block.self_attn.backend} is not supported." + ) + + block.self_attn.register_module("attn_op", sparse_attn_op) + + return model diff --git a/REGEN-main/cosmos_policy/_src/predict2/checkpointer/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/checkpointer/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/checkpointer/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/checkpointer/dcp.py b/REGEN-main/cosmos_policy/_src/predict2/checkpointer/dcp.py new file mode 100644 index 0000000000000000000000000000000000000000..6f0a45cbc688bd2e20bcc52d20417b9e932fa321 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/checkpointer/dcp.py @@ -0,0 +1,760 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Distributed checkpoint (DCP) directory structure and storage backends. + +The checkpointer saves model state in a sharded format across multiple processes: + +self.save_dirname/ +├── iter_000000005/ # Checkpoint at iteration 5 +│ ├── model/ # Model state shards +│ │ ├── __0_0.distcp # Shard 0 from rank 0 +│ │ └── __1_0.distcp # Shard 1 from rank 1 +│ ├── optim/ # Optimizer state shards +│ │ ├── __0_0.distcp # Shard 0 from rank 0 +│ │ └── __1_0.distcp # Shard 1 from rank 1 +│ ├── scheduler/ # Learning rate scheduler state +│ │ ├── __0_0.distcp # Shard 0 from rank 0 +│ │ └── __1_0.distcp # Shard 1 from rank 1 +│ └── trainer/ # Additional training state +│ ├── __0_0.distcp # Shard 0 from rank 0 +│ └── __1_0.distcp # Shard 1 from rank 1 +└── latest_checkpoint.txt # Points to most recent checkpoint folder, e.g. iter_000000005 + +Storage backends: +- Local filesystem: + self.save_dirname = "{config_job.path_local}/checkpoints" + +- S3 object store: + self.save_dirname = "s3://{bucket}/{config_job.path}/checkpoints" + where bucket = self.config_checkpoint.save_to_object_store.bucket + +The sharded format enables efficient distributed saving/loading by: +1. Parallelizing I/O across processes +2. Reducing memory usage per process +3. Supporting both local and cloud storage backends +""" + +import enum +import functools +import multiprocessing +import os +import queue +import re +import time +from collections import namedtuple +from multiprocessing import get_context +from typing import Any, Dict, List, Optional, Set, Tuple, Union + +import torch +import torch.distributed +import torch.distributed.checkpoint as dcp +from torch import nn +from torch.distributed.checkpoint import FileSystemReader, FileSystemWriter +from torch.distributed.checkpoint.default_planner import DefaultSavePlanner +from torch.distributed.checkpoint.state_dict import ( + StateDictOptions, + get_model_state_dict, + get_optimizer_state_dict, + set_model_state_dict, + set_optimizer_state_dict, +) +from torch.distributed.checkpoint.stateful import Stateful + +from cosmos_policy._src.imaginaire.checkpointer.base import AbstractCheckpointer +from cosmos_policy._src.imaginaire.checkpointer.s3_filesystem import S3StorageReader, S3StorageWriter +from cosmos_policy._src.imaginaire.config import CheckpointConfig, JobConfig +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import callback, distributed, log, misc +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + +try: + from torch.distributed.checkpoint.default_planner import DefaultLoadPlanner as _DefaultLoadPlanner + from torch.distributed.checkpoint.default_planner import ( + DTensor, + LoadPlan, + _create_read_items, + _version, + flatten_state_dict, + ) + from torch.distributed.checkpoint.metadata import Metadata, TensorStorageMetadata + + def create_default_local_load_plan( + state_dict: dict[str, Any], metadata: Metadata, strict: bool = True, dcp_allow_mismatched_size: bool = False + ) -> LoadPlan: + requests = [] + """ + Create the ``LoadPlan`` used by DefaultLoadPlanner. + + It produces one read item per value in ``state_dict`` using the metadata in ``metadata``. + + The default behavior is to match key exactly between state_dict and metadata. + It handles resharding by issuing multiple read requests against storage in order to match + load requirements. + """ + + for fqn, obj in state_dict.items(): + if fqn.endswith("._extra_state"): # dirty TE attention package! + continue + # ignore state_dict keys which do not exist in `state_dict` if strict=False + if fqn not in metadata.state_dict_metadata: + if strict: + raise RuntimeError(f"Missing key in checkpoint state_dict: {fqn}.") + else: + continue + + md = metadata.state_dict_metadata[fqn] + + if not dcp_allow_mismatched_size: + if ( + isinstance(md, TensorStorageMetadata) + and getattr(obj, "size", None) is not None + and md.size != obj.size() + ): + if not strict: + log.critical(f"Size mismatch between saved {md.size} and current: {obj.size()} for {fqn}") + continue + else: + raise ValueError( + f"Size mismatch between saved {md.size} and current: {obj.size()} for {fqn}", + ) + # Since DTensor supports submesh, adding extra check to ensure _create_read_items() + # gets called only when the current rank is part of the mesh for the corresponding DTensor. + if isinstance(obj, DTensor): + if obj.device_mesh.get_coordinate() is not None: + requests += _create_read_items(fqn, md, obj) + else: + requests += _create_read_items(fqn, md, obj) + + return LoadPlan(requests) + + class DefaultLoadPlanner(_DefaultLoadPlanner): + def set_partial_channel_weight(self, dcp_allow_mismatched_size: bool): + self.dcp_allow_mismatched_size = dcp_allow_mismatched_size + + def create_local_plan(self) -> LoadPlan: + assert self.metadata is not None + if self.flatten_state_dict: + # To support checkpoints that are saved before v2.4, we have to + # differentiate if the missing keys are due to old checkpoints. + # The contracts are: + # 1. There are 3 cases when we found a missing key. + # 1.1 Actual missing key, but allow_partial_load is False + # 1.2 Actual missing key, but allow_partial load is True + # 1.3 Old checkpoint, but allow_partial_load is False + # 1.4 Old checkpoint, but allow_partial_load is True + # 2. If we found a missing key, we first convert the keys back to + # the key format of v2.3 + # 3. If the previous missing keys are in the v2.3 keys, we assume + # this is a old checkpoint. + # 4. Pass the state_dict to `create_default_local_load_plan()`, + # which has the logic to check missing for allow_partial_load. + # So for 1.2 and 1.4 cases, we delegate allow_partial_load check to + # `create_default_local_load_plan()`. The logic here is to determine + # whether the checkpoint belong to 2.3 (or before) or 2.4 (or after). + current_keys = set(self.state_dict.keys()) + load_keys = set(self.metadata.state_dict_metadata.keys()) + missing_keys = load_keys - current_keys + if missing_keys: + _version._derived_version = "2_3" + old_state_dict, old_mappings = flatten_state_dict(self.original_state_dict) + old_keys = set(old_state_dict.keys()) + if old_keys & missing_keys: + self.state_dict, self.mappings = old_state_dict, old_mappings + # _derived_version is only used by flatten_state_dict now. + # Set it back to None so that later we can save to a new version. + _version._derived_version = None + + return create_default_local_load_plan( + self.state_dict, + self.metadata, + not self.allow_partial_load, + getattr(self, "dcp_allow_mismatched_size", False), + ) + + log.critical("for the back comptiable pytorch! New DefaultLoadPlanner class is created.") +except ImportError as e: + from torch.distributed.checkpoint.default_planner import DefaultLoadPlanner + + log.critical(f"{e}, using default planner") + + +StateDictItemPath = namedtuple("StateDictItemPath", ["state_dict", "save_path"]) + +# (qsh 2025-01-01) the design is from https://github.com/pytorch/torchtitan/blob/1060feacc1b51cb6b339a04e53a5243b8466552b/torchtitan/checkpoint.py +# we recreate wrapper when needed instead of creating one from the beginning. +# to people who find it difficult to digest the code, official tutorial for torch dcp may be helpful + + +def dcp_load_state_dict(_state_dict, storage_reader, load_planner): + dcp.load( + _state_dict, + storage_reader=storage_reader, + planner=load_planner, + ) + # Check for missing and unexpected keys by comparing with checkpoint metadata + missing_keys = [] + if hasattr(load_planner, "metadata") and load_planner.metadata is not None: + checkpoint_keys = set(load_planner.metadata.state_dict_metadata.keys()) + model_keys = set(_state_dict.keys()) + missing_keys = list(model_keys - checkpoint_keys) + unexpected_keys = list(checkpoint_keys - model_keys) + # Log missing keys if any are found + if missing_keys: + # Only log keys in blocks.0 since other blocks are the same as blocks.0 + missing_keys = [key for key in missing_keys if "blocks.0" in key or "blocks." not in key] + missing_keys_str = "\n".join(sorted(set(".".join(k.split(".")[:10]) for k in missing_keys))) + log.critical(f"Missing keys in pretrained model: {missing_keys_str}") + if unexpected_keys: + unexpected_keys = [key for key in unexpected_keys if "_extra_state" not in key] + unexpected_keys = [key for key in unexpected_keys if "blocks.0" in key or "blocks." not in key] + unexpected_keys_str = "\n".join(sorted(set(".".join(k.split(".")[:10]) for k in unexpected_keys))) + log.critical(f"Unexpected keys in pretrained model: {unexpected_keys_str}") + + +class ModelWrapper(Stateful): + """Wrapper for model state dict handling""" + + def __init__(self, model: Union[nn.Module, List[nn.Module]], load_ema_to_reg: bool = False): + self.model = [model] if isinstance(model, nn.Module) else model + self.load_ema_to_reg = load_ema_to_reg + if self.load_ema_to_reg: + supported_model_types = [] + from cosmos_policy._src.predict2.models.text2world_model import DiffusionModel as predict2_DiffusionModel + + supported_model_types.append(predict2_DiffusionModel) + from cosmos_policy._src.predict2.models.text2world_model_rectified_flow import ( + Text2WorldModelRectifiedFlow as predict2_DiffusionModel_rectified_flow, + ) + + supported_model_types.append(predict2_DiffusionModel_rectified_flow) + from cosmos_policy._src.predict2.models.text2world_wan2pt1_model import ( + WANDiffusionModel as wan2pt1_DiffusionModel, + ) + + supported_model_types.append(wan2pt1_DiffusionModel) + + assert any(isinstance(model, cls) for cls in supported_model_types), ( + f"ModelWrapper only supports DiffusionModel when load_ema_to_reg is True, but got {type(model)}" + ) + + def state_dict(self, mapping_keys: dict[str, str] = {}) -> Dict[str, Any]: + _state_dict = {k: v for sd in map(get_model_state_dict, self.model) for k, v in sd.items()} + if self.load_ema_to_reg: + assert not self.model[0].config.ema.enabled, ( + "EMA is enabled, can not load EMA weights to regular model weights" + ) + all_keys = list(_state_dict.keys()) + assert all(k.startswith("net.") for k in all_keys), "All keys must start with net." + for k in all_keys: + _state_dict[k.replace("net.", "net_ema.")] = _state_dict.pop(k) + + if hasattr(self.model[0].config, "use_lora") and self.model[0].config.use_lora: + """ + When using LoRA, `inject_adapter_in_model` modifies the target modules in place. + For example, `blocks[0].attn.q_proj.weight` will be modified to `blocks[0].attn.q_proj.base_layer.weight`. + This means that the model will have the key `blocks[0].attn.q_proj.base_layer.weight`, + but the checkpoint will have the key `blocks[0].attn.q_proj.weight`. + We need to map the model key to the checkpoint key. + """ + self.checkpoint_to_model_key = {} + mapping_keys.update( + { + "base_layer.": "", + "base_model.model.": "", + } + ) + keys_to_update = [] + for k in _state_dict.keys(): + new_key = k + for from_key, to_key in mapping_keys.items(): + new_key = new_key.replace(from_key, to_key) + if new_key != k: + keys_to_update.append((k, new_key)) + self.checkpoint_to_model_key[new_key] = k + for k, new_key in keys_to_update: + _state_dict[new_key] = _state_dict.pop(k) + + return _state_dict + + def load_state_dict(self, state_dict: Dict[str, Any]) -> None: + if hasattr(self.model[0].config, "use_lora") and self.model[0].config.use_lora: + if hasattr(self, "checkpoint_to_model_key"): + for checkpoint_key, model_key in self.checkpoint_to_model_key.items(): + state_dict[model_key] = state_dict.pop(checkpoint_key) + else: + raise ValueError("checkpoint_to_model_key is not set by `state_dict`") + if self.load_ema_to_reg: + assert not self.model[0].config.ema.enabled, ( + "EMA is enabled, can not load EMA weights to regular model weights" + ) + all_keys = list(state_dict.keys()) + assert all(k.startswith("net_ema.") for k in all_keys), "All keys must start with net_ema." + for k in all_keys: + state_dict[k.replace("net_ema.", "net.")] = state_dict.pop(k) + + func = functools.partial( + set_model_state_dict, + model_state_dict=state_dict, + options=StateDictOptions(strict=False), + ) + list(map(func, self.model)) + + +class OptimizerWrapper(Stateful): + def __init__( + self, + model: Union[nn.Module, List[nn.Module]], + optim: Union[torch.optim.Optimizer, List[torch.optim.Optimizer]], + ) -> None: + self.model = [model] if isinstance(model, nn.Module) else model + self.optim = [optim] if isinstance(optim, torch.optim.Optimizer) else optim + + def state_dict(self) -> Dict[str, Any]: + func = functools.partial( + get_optimizer_state_dict, + options=StateDictOptions(flatten_optimizer_state_dict=True), + ) + return {k: v for sd in map(func, self.model, self.optim) for k, v in sd.items()} + + def load_state_dict(self, state_dict: Dict[str, Any]) -> None: + func = functools.partial( + set_optimizer_state_dict, + optim_state_dict=state_dict, + options=StateDictOptions(flatten_optimizer_state_dict=True), + ) + list(map(func, self.model, self.optim)) + + +class AsyncMode(str, enum.Enum): + DISABLED = "disabled" + ASYNC_WITH_PINNED_MEM = "async_with_pinned_mem" + + +class Terminate: + pass + + +class SaveDone: + def __init__(self, iteration: int, elapsed_time: float, succeeded: bool): + self.iteration = iteration + self.elapsed_time = elapsed_time + self.succeeded = succeeded + + def __str__(self): + return f"SaveDone(iteration={self.iteration}, elapsed_time={self.elapsed_time}, succeeded={self.succeeded})" + + +def save_checkpoint_in_background( + receiver_queue: multiprocessing.Queue, + sender_queue: multiprocessing.Queue, + checkpoint_config: CheckpointConfig, + job_config: JobConfig, +) -> None: + """ + Handles model checkpoint saving in a separate background process using PyTorch's distributed functionality. + This function runs in a dedicated process to avoid blocking the main training loop. + + Args: + receiver_queue: Queue to receive state dictionaries and commands from the main process + sender_queue: Queue to send completion signals back to the main process + checkpoint_config: Configuration settings for checkpoint saving behavior + job_config: Configuration settings for the training job + + Flow: + 1. Initializes distributed processing environment + 2. Continuously waits for state dictionaries to save + 3. Saves checkpoints asynchronously + 4. Signals completion back to main process + 5. Terminates when receiving a Terminate signal + + Raises: + AssertionError: If received object is neither Terminate signal nor valid state dict tuple + + Note: + - Uses a different port than the main process to avoid conflicts + - Disables TorchElastic agent store for checkpoint operations + - Automatically cleans up distributed process group on exit + """ + # Configure distributed environment + os.environ["MASTER_PORT"] = str(int(os.environ["MASTER_PORT"]) + 2) + os.environ["TORCHELASTIC_USE_AGENT_STORE"] = "False" + + # Set up GPU device and distributed processing + torch.cuda.set_device(int(os.environ["LOCAL_RANK"])) + distributed.init() + + # Initialize checkpointing mechanism + checkpoint_handler = DistributedCheckpointer(checkpoint_config, job_config, None, disable_async=True) + + try: + while True: + log.debug("Checkpoint background process is ready for next task, waiting for new state_dict") + received_data = receiver_queue.get() + log.debug("Received new state_dict") + + if isinstance(received_data, Terminate): + log.info("Received termination signal in checkpoint background process, closing sender queue") + sender_queue.put(Terminate()) + sender_queue.close() + return + + assert isinstance(received_data, tuple), "Received data must be a tuple of (state_dict, checkpoint_path)" + state_dict, checkpoint_path = received_data + + # Save checkpoint and measure time taken + start_time = time.monotonic() + iteration = state_dict["trainer"][0]["iteration"] + elapsed_time = 0 + succeeded = False + try: + checkpoint_handler.save_state_dict_worker(state_dict, checkpoint_path) + elapsed_time = time.monotonic() - start_time + log.info( + f"Checkpoint saved successfully in background process. Time taken: {elapsed_time:.2f} seconds, iteration: {iteration}" + ) + succeeded = True + except Exception as e: + log.error(f"Error saving checkpoint to {checkpoint_path}: {e}") + # continue because if the thread exits, the main thread keeps on adding to the queue + finally: + if elapsed_time == 0: + elapsed_time = time.monotonic() - start_time + sender_queue.put(SaveDone(iteration, elapsed_time, succeeded)) + + finally: + log.info("Cleaning up: destroying distributed process group") + torch.distributed.destroy_process_group() + + +class DistributedCheckpointer(AbstractCheckpointer): + KEYS_TO_SAVE = ["model", "optim", "scheduler", "trainer"] + + def __init__( + self, + config_checkpoint: CheckpointConfig, + config_job: JobConfig, + callbacks: Optional[callback.CallBackGroup] = None, + disable_async: bool = False, + ): + super().__init__(config_checkpoint, config_job, callbacks) + self.config_checkpoint = config_checkpoint + if config_checkpoint.dcp_async_mode_enabled: + self.async_mode = AsyncMode.ASYNC_WITH_PINNED_MEM + else: + self.async_mode = AsyncMode.DISABLED + + if disable_async: + self.async_mode = AsyncMode.DISABLED + + if self.async_mode == AsyncMode.ASYNC_WITH_PINNED_MEM: + ctx = get_context("spawn") + self.mp_queue_send = ctx.Queue() + self.mp_queue_recv = ctx.Queue() + self.mp = ctx.Process( + target=save_checkpoint_in_background, + args=( + self.mp_queue_send, + self.mp_queue_recv, + config_checkpoint, + config_job, + ), + daemon=True, + ) + self.mp.start() + self.cpu_offload_state_dict = None + self.staging = False + self.staging_ckpt_file = None + self.staging_stream = torch.cuda.Stream() + + def keys_to_resume_during_load(self) -> Tuple[Set, Union[str, None]]: + latest_checkpoint_file = self._read_latest_checkpoint_file() + + resume_keys = [] + + if latest_checkpoint_file is not None: + # 1. Resume training from latest_checkpoint.txt under the same name. + checkpoint_path = os.path.join(self.load_dirname, latest_checkpoint_file) + resume_keys.extend(self.KEYS_TO_SAVE) + else: + if self.load_path and not str(self.load_path).endswith(".pt"): + # 2. Load the module weights specified by config_checkpoint.path. + checkpoint_path = self.load_path + if self.load_s3_backend_key: + checkpoint_path = f"s3://{self.config_checkpoint.load_from_object_store.bucket}/{checkpoint_path}" + if not re.search(r"/checkpoints/iter_\d{9}/?$", checkpoint_path): + old_ckpt_path = checkpoint_path + # If path doesn't end with specific checkpoint, read latest checkpoint file + latest_ckpt_path = os.path.join(checkpoint_path, "checkpoints/latest_checkpoint.txt") + if easy_io.exists(latest_ckpt_path, backend_key=self.load_s3_backend_key): + checkpoint_file = easy_io.load( + latest_ckpt_path, backend_key=self.load_s3_backend_key + ).strip() + checkpoint_path = f"{checkpoint_path}/checkpoints/{checkpoint_file}" + else: + log.warning( + f"Latest checkpoint file {latest_ckpt_path} not found, load from {old_ckpt_path}" + ) + checkpoint_path = old_ckpt_path + + if self.load_training_state: + resume_keys.extend(self.KEYS_TO_SAVE) + else: + resume_keys.append("model") + if self.only_load_scheduler_state: + resume_keys.append("scheduler") + else: + checkpoint_path = None + if len(self.keys_not_to_resume) > 0: + for key in self.keys_not_to_resume: + assert key in self.KEYS_TO_SAVE, f"Invalid key to resume: {key} not in {self.KEYS_TO_SAVE}" + resume_keys = [key for key in resume_keys if key not in self.keys_not_to_resume] + return set(resume_keys), checkpoint_path + + @misc.timer("checkpoint loading") + def load( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer | None = None, + scheduler: torch.optim.lr_scheduler.LRScheduler | None = None, + grad_scaler: torch.amp.GradScaler | None = None, + ) -> int: + if self.callbacks is not None: + self.callbacks.on_load_checkpoint_start(model) + + resume_keys, checkpoint_path = self.keys_to_resume_during_load() + resume_keys = sorted(resume_keys) + log.critical(f"Resuming ckpt {checkpoint_path} with keys: {resume_keys}") + + iteration = 0 + + if checkpoint_path is not None: + self._check_checkpoint_exists(checkpoint_path) + for key in resume_keys: + load_planner = DefaultLoadPlanner(allow_partial_load=True) + if hasattr(load_planner, "set_partial_channel_weight"): + log.critical(f"set_partial_channel_weight: {self.config_checkpoint.dcp_allow_mismatched_size}") + load_planner.set_partial_channel_weight(self.config_checkpoint.dcp_allow_mismatched_size) + cur_key_ckpt_full_path = os.path.join(checkpoint_path, key) + log.critical(f"Start loading checkpoint from {checkpoint_path}") + storage_reader = self.get_storage_reader(cur_key_ckpt_full_path) + torch.distributed.barrier() + log.critical(f"starting {cur_key_ckpt_full_path}", rank0_only=False) + if key == "model": + log.info("- Loading the model...") + _model_wrapper = ModelWrapper(model) + _state_dict = _model_wrapper.state_dict() + + dcp_load_state_dict(_state_dict, storage_reader, load_planner) + _model_wrapper.load_state_dict(_state_dict) + elif key == "optim": + log.info("- Loading the optimizer...") + _optim_wrapper = OptimizerWrapper(model, optimizer) + _state_dict = _optim_wrapper.state_dict() + dcp.load( + _state_dict, + storage_reader=storage_reader, + planner=load_planner, + ) + _optim_wrapper.load_state_dict(_state_dict) + elif key == "scheduler": + log.info("- Loading the scheduler...") + _state_dict = scheduler.state_dict() + dcp.load( + _state_dict, + storage_reader=storage_reader, + planner=load_planner, + ) + scheduler.load_state_dict(_state_dict) + elif key == "trainer": + log.info("- Loading the trainer...") + _state_dict = { + "grad_scaler": grad_scaler.state_dict(), + "iteration": iteration, + } + dcp.load( + _state_dict, + storage_reader=storage_reader, + planner=load_planner, + ) + grad_scaler.load_state_dict(_state_dict["grad_scaler"]) + iteration = _state_dict["iteration"] + else: + raise ValueError(f"Invalid key: {key}. not support to resume.") + if self.callbacks is not None: + self.callbacks.on_load_checkpoint(model, state_dict=_state_dict) + log.critical(f"Loaded checkpoint from {checkpoint_path} in iteration {iteration}") + else: + log.info("Training from scratch.") + torch.cuda.empty_cache() + + if self.callbacks is not None: + self.callbacks.on_load_checkpoint_end(model, iteration=iteration, checkpoint_path=checkpoint_path) + return iteration + + def _async_with_pinned_memory(self, checkpoint_file: str, state_dict: Dict[str, Tuple[Any, str]]) -> None: + try: + from torch.distributed._state_dict_utils import _copy_state_dict, _create_cpu_state_dict + except ImportError as e: + raise ImportError( + "Please install the latest PyTorch nightly to use async checkpointing with pinned memory." + ) from e + if self.cpu_offload_state_dict is None: + log.debug(f"Preparing the CPU memory, {time.monotonic()=}.:.2f") + self.cpu_offload_state_dict = _create_cpu_state_dict(state_dict, pin_memory=True, share_memory=True) + + log.debug(f"Staging the state_dict, {time.monotonic()=}.:.2f") + with torch.cuda.stream(self.staging_stream): + self.cpu_offload_state_dict = _copy_state_dict( + state_dict, + self.cpu_offload_state_dict, + non_blocking=True, + ) + self.staging = True + self.staging_ckpt_file = checkpoint_file + + self.maybe_wait_for_staging() + + def maybe_wait_for_staging(self) -> None: + if self.async_mode == AsyncMode.ASYNC_WITH_PINNED_MEM and self.staging: + if not self.staging_stream.query(): + self.staging_stream.synchronize() + + def sync_func(): + self.mp_queue_send.put_nowait((self.cpu_offload_state_dict, self.staging_ckpt_file)) + + sync_func() + self.staging = False + + def get_previous_checkpoint_results(self, wait_for: int = 0) -> None: + """Get the results of previously submitted checkpoints and pass them to callbacks if checkpoint succeeded""" + if self.async_mode == AsyncMode.ASYNC_WITH_PINNED_MEM: + try: + start_time = time.monotonic() + while not self.mp_queue_recv.empty() or wait_for > 0: + try: + ret = self.mp_queue_recv.get(timeout=1) + if isinstance(ret, Terminate): + log.info("Received termination event from checkpoint background process") + break + save_done: SaveDone = ret + log.logger.info(f"Received checkpoint save result: {save_done}") + if self.callbacks is not None and save_done.succeeded: + self.callbacks.on_save_checkpoint_success( + iteration=save_done.iteration, elapsed_time=save_done.elapsed_time + ) + except queue.Empty: + elapsed_time = time.monotonic() - start_time + if elapsed_time > wait_for: + break + except (EOFError, BrokenPipeError): + log.info("Queue was closed by checkpoint background process") + + def get_storage_writer(self, checkpoint_path: str) -> Union[S3StorageWriter, FileSystemWriter]: + if self.save_to_object_store: + return S3StorageWriter( + credential_path=self.config_checkpoint.save_to_object_store.credentials, + path=checkpoint_path, + ) + return FileSystemWriter(path=checkpoint_path) + + def get_storage_reader(self, checkpoint_path: str) -> Union[S3StorageReader, FileSystemReader]: + if self.load_from_object_store: + return S3StorageReader( + credential_path=self.config_checkpoint.load_from_object_store.credentials, + path=checkpoint_path, + ) + return FileSystemReader(checkpoint_path) + + def save_state_dict_worker(self, to_save_dict: Dict[str, Tuple[Any, str]], checkpoint_file: str) -> None: + for k, (v, full_checkpoint_path) in to_save_dict.items(): + storage_writer = self.get_storage_writer(full_checkpoint_path) + dcp.save( + v, + storage_writer=storage_writer, + planner=DefaultSavePlanner(dedup_save_to_lowest_rank=True), + ) + + if distributed.is_rank0(): + print(f"Saving last checkpoint file {checkpoint_file}") + self._write_latest_checkpoint_file(checkpoint_file) + + log.critical(f"Saved checkpoint to {os.path.join(self.save_dirname, checkpoint_file)}", rank0_only=True) + + def save( + self, + model: ImaginaireModel, + optimizer: torch.optim.Optimizer, + scheduler: torch.optim.lr_scheduler.LRScheduler, + grad_scaler: torch.amp.GradScaler, + iteration: int, + ) -> None: + """Save network weights, optimizer parameters, scheduler parameters to a checkpoint. + + Args: + model (ImaginaireModel): The PyTorch model. + optimizer (torch.optim.Optimizer): The model optimizer. + scheduler (torch.optim.lr_scheduler.LRScheduler): The optimization scheduler. + grad_scaler (torch.amp.GradScaler): The gradient scaler (for mixed precision training). + iteration (int): Current iteration number. + """ + if self.async_mode == AsyncMode.ASYNC_WITH_PINNED_MEM: + self.get_previous_checkpoint_results(wait_for=0) + + if self.callbacks is not None: + self.callbacks.on_save_checkpoint_start(model, iteration) + + checkpoint_file = f"iter_{iteration:09}" + to_save_dict = { + "model": ModelWrapper(model).state_dict(), + "optim": OptimizerWrapper(model, optimizer).state_dict(), + "scheduler": scheduler.state_dict(), + "trainer": { + "grad_scaler": grad_scaler.state_dict(), + "iteration": iteration, + }, + } + for k in to_save_dict.keys(): + output_dirname = os.path.join(self.save_dirname, f"iter_{iteration:09}/{k}") + to_save_dict[k] = (to_save_dict[k], output_dirname) + + if self.callbacks is not None: + self.callbacks.on_save_checkpoint(model, state_dict=to_save_dict) + + if self.async_mode == AsyncMode.ASYNC_WITH_PINNED_MEM: + self._async_with_pinned_memory(checkpoint_file, to_save_dict) + else: + start_time = time.monotonic() + try: + self.save_state_dict_worker(to_save_dict, checkpoint_file) + finally: + if self.callbacks is not None: + self.callbacks.on_save_checkpoint_success( + iteration=iteration, elapsed_time=time.monotonic() - start_time + ) + + # This measures exposed (synchronous) checkpoint time, on_save_checkpoint_success() + # is instead called to measure the entire duration for asynchronous checkpoint for the async case too. + if self.callbacks is not None: + self.callbacks.on_save_checkpoint_end(model=None, iteration=iteration) + + def finalize(self) -> None: + super().finalize() + if self.async_mode == AsyncMode.ASYNC_WITH_PINNED_MEM: + if self.mp and self.mp.is_alive(): + self.mp_queue_send.put(Terminate()) + self.get_previous_checkpoint_results(wait_for=60) + self.mp.join() diff --git a/REGEN-main/cosmos_policy/_src/predict2/conditioner.py b/REGEN-main/cosmos_policy/_src/predict2/conditioner.py new file mode 100644 index 0000000000000000000000000000000000000000..433dfa9d7c66a380ed5c31407b11cd18289054a7 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/conditioner.py @@ -0,0 +1,575 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import copy +from abc import ABC, abstractmethod +from collections import defaultdict +from contextlib import nullcontext +from dataclasses import dataclass, fields +from enum import Enum +from typing import Any, Dict, List, Optional, Tuple, TypeVar, Union + +import omegaconf +import torch +import torch.nn as nn +from torch.distributed import ProcessGroup + +from cosmos_policy._src.imaginaire.functional.batch_ops import batch_mul +from cosmos_policy._src.imaginaire.lazy_config import instantiate +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.context_parallel import broadcast +from cosmos_policy._src.imaginaire.utils.count_params import count_params +from cosmos_policy._src.imaginaire.utils.disabled_train import disabled_train +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + +T = TypeVar("T", bound="BaseCondition") + + +class DataType(str, Enum): + IMAGE = "image" + VIDEO = "video" + MIX = "mix" + + def __str__(self) -> str: + return self.value + + +def broadcast_condition(condition: BaseCondition, process_group: Optional[ProcessGroup] = None) -> BaseCondition: + """ + Broadcast the condition from the minimum rank in the specified group(s). + """ + if condition.is_broadcasted: + return condition + + kwargs = condition.to_dict(skip_underscore=False) + for key, value in kwargs.items(): + if value is not None: + if isinstance(value, torch.Tensor): + value = value.cuda() + kwargs[key] = broadcast(value, process_group) + kwargs["_is_broadcasted"] = True + return type(condition)(**kwargs) + + +@dataclass(frozen=True) +class BaseCondition(ABC): + """ + Attributes: + _is_broadcasted: Flag indicating if parallel broadcast splitting + has been performed. This is an internal implementation detail. + """ + + _is_broadcasted: bool = False + + def to_dict(self, skip_underscore: bool = True) -> Dict[str, Any]: + """Converts the condition to a dictionary. + + Returns: + Dictionary containing the condition's fields and values. + """ + # return {f.name: getattr(self, f.name) for f in fields(self) if not f.name.startswith("_")} + return {f.name: getattr(self, f.name) for f in fields(self) if not (f.name.startswith("_") and skip_underscore)} + + @property + def is_broadcasted(self) -> bool: + return self._is_broadcasted + + def broadcast(self, process_group: torch.distributed.ProcessGroup) -> BaseCondition: + """Broadcasts and splits the condition across the checkpoint parallelism group. + For most condition, such as Text2WorldCondition, we do not need split. + + Args: + process_group: The process group for broadcast and split + + Returns: + A new BaseCondition instance with the broadcasted and split condition. + """ + if self.is_broadcasted: + return self + return broadcast_condition(self, process_group) + + +@dataclass(frozen=True) +class Text2WorldCondition(BaseCondition): + crossattn_emb: Optional[torch.Tensor] = None + data_type: DataType = DataType.VIDEO + padding_mask: Optional[torch.Tensor] = None + fps: Optional[torch.Tensor] = None + + def edit_data_type(self, data_type: DataType) -> Text2WorldCondition: + """Edit the data type of the condition. + + Args: + data_type: The new data type. + + Returns: + A new Text2WorldCondition instance with the new data type. + """ + kwargs = self.to_dict(skip_underscore=False) + kwargs["data_type"] = data_type + return type(self)(**kwargs) + + @property + def is_video(self) -> bool: + return self.data_type == DataType.VIDEO + + +@dataclass(frozen=True) +class GR00TV1Img2VidCondition(Text2WorldCondition): + gt_first_frame: Optional[torch.Tensor] = None + use_image_condition: bool = False + condition_video_input_mask_B_C_T_H_W: Optional[torch.Tensor] = None + + def edit_video_condition( + self, x0_B_C_T_H_W, process_group: Optional[ProcessGroup] = None + ) -> GR00TV1Img2VidCondition: + """Edit the video condition to include the video mask information. + + Args: + x0_B_C_T_H_W: The first frame of the video. + + Returns: + A new GR00TV1Img2VidCondition instance with the video mask information. + """ + pg_size = 1 if process_group is None else process_group.size() + kwargs = self.to_dict(skip_underscore=False) + B, _, T, H, W = x0_B_C_T_H_W.shape + condition_video_input_mask = torch.zeros((B, 1, T, H, W), dtype=x0_B_C_T_H_W.dtype, device=x0_B_C_T_H_W.device) + if pg_size == 1 or process_group.rank() == 0: + kwargs["gt_first_frame"] = x0_B_C_T_H_W[:, :, 0].detach() + condition_video_input_mask[:, :, 0] += 1 + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask + return type(self)(**kwargs) + + +class AbstractEmbModel(nn.Module): + def __init__(self): + super().__init__() + + self._is_trainable = None + self._dropout_rate = None + self._input_key = None + + self._return_dict = False + + @property + def is_trainable(self) -> bool: + return self._is_trainable + + @property + def dropout_rate(self) -> Union[float, torch.Tensor]: + return self._dropout_rate + + @property + def input_key(self) -> str: + return self._input_key + + @property + def is_return_dict(self) -> bool: + return self._return_dict + + @is_trainable.setter + def is_trainable(self, value: bool): + self._is_trainable = value + + @dropout_rate.setter + def dropout_rate(self, value: Union[float, torch.Tensor]): + self._dropout_rate = value + + @input_key.setter + def input_key(self, value: str): + self._input_key = value + + @is_return_dict.setter + def is_return_dict(self, value: bool): + self._return_dict = value + + @is_trainable.deleter + def is_trainable(self): + del self._is_trainable + + @dropout_rate.deleter + def dropout_rate(self): + del self._dropout_rate + + @input_key.deleter + def input_key(self): + del self._input_key + + @is_return_dict.deleter + def is_return_dict(self): + del self._return_dict + + def random_dropout_input( + self, in_tensor: torch.Tensor, dropout_rate: Optional[float] = None, key: Optional[str] = None + ) -> torch.Tensor: + del key + dropout_rate = dropout_rate if dropout_rate is not None else self.dropout_rate + return batch_mul( + torch.bernoulli((1.0 - dropout_rate) * torch.ones(in_tensor.shape[0])).type_as(in_tensor), + in_tensor, + ) + + def details(self) -> str: + return "" + + def summary(self) -> str: + input_key = self.input_key if self.input_key is not None else getattr(self, "input_keys", None) + return ( + f"{self.__class__.__name__} \n\tinput key: {input_key}" + f"\n\tParam count: {count_params(self, False)} \n\tTrainable: {self.is_trainable}" + f"\n\tDropout rate: {self.dropout_rate}" + f"\n\t{self.details()}" + ) + + +class TextAttr(AbstractEmbModel): + def __init__( + self, + input_key: List[str], + dropout_rate: Optional[float] = 0.0, + use_empty_string: bool = False, + empty_string_embeddings_path: str = "s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt", + credential_path: str = "credentials/s3_training.secret", + ): + super().__init__() + self._input_key = input_key + self._dropout_rate = dropout_rate + # if True, will use empty string embeddings + # otherwise use zero tensor embeddings + self.use_empty_string = use_empty_string + self._empty_string_embeddings_cache = None + self.empty_string_embeddings_path = empty_string_embeddings_path + self.credential_path = credential_path + + def forward(self, token: torch.Tensor): + return {"crossattn_emb": token} + + def _get_empty_string_embeddings(self) -> torch.Tensor: + """Lazy load and cache empty string embeddings.""" + if self._empty_string_embeddings_cache is None: + self._empty_string_embeddings_cache = easy_io.load( + self.empty_string_embeddings_path, + backend_args={"backend": "s3", "s3_credential_path": self.credential_path}, + ) + return self._empty_string_embeddings_cache + + def random_dropout_input( + self, in_tensor: torch.Tensor, dropout_rate: Optional[float] = None, key: Optional[str] = None + ) -> torch.Tensor: + if key is not None and "mask" in key: + return in_tensor + if not self.use_empty_string: + return super().random_dropout_input(in_tensor, dropout_rate, key) + B = in_tensor.shape[0] + dropout_rate = dropout_rate if dropout_rate is not None else self.dropout_rate + empty_string_embeddings = self._get_empty_string_embeddings() + empty_string_embeddings = empty_string_embeddings.expand(in_tensor.shape).to( + dtype=in_tensor.dtype, device=in_tensor.device + ) + + keep_mask = torch.bernoulli((1.0 - dropout_rate) * torch.ones(B, device=in_tensor.device)).type_as(in_tensor) + keep_mask = keep_mask.view(B, *[1] * (in_tensor.dim() - 1)) # broadcastable shape + return keep_mask * in_tensor + (1.0 - keep_mask) * empty_string_embeddings + + def details(self) -> str: + return "Output key: [crossattn_emb]" + + +class TextAttrEmptyStringDrop(AbstractEmbModel): + def __init__(self, input_key: List[str], dropout_rate: Optional[float] = 0.0): + super().__init__() + self._input_key = input_key + self._dropout_rate = dropout_rate + self.empty_prompt_data = None + + def forward(self, token: torch.Tensor): + return {"crossattn_emb": token} + + def random_dropout_input( + self, in_tensor: torch.Tensor, dropout_rate: Optional[float] = None, key: Optional[str] = None + ) -> torch.Tensor: + if key is not None and "mask" in key: + return in_tensor + del key + if self.empty_prompt_data is None: + self.empty_prompt_data = easy_io.load( + "s3://bucket/edify_video/v4/validation/item_dataset/negative_prompt/empty_string_umt5.pt", + backend_args={"backend": "s3", "s3_credential_path": "credentials/s3_training.secret"}, + ) + dropout_rate = dropout_rate if dropout_rate is not None else self.dropout_rate + + B = in_tensor.shape[0] # batch size + # Create dropout mask: 1 -> keep in_tensor, 0 -> use empty_prompt_data + keep_mask = torch.bernoulli((1.0 - dropout_rate) * torch.ones(B, device=in_tensor.device)).type_as(in_tensor) + keep_mask = keep_mask.view(B, *[1] * (in_tensor.dim() - 1)) # broadcastable shape + # Prepare empty_prompt_data with correct shape, dtype, and device + empty_prompt = self.empty_prompt_data.to(dtype=in_tensor.dtype, device=in_tensor.device) + # Repeat empty_prompt along batch dimension if needed + if empty_prompt.shape[0] != B: + if empty_prompt.shape[0] == 1: + empty_prompt = empty_prompt.expand(B, *empty_prompt.shape[1:]) + else: + raise ValueError( + f"empty_prompt_data batch size {empty_prompt.shape[0]} does not match in_tensor batch size {B}" + ) + + # Mix using the dropout mask + return keep_mask * in_tensor + (1.0 - keep_mask) * empty_prompt + + def details(self) -> str: + return "Output key: [crossattn_emb]" + + +class ReMapkey(AbstractEmbModel): + def __init__( + self, + input_key: str, + output_key: Optional[str] = None, + dropout_rate: Optional[float] = 0.0, + dtype: Optional[str] = None, + ): + super().__init__() + self.output_key = output_key + self.dtype = { + None: None, + "float": torch.float32, + "bfloat16": torch.bfloat16, + "half": torch.float16, + "float16": torch.float16, + "int": torch.int32, + "long": torch.int64, + }[dtype] + self._input_key = input_key + self._output_key = output_key + self._dropout_rate = dropout_rate + + def forward(self, element: torch.Tensor) -> Dict[str, torch.Tensor]: + key = self.output_key if self.output_key else self.input_key + if isinstance(element, torch.Tensor): + element = element.to(dtype=self.dtype) + return {key: element} + + def details(self) -> str: + key = self.output_key if self.output_key else self.input_key + return f"Output key: {key} \n\tDtype: {self.dtype}" + + +class BooleanFlag(AbstractEmbModel): + def __init__(self, input_key: str, output_key: Optional[str] = None, dropout_rate: Optional[float] = 0.0): + super().__init__() + self._input_key = input_key + self._dropout_rate = dropout_rate + self.output_key = output_key + + def forward(self, *args, **kwargs) -> Dict[str, torch.Tensor]: + del args, kwargs + key = self.output_key if self.output_key else self.input_key + return {key: self.flag} + + def random_dropout_input( + self, in_tensor: torch.Tensor, dropout_rate: Optional[float] = None, key: Optional[str] = None + ) -> torch.Tensor: + del key + dropout_rate = dropout_rate if dropout_rate is not None else self.dropout_rate + self.flag = torch.bernoulli((1.0 - dropout_rate) * torch.ones(1)).bool().to(device=in_tensor.device) + return in_tensor + + def details(self) -> str: + key = self.output_key if self.output_key else self.input_key + return f"Output key: {key} \n\t This is a boolean flag" + + +class GeneralConditioner(nn.Module, ABC): + """ + An abstract module designed to handle various embedding models with conditional and unconditional configurations. + This abstract base class initializes and manages a collection of embedders that can dynamically adjust + their dropout rates based on conditioning. + + Attributes: + KEY2DIM (dict): A mapping from output keys to dimensions used for concatenation. + embedders (nn.ModuleDict): A dictionary containing all embedded models initialized and configured + based on the provided configurations. + + Parameters: + emb_models (Union[List, Any]): A dictionary where keys are embedder names and values are configurations + for initializing the embedders. + + Example: + See Edify4ConditionerConfig + """ + + KEY2DIM = {"crossattn_emb": 1} + + def __init__(self, **emb_models: Union[List, Any]): + super().__init__() + self.embedders = nn.ModuleDict() + for n, (emb_name, emb_config) in enumerate(emb_models.items()): + embedder = instantiate(emb_config) + assert isinstance(embedder, AbstractEmbModel), ( + f"embedder model {embedder.__class__.__name__} has to inherit from AbstractEmbModel" + ) + embedder.is_trainable = getattr(emb_config, "is_trainable", True) + embedder.dropout_rate = getattr(emb_config, "dropout_rate", 0.0) + if not embedder.is_trainable: + embedder.train = disabled_train + for param in embedder.parameters(): + param.requires_grad = False + embedder.eval() + + log.info(f"Initialized embedder #{n}-{emb_name}: \n {embedder.summary()}") + self.embedders[emb_name] = embedder + + @abstractmethod + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> Any: + """Should be implemented in subclasses to handle conditon datatype""" + raise NotImplementedError + + def _forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> Dict: + """ + Processes the input batch through all configured embedders, applying conditional dropout rates if specified. + Output tensors for each key are concatenated along the dimensions specified in KEY2DIM. + + Parameters: + batch (Dict): The input data batch to process. + override_dropout_rate (Optional[Dict[str, float]]): Optional dictionary to override default dropout rates + per embedder key. + + Returns: + Dict: A dictionary of output tensors concatenated by specified dimensions. + + Note: + In case the network code is sensitive to the order of concatenation, you can either control the order via \ + config file or make sure the embedders return a unique key for each output. + """ + output = defaultdict(list) + if override_dropout_rate is None: + override_dropout_rate = {} + + # make sure emb_name in override_dropout_rate is valid + for emb_name in override_dropout_rate.keys(): + assert emb_name in self.embedders, f"invalid name found {emb_name}" + + for emb_name, embedder in self.embedders.items(): + embedding_context = nullcontext if embedder.is_trainable else torch.no_grad + with embedding_context(): + if isinstance(embedder.input_key, str): + emb_out = embedder( + embedder.random_dropout_input( + batch[embedder.input_key], override_dropout_rate.get(emb_name, None) + ) + ) + elif isinstance(embedder.input_key, (list, omegaconf.listconfig.ListConfig)): + emb_out = embedder( + *[ + embedder.random_dropout_input(batch.get(k), override_dropout_rate.get(emb_name, None), k) + for k in embedder.input_key + ] + ) + else: + raise KeyError( + f"Embedder '{embedder.__class__.__name__}' requires an 'input_key' attribute to be defined as either a string or list of strings" + ) + for k, v in emb_out.items(): + output[k].append(v) + # Concatenate the outputs + return {k: torch.cat(v, dim=self.KEY2DIM.get(k, -1)) for k, v in output.items()} + + def get_condition_uncondition( + self, + data_batch: Dict, + ) -> Tuple[Any, Any]: + """ + Processes the provided data batch to generate two sets of outputs: conditioned and unconditioned. This method + manipulates the dropout rates of embedders to simulate two scenarios — one where all conditions are applied + (conditioned), and one where they are removed or reduced to the minimum (unconditioned). + + This method first sets the dropout rates to zero for the conditioned scenario to fully apply the embedders' effects. + For the unconditioned scenario, it sets the dropout rates to 1 (or to 0 if the initial unconditional dropout rate + is insignificant) to minimize the embedders' influences, simulating an unconditioned generation. + + Parameters: + data_batch (Dict): The input data batch that contains all necessary information for embedding processing. The + data is expected to match the required format and keys expected by the embedders. + + Returns: + Tuple[Any, Any]: A tuple containing two condition: + - The first one contains the outputs with all embedders fully applied (conditioned outputs). + - The second one contains the outputs with embedders minimized or not applied (unconditioned outputs). + """ + cond_dropout_rates, dropout_rates = {}, {} + for emb_name, embedder in self.embedders.items(): + cond_dropout_rates[emb_name] = 0.0 + dropout_rates[emb_name] = 1.0 if embedder.dropout_rate > 1e-4 else 0.0 + + condition: Any = self(data_batch, override_dropout_rate=cond_dropout_rates) + un_condition: Any = self(data_batch, override_dropout_rate=dropout_rates) + return condition, un_condition + + def get_condition_with_negative_prompt( + self, + data_batch: Dict, + ) -> Tuple[Any, Any]: + """ + Similar functionality as get_condition_uncondition + But use negative prompts for unconditon + """ + cond_dropout_rates, uncond_dropout_rates = {}, {} + for emb_name, embedder in self.embedders.items(): + cond_dropout_rates[emb_name] = 0.0 + if isinstance(embedder, TextAttr): + uncond_dropout_rates[emb_name] = 0.0 + else: + uncond_dropout_rates[emb_name] = 1.0 if embedder.dropout_rate > 1e-4 else 0.0 + + data_batch_neg_prompt = copy.deepcopy(data_batch) + if "neg_t5_text_embeddings" in data_batch_neg_prompt: + if isinstance(data_batch_neg_prompt["neg_t5_text_embeddings"], torch.Tensor): + data_batch_neg_prompt["t5_text_embeddings"] = data_batch_neg_prompt["neg_t5_text_embeddings"] + + condition: Any = self(data_batch, override_dropout_rate=cond_dropout_rates) + un_condition: Any = self(data_batch_neg_prompt, override_dropout_rate=uncond_dropout_rates) + + return condition, un_condition + + +class VideoConditioner(GeneralConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> Text2WorldCondition: + output = super()._forward(batch, override_dropout_rate) + return Text2WorldCondition(**output) + + +class GR00TV1Img2VidConditioner(GeneralConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> GR00TV1Img2VidCondition: + output = super()._forward(batch, override_dropout_rate) + return GR00TV1Img2VidCondition(**output) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/callbacks.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/callbacks.py new file mode 100644 index 0000000000000000000000000000000000000000..33dbef729e33a3227071dcc51697ab8fc4b40495 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/callbacks.py @@ -0,0 +1,72 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from cosmos_policy._src.imaginaire.callbacks.manual_gc import ManualGarbageCollection +from cosmos_policy._src.imaginaire.lazy_config import PLACEHOLDER +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.utils.callback import LowPrecisionCallback +from cosmos_policy._src.predict2.callbacks.compile_tokenizer import CompileTokenizer +from cosmos_policy._src.predict2.callbacks.dataloading_monitor import DetailedDataLoadingSpeedMonitor +from cosmos_policy._src.predict2.callbacks.device_monitor import DeviceMonitor +from cosmos_policy._src.predict2.callbacks.grad_clip import GradClip +from cosmos_policy._src.predict2.callbacks.heart_beat import HeartBeat +from cosmos_policy._src.predict2.callbacks.iter_speed import IterSpeed +from cosmos_policy._src.predict2.callbacks.wandb_log import WandbCallback + +BASIC_CALLBACKS = dict( + grad_clip=L(GradClip)(), + low_prec=L(LowPrecisionCallback)(config=PLACEHOLDER, trainer=PLACEHOLDER, update_iter=1), + iter_speed=L(IterSpeed)( + every_n="${trainer.logging_iter}", + save_s3="${upload_reproducible_setup}", + save_s3_every_log_n=10, + ), + heart_beat=L(HeartBeat)( + every_n=10, + update_interval_in_minute=20, + save_s3="${upload_reproducible_setup}", + ), + device_monitor=L(DeviceMonitor)( + every_n="${trainer.logging_iter}", + save_s3="${upload_reproducible_setup}", + upload_every_n_mul=10, + ), + manual_gc=L(ManualGarbageCollection)(every_n=5), + compile_tokenizer=L(CompileTokenizer)( + enabled=True, + compile_after_iterations=4, + dynamic=False, # If there are issues with constant recompilations you may set this value to None or True + ), +) + +WANDB_CALLBACK = dict( + wandb=L(WandbCallback)( + save_s3="${upload_reproducible_setup}", + logging_iter_multipler=1, + save_logging_iter_multipler=10, + ), + wandb_10x=L(WandbCallback)( + logging_iter_multipler=10, + save_logging_iter_multipler=1, + save_s3="${upload_reproducible_setup}", + ), +) + +SPEED_CALLBACKS = dict( + dataloader_speed=L(DetailedDataLoadingSpeedMonitor)( + every_n="${trainer.logging_iter}", + save_s3="${upload_reproducible_setup}", + ), +) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/checkpoint.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/checkpoint.py new file mode 100644 index 0000000000000000000000000000000000000000..eb23e1418bd4a0107512b61a244bfe1a8fed01c9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/checkpoint.py @@ -0,0 +1,69 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.config import CheckpointConfig, ObjectStoreConfig + +pbss_object_store = ObjectStoreConfig( + enabled=False, + credentials="credentials/pbss_checkpoint.secret", + bucket="checkpoints", +) +s3_object_store = ObjectStoreConfig( + enabled=False, + credentials="credentials/s3_checkpoint.secret", + bucket="bucket", +) +gcp_object_store = ObjectStoreConfig( + enabled=False, + credentials="credentials/gcp_checkpoint.secret", + bucket="bucket", +) + + +CHECKPOINT_PBSS = CheckpointConfig( + save_to_object_store=pbss_object_store, + save_iter=1000, + load_from_object_store=pbss_object_store, + load_path="", + load_training_state=False, + strict_resume=True, +) + +CHECKPOINT_S3 = CheckpointConfig( + save_to_object_store=s3_object_store, + save_iter=1000, + load_from_object_store=s3_object_store, + load_path="", + load_training_state=False, + strict_resume=True, +) + +CHECKPOINT_GCP = CheckpointConfig( + save_to_object_store=gcp_object_store, + save_iter=1000, + load_from_object_store=gcp_object_store, + load_path="", + load_training_state=False, + strict_resume=True, +) + + +def register_checkpoint(): + cs = ConfigStore.instance() + cs.store(group="checkpoint", package="checkpoint", name="pbss", node=CHECKPOINT_PBSS) + cs.store(group="checkpoint", package="checkpoint", name="s3", node=CHECKPOINT_S3) + cs.store(group="checkpoint", package="checkpoint", name="gcp", node=CHECKPOINT_GCP) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/ckpt_type.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/ckpt_type.py new file mode 100644 index 0000000000000000000000000000000000000000..2bffd24b5a0aa8e14c744b8c220009c51ba7654c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/ckpt_type.py @@ -0,0 +1,31 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Dict + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.checkpointer.dummy import Checkpointer as DummyCheckpointer +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.checkpointer.dcp import DistributedCheckpointer + +DUMMY_CHECKPOINTER: Dict[str, str] = L(DummyCheckpointer)() +DISTRIBUTED_CHECKPOINTER: Dict[str, str] = L(DistributedCheckpointer)() + + +def register_ckpt_type(): + cs = ConfigStore.instance() + cs.store(group="ckpt_type", package="checkpoint.type", name="dummy", node=DUMMY_CHECKPOINTER) + cs.store(group="ckpt_type", package="checkpoint.type", name="dcp", node=DISTRIBUTED_CHECKPOINTER) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/dataloader.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/dataloader.py new file mode 100644 index 0000000000000000000000000000000000000000..c595124888c0ae2801ce41a26ca1dc48f2719b90 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/dataloader.py @@ -0,0 +1,131 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import omegaconf +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.configs.common.mock_data import ( + MOCK_DATA_IMAGE_ONLY_CONFIG, + MOCK_DATA_INTERLEAVE_CONFIG, + MOCK_DATA_VIDEO_ONLY_CONFIG, +) +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import get_cached_replay_dataloader +from cosmos_policy._src.predict2.datasets.dataset_provider import get_image_dataset, get_video_dataset +from cosmos_policy._src.predict2.datasets.joint_dataloader import IterativeJointDataLoader + + +def get_image_dataloader(dataset_name: str, object_store: str) -> omegaconf.dictconfig.DictConfig: + return L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name=dataset_name, + resolution="720", + is_train=True, + object_store=object_store, + ), + num_workers=8, + prefetch_factor=4, + batch_size=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + ) + + +def get_video_dataloader(dataset_name: str, object_store: str) -> omegaconf.dictconfig.DictConfig: + return L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name=dataset_name, + video_decoder_name="chunked_video_decoder", + resolution="720", + is_train=True, + object_store=object_store, + chunk_size=256, + ), + batch_size=1, + num_workers=8, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + ) + + +def get_joint_image_video_dataloader( + image_dataset_name: str, + video_dataset_name: str, + object_store="s3", +) -> omegaconf.dictconfig.DictConfig: + image_dataloader = get_image_dataloader(dataset_name=image_dataset_name, object_store=object_store) + video_dataloader = get_video_dataloader(dataset_name=video_dataset_name, object_store=object_store) + c = L(IterativeJointDataLoader)( + dataloaders={ + "image_data": { + "dataloader": image_dataloader, + "ratio": 1, + }, + "video_data": { + "dataloader": video_dataloader, + "ratio": 1, + }, + } + ) + return c + + +def get_joint_image_two_video_dataloader( + image_dataset_name: str, + video_dataset_name: str, + object_store="s3", +) -> omegaconf.dictconfig.DictConfig: + """ + This dataloader is used for training with multi-resolution multi-fps video. + """ + image_dataloader = get_image_dataloader(dataset_name=image_dataset_name, object_store=object_store) + video_dataloader = get_video_dataloader(dataset_name=video_dataset_name, object_store=object_store) + # Why do we name it video_data video_data_1 instead of video_data_1 video_data_2? + # In our exp config, if we inherit the exp which has video_data (joint_image_video), the video_data dict will + # stay in the config, and we end up with having + # video_data: not used, stay here due to inheritance, can confuse users when checking config.yaml. This is an known issue of our config system. + # video_data_1: used + # video_data_2: used + # Therefore, we keep the same video_data entry here, and add video_data_1 + c = L(IterativeJointDataLoader)( + dataloaders={ + "image_data": { + "dataloader": image_dataloader, + "ratio": 2, + }, + "video_data": { + "dataloader": video_dataloader, + "ratio": 1, + }, + "video_data_1": { + "dataloader": video_dataloader, + "ratio": 1, + }, + } + ) + return c + + +def register_training_and_val_data(): + cs = ConfigStore() + cs.store(group="data_train", package="dataloader_train", name="mock", node=MOCK_DATA_INTERLEAVE_CONFIG) + cs.store(group="data_train", package="dataloader_train", name="mock_image", node=MOCK_DATA_IMAGE_ONLY_CONFIG) + cs.store(group="data_train", package="dataloader_train", name="mock_video", node=MOCK_DATA_VIDEO_ONLY_CONFIG) + cs.store(group="data_val", package="dataloader_val", name="mock", node=MOCK_DATA_INTERLEAVE_CONFIG) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/ema.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/ema.py new file mode 100644 index 0000000000000000000000000000000000000000..e0edbb00abde9953beb6c83510e6f2a711ff402d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/ema.py @@ -0,0 +1,29 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.predict2.models.text2world_model import EMAConfig + +PowerEMAConfig: EMAConfig = EMAConfig( + enabled=True, + rate=0.10, + iteration_shift=0, +) + + +def register_ema(): + cs = ConfigStore.instance() + cs.store(group="ema", package="model.config.ema", name="power", node=PowerEMAConfig) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/optimizer.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/optimizer.py new file mode 100644 index 0000000000000000000000000000000000000000..d470d00b35034256575b41fdb2d1c699f7f92d2a --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/optimizer.py @@ -0,0 +1,47 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import PLACEHOLDER, LazyDict +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.utils.optim_instantiate import get_base_optimizer + +AdamWConfig = L(get_base_optimizer)( + model=PLACEHOLDER, + lr=1e-4, + weight_decay=0.1, + betas=[0.9, 0.99], + optim_type="adamw", + eps=1e-8, + fused=True, +) + +FusedAdamWConfig: LazyDict = L(get_base_optimizer)( + model=PLACEHOLDER, + lr=1e-4, + weight_decay=0.1, + betas=[0.9, 0.99], + optim_type="fusedadam", + eps=1e-8, + master_weights=True, + capturable=True, +) + + +def register_optimizer(): + cs = ConfigStore.instance() + cs.store(group="optimizer", package="optimizer", name="fusedadamw", node=FusedAdamWConfig) + cs.store(group="optimizer", package="optimizer", name="adamw", node=AdamWConfig) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/scheduler.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/scheduler.py new file mode 100644 index 0000000000000000000000000000000000000000..6839994079d8a71d0501dc2ca241a2e70e96e894 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/scheduler.py @@ -0,0 +1,23 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.configs.lr_scheduler import LambdaLinearSchedulerConfig + + +def register_scheduler(): + cs = ConfigStore.instance() + cs.store(group="scheduler", package="scheduler", name="lambdalinear", node=LambdaLinearSchedulerConfig) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/tokenizer.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/tokenizer.py new file mode 100644 index 0000000000000000000000000000000000000000..a78e76144cae364f4f20dee25ddf1a2c39ab6607 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/defaults/tokenizer.py @@ -0,0 +1,33 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.predict2.tokenizers.cosmos import ( + Wan2pt1VAEConfig, + Wan2pt1VAEConfig_GCP, + Wan2pt2VAEConfig, +) + + +def register_tokenizer(): + cs = ConfigStore.instance() + + # Wan2pt1 and Wan2pt2 tokenizers + cs.store(group="tokenizer", package="model.config.tokenizer", name="wan2pt1_tokenizer", node=Wan2pt1VAEConfig) + cs.store( + group="tokenizer", package="model.config.tokenizer", name="wan2pt1_tokenizer_gcp", node=Wan2pt1VAEConfig_GCP + ) + cs.store(group="tokenizer", package="model.config.tokenizer", name="wan2pt2_tokenizer", node=Wan2pt2VAEConfig) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/common/mock_data.py b/REGEN-main/cosmos_policy/_src/predict2/configs/common/mock_data.py new file mode 100644 index 0000000000000000000000000000000000000000..0c5ef3cffbf0e301a548383aefef195d283aeeaf --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/common/mock_data.py @@ -0,0 +1,63 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import get_cached_replay_dataloader +from cosmos_policy._src.predict2.datasets.data_sources.mock_data import get_image_dataset, get_video_dataset +from cosmos_policy._src.predict2.datasets.joint_dataloader import IterativeJointDataLoader + +_IMAGE_LOADER = L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + resolution="512", + ), + batch_size=2, + shuffle=False, + num_workers=8, + pin_memory=True, + webdataset=False, + cache_replay_name="image_dataloader", +) + +_VIDEO_LOADER = L( + get_cached_replay_dataloader +)( + dataset=L(get_video_dataset)( + resolution="512", + num_video_frames=136, # number of pixel frames, the number needs to agree with tokenizer encoder since tokenizer can not handle arbitrary length + ), + batch_size=1, + shuffle=False, + num_workers=8, + pin_memory=True, + webdataset=False, + cache_replay_name="video_dataloader", +) + +MOCK_DATA_INTERLEAVE_CONFIG = L(IterativeJointDataLoader)( + dataloaders={ + "image_data": { + "dataloader": _IMAGE_LOADER, + "ratio": 1, + }, + "video_data": { + "dataloader": _VIDEO_LOADER, + "ratio": 1, + }, + } +) + +MOCK_DATA_IMAGE_ONLY_CONFIG = _IMAGE_LOADER + +MOCK_DATA_VIDEO_ONLY_CONFIG = _VIDEO_LOADER diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/config.py b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/config.py new file mode 100644 index 0000000000000000000000000000000000000000..932f945c11e7e8499c33948c9991fe0d338e456c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/config.py @@ -0,0 +1,99 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any, List + +import attrs + +from cosmos_policy._src.imaginaire import config +from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer as Trainer +from cosmos_policy._src.imaginaire.utils.config_helper import import_all_modules_from_package +from cosmos_policy._src.predict2.configs.common.defaults.checkpoint import register_checkpoint +from cosmos_policy._src.predict2.configs.common.defaults.ckpt_type import register_ckpt_type +from cosmos_policy._src.predict2.configs.common.defaults.dataloader import register_training_and_val_data +from cosmos_policy._src.predict2.configs.common.defaults.ema import register_ema +from cosmos_policy._src.predict2.configs.common.defaults.optimizer import register_optimizer +from cosmos_policy._src.predict2.configs.common.defaults.scheduler import register_scheduler +from cosmos_policy._src.predict2.configs.common.defaults.tokenizer import register_tokenizer +from cosmos_policy._src.predict2.configs.text2world.defaults.callbacks import register_callbacks +from cosmos_policy._src.predict2.configs.text2world.defaults.conditioner import register_conditioner +from cosmos_policy._src.predict2.configs.text2world.defaults.model import register_model +from cosmos_policy._src.predict2.configs.text2world.defaults.net import register_net + + +@attrs.define(slots=False) +class Config(config.Config): + # default config groups that will be used unless overwritten + # see config groups in registry.py + defaults: List[Any] = attrs.field( + factory=lambda: [ + "_self_", + {"data_train": "mock"}, + {"data_val": "mock"}, + {"optimizer": "fusedadamw"}, + {"scheduler": "lambdalinear"}, + {"model": "ddp"}, + {"callbacks": "basic"}, + {"net": None}, + {"conditioner": "add_fps_padding_mask"}, + {"ema": "power"}, + {"tokenizer": "cosmos_tokenizer_causal_cv8x8x8_c16_res720_t121_it121_v1_0"}, + {"checkpoint": "s3"}, + {"ckpt_type": "dummy"}, + # the list is with order, we need global experiment to be the last one + {"experiment": None}, + ] + ) + + +def make_config() -> Config: + c = Config( + model=None, + optimizer=None, + scheduler=None, + dataloader_train=None, + dataloader_val=None, + ) + + # Specifying values through instances of attrs + c.job.project = "cosmos_diffusion_v2" + c.job.group = "debug" + c.job.name = "delete_${now:%Y-%m-%d}_${now:%H-%M-%S}" + + c.trainer.type = Trainer + c.trainer.straggler_detection.enabled = False + c.trainer.max_iter = 400_000 + c.trainer.logging_iter = 10 + c.trainer.validation_iter = 100 + c.trainer.run_validation = False + c.trainer.callbacks = None + + # Call this function to register config groups for advanced overriding. the order follows the default config groups + register_training_and_val_data() + register_optimizer() + register_scheduler() + register_model() + register_callbacks() + register_net() + register_conditioner() + register_ema() + register_tokenizer() + register_checkpoint() + register_ckpt_type() + + # experiment config are defined in the experiment folder + # call import_all_modules_from_package to register them + import_all_modules_from_package("cosmos_policy._src.predict2.configs.text2world.experiment", reload=True) + return c diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/config_test.py b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/config_test.py new file mode 100644 index 0000000000000000000000000000000000000000..f2f12f25a3f7edc8ab2701052d0f1da9e6ba7b02 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/config_test.py @@ -0,0 +1,27 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import pytest + +from cosmos_policy._src.imaginaire.utils.config_helper import override +from cosmos_policy._src.predict2.configs.text2world.config import make_config + + +@pytest.mark.L1 +def test_make_config(): + config = make_config() + config = override(config, ["--", "experiment=error-free_fsdp_mock-data_base-cb", "trainer.max_iter=1"]) + + assert config.trainer.max_iter == 1 diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/callbacks.py b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/callbacks.py new file mode 100644 index 0000000000000000000000000000000000000000..2244c7f0b391a76100a470532aae23954f1fbeca --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/callbacks.py @@ -0,0 +1,56 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import copy + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.callbacks.every_n_draw_sample import EveryNDrawSample +from cosmos_policy._src.predict2.configs.common.defaults.callbacks import ( + BASIC_CALLBACKS, + SPEED_CALLBACKS, + WANDB_CALLBACK, +) + +_basic_callback = copy.deepcopy(BASIC_CALLBACKS) + +DEBUG_CALLBACKS = dict() +LONG_RUNNING_CALLBACKS = dict() + +VIZ_ONLINE_SAMPLING_CALLBACKS = dict( + every_n_sample_reg=L(EveryNDrawSample)( + every_n=5000, + save_s3=True, + ), + every_n_sample_ema=L(EveryNDrawSample)( + every_n=5000, + is_ema=True, + save_s3=True, + ), +) + + +def register_callbacks(): + cs = ConfigStore.instance() + cs.store(group="callbacks", package="trainer.callbacks", name="basic", node=_basic_callback) + cs.store(group="callbacks", package="trainer.callbacks", name="wandb", node=WANDB_CALLBACK) + cs.store(group="callbacks", package="trainer.callbacks", name="debug", node=DEBUG_CALLBACKS) + cs.store( + group="callbacks", package="trainer.callbacks", name="viz_online_sampling", node=VIZ_ONLINE_SAMPLING_CALLBACKS + ) + + cs.store(group="callbacks", package="trainer.callbacks", name="long", node=LONG_RUNNING_CALLBACKS) + cs.store(group="callbacks", package="trainer.callbacks", name="cluster_speed", node=SPEED_CALLBACKS) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/conditioner.py b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/conditioner.py new file mode 100644 index 0000000000000000000000000000000000000000..4f35d0553fe8b6340641bc2c788968d8eb4d38b1 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/conditioner.py @@ -0,0 +1,75 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.conditioner import ReMapkey, TextAttr, TextAttrEmptyStringDrop, VideoConditioner + +VideoConditionerFpsPaddingConfig: LazyDict = L(VideoConditioner)( + text=L(TextAttr)( + input_key=["t5_text_embeddings"], + dropout_rate=0.2, + ), + fps=L(ReMapkey)( + input_key="fps", + output_key="fps", + dropout_rate=0.0, + dtype=None, + ), + padding_mask=L(ReMapkey)( + input_key="padding_mask", + output_key="padding_mask", + dropout_rate=0.0, + dtype=None, + ), +) + + +VideoConditionerFpsPaddingEmptyStringDrppConfig: LazyDict = L(VideoConditioner)( + text=L(TextAttrEmptyStringDrop)( + input_key=["t5_text_embeddings"], + dropout_rate=0.2, + ), + fps=L(ReMapkey)( + input_key="fps", + output_key="fps", + dropout_rate=0.0, + dtype=None, + ), + padding_mask=L(ReMapkey)( + input_key="padding_mask", + output_key="padding_mask", + dropout_rate=0.0, + dtype=None, + ), +) + + +def register_conditioner(): + cs = ConfigStore.instance() + cs.store( + group="conditioner", + package="model.config.conditioner", + name="add_fps_padding_mask", + node=VideoConditionerFpsPaddingConfig, + ) + cs.store( + group="conditioner", + package="model.config.conditioner", + name="add_fps_padding_mask_empty_string_drop", + node=VideoConditionerFpsPaddingEmptyStringDrppConfig, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/model.py b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/model.py new file mode 100644 index 0000000000000000000000000000000000000000..5da91a596631887379150de35da10ecf064fe223 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/model.py @@ -0,0 +1,63 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.models.text2world_model import DiffusionModel as Text2WorldModel +from cosmos_policy._src.predict2.models.text2world_model import Text2WorldModelConfig +from cosmos_policy._src.predict2.models.text2world_wan2pt1_model import Text2WorldModelWan2pt1Config +from cosmos_policy._src.predict2.models.text2world_wan2pt1_model import WANDiffusionModel as Text2WorldWan2pt1Model + +DDP_CONFIG = dict( + trainer=dict( + distributed_parallelism="ddp", + ), + model=L(Text2WorldModel)( + config=Text2WorldModelConfig(), + _recursive_=False, + ), +) + +FSDP_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(Text2WorldModel)( + config=Text2WorldModelConfig( + fsdp_shard_size=8, + ), + _recursive_=False, + ), +) + +FSDP_WAN2PT1_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(Text2WorldWan2pt1Model)( + config=Text2WorldModelWan2pt1Config( + fsdp_shard_size=8, + ), + _recursive_=False, + ), +) + + +def register_model(): + cs = ConfigStore.instance() + cs.store(group="model", package="_global_", name="ddp", node=DDP_CONFIG) + cs.store(group="model", package="_global_", name="fsdp", node=FSDP_CONFIG) + cs.store(group="model", package="_global_", name="fsdp_wan2pt1", node=FSDP_WAN2PT1_CONFIG) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/net.py b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/net.py new file mode 100644 index 0000000000000000000000000000000000000000..f1783596f2f50be615ca36793cab23b67725dc44 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/text2world/defaults/net.py @@ -0,0 +1,125 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import copy + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.networks.minimal_v4_dit import MiniTrainDIT, SACConfig +from cosmos_policy._src.predict2.networks.wan2pt1 import WanModel + +COSMOS_V1_7B_NET_MININET: LazyDict = L(MiniTrainDIT)( + max_img_h=240, + max_img_w=240, + max_frames=128, + in_channels=16, + out_channels=16, + patch_spatial=2, + patch_temporal=1, + model_channels=4096, + num_blocks=28, + num_heads=32, + concat_padding_mask=True, + pos_emb_cls="rope3d", + pos_emb_learnable=True, + pos_emb_interpolation="crop", + use_adaln_lora=True, + adaln_lora_dim=256, + atten_backend="minimal_a2a", + extra_per_block_abs_pos_emb=True, + rope_h_extrapolation_ratio=1.0, + rope_w_extrapolation_ratio=1.0, + rope_t_extrapolation_ratio=2.0, + sac_config=SACConfig(), +) +COSMOS_V1_2B_NET_MININET = copy.deepcopy(COSMOS_V1_7B_NET_MININET) +COSMOS_V1_2B_NET_MININET.model_channels = 2048 +COSMOS_V1_2B_NET_MININET.num_heads = 16 +COSMOS_V1_2B_NET_MININET.num_blocks = 28 +COSMOS_V1_2B_NET_MININET.extra_per_block_abs_pos_emb = False +COSMOS_V1_2B_NET_MININET.rope_t_extrapolation_ratio = 1.0 + +COSMOS_V1_14B_NET_MININET = copy.deepcopy(COSMOS_V1_7B_NET_MININET) +COSMOS_V1_14B_NET_MININET.model_channels = 5120 +COSMOS_V1_14B_NET_MININET.num_heads = 40 +COSMOS_V1_14B_NET_MININET.num_blocks = 36 +COSMOS_V1_14B_NET_MININET.extra_per_block_abs_pos_emb = False +COSMOS_V1_14B_NET_MININET.rope_t_extrapolation_ratio = 1.0 + +COSMOS_V2_0P6B_NET = copy.deepcopy(COSMOS_V1_2B_NET_MININET) +COSMOS_V2_0P6B_NET.model_channels = 20 * 64 +COSMOS_V2_0P6B_NET.num_heads = 20 +COSMOS_V2_0P6B_NET.num_blocks = 20 + +COSMOS_V2_1B_NET = copy.deepcopy(COSMOS_V1_2B_NET_MININET) +COSMOS_V2_1B_NET.model_channels = 24 * 64 +COSMOS_V2_1B_NET.num_heads = 24 +COSMOS_V2_1B_NET.num_blocks = 24 + +mini_net = copy.deepcopy(COSMOS_V1_7B_NET_MININET) +mini_net.model_channels = 1024 +mini_net.num_heads = 8 +mini_net.num_blocks = 2 +mini_net.rope_t_extrapolation_ratio = 1.0 + +COSMOS_V2_2B_NET: LazyDict = L(MiniTrainDIT)( + max_img_h=240, + max_img_w=240, + max_frames=128, + in_channels=16, + out_channels=16, + patch_spatial=2, + patch_temporal=1, + model_channels=2048, + num_blocks=28, + num_heads=16, + concat_padding_mask=True, + pos_emb_cls="rope3d", + pos_emb_learnable=True, + pos_emb_interpolation="crop", + use_adaln_lora=True, + adaln_lora_dim=256, + extra_per_block_abs_pos_emb=True, +) + + +WAN2PT1_1PT3B: LazyDict = L(WanModel)( + dim=1536, + eps=1e-06, + ffn_dim=8960, + freq_dim=256, + in_dim=16, + model_type="t2v", + num_heads=12, + num_layers=30, + out_dim=16, + text_len=512, + cp_comm_type="p2p", + sac_config=L(SACConfig)(mode="block_wise"), +) + + +def register_net(): + cs = ConfigStore.instance() + cs.store(group="net", package="model.config.net", name="mini_net", node=mini_net) + cs.store(group="net", package="model.config.net", name="cosmos_v1_7B", node=COSMOS_V1_7B_NET_MININET) + cs.store(group="net", package="model.config.net", name="cosmos_v1_2B", node=COSMOS_V1_2B_NET_MININET) + cs.store(group="net", package="model.config.net", name="cosmos_v1_14B", node=COSMOS_V1_14B_NET_MININET) + cs.store(group="net", package="model.config.net", name="cosmos_v2_2B", node=COSMOS_V2_2B_NET) + cs.store(group="net", package="model.config.net", name="cosmos_v2_0p6B", node=COSMOS_V2_0P6B_NET) + cs.store(group="net", package="model.config.net", name="cosmos_v2_1B", node=COSMOS_V2_1B_NET) + cs.store(group="net", package="model.config.net", name="wan2pt1_1pt3B", node=WAN2PT1_1PT3B) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/config.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/config.py new file mode 100644 index 0000000000000000000000000000000000000000..66704f356a3ec8a68508b7e35a1fafbaa230363d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/config.py @@ -0,0 +1,107 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import importlib +from typing import Any, List + +import attrs + +from cosmos_policy._src.imaginaire import config +from cosmos_policy._src.imaginaire.trainer import ImaginaireTrainer as Trainer +from cosmos_policy._src.imaginaire.utils.config_helper import import_all_modules_from_package +from cosmos_policy._src.predict2.configs.common.defaults.checkpoint import register_checkpoint +from cosmos_policy._src.predict2.configs.common.defaults.ckpt_type import register_ckpt_type +from cosmos_policy._src.predict2.configs.common.defaults.dataloader import register_training_and_val_data +from cosmos_policy._src.predict2.configs.common.defaults.ema import register_ema +from cosmos_policy._src.predict2.configs.common.defaults.optimizer import register_optimizer +from cosmos_policy._src.predict2.configs.common.defaults.scheduler import register_scheduler +from cosmos_policy._src.predict2.configs.common.defaults.tokenizer import register_tokenizer +from cosmos_policy._src.predict2.configs.video2world.defaults.callbacks import register_callbacks +from cosmos_policy._src.predict2.configs.video2world.defaults.conditioner import register_conditioner +from cosmos_policy._src.predict2.configs.video2world.defaults.model import register_model +from cosmos_policy._src.predict2.configs.video2world.defaults.net import register_net + + +@attrs.define(slots=False) +class Config(config.Config): + # default config groups that will be used unless overwritten + # see config groups in registry.py + defaults: List[Any] = attrs.field( + factory=lambda: [ + "_self_", + {"data_train": "mock"}, + {"data_val": "mock"}, + {"optimizer": "fusedadamw"}, + {"scheduler": "lambdalinear"}, + {"model": "ddp"}, + {"callbacks": "basic"}, + {"net": None}, + {"conditioner": "video_prediction_conditioner"}, + {"ema": "power"}, + {"tokenizer": "cosmos_tokenizer_causal_cv8x8x8_c16_res720_t121_it121_v1_0"}, + {"checkpoint": "s3"}, + {"ckpt_type": "dummy"}, + # the list is with order, we need global experiment to be the last one + {"experiment": None}, + ] + ) + + +def make_config() -> Config: + c = Config( + model=None, + optimizer=None, + scheduler=None, + dataloader_train=None, + dataloader_val=None, + ) + + # Specifying values through instances of attrs + c.job.project = "cosmos_diffusion_v2" + c.job.group = "debug" + c.job.name = "delete_${now:%Y-%m-%d}_${now:%H-%M-%S}" + + c.trainer.type = Trainer + c.trainer.straggler_detection.enabled = False + c.trainer.max_iter = 400_000 + c.trainer.logging_iter = 10 + c.trainer.validation_iter = 100 + c.trainer.run_validation = False + c.trainer.callbacks = None + + # Call this function to register config groups for advanced overriding. the order follows the default config groups + register_training_and_val_data() + register_optimizer() + register_scheduler() + register_model() + register_callbacks() + register_net() + register_conditioner() + register_ema() + register_tokenizer() + register_checkpoint() + register_ckpt_type() + + # experiment config are defined in the experiment folder + # call import_all_modules_from_package to register them + import_all_modules_from_package("cosmos_policy._src.predict2.configs.video2world.experiment", reload=True) + try: + if importlib.util.find_spec("cosmos_predict2.experiments.internal") is not None: + import_all_modules_from_package("cosmos_predict2.experiments.internal", reload=True) + except ModuleNotFoundError: + pass # Module or parent package doesn't exist + import_all_modules_from_package("cosmos_predict2.experiments", reload=True) + + return c diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/callbacks.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/callbacks.py new file mode 100644 index 0000000000000000000000000000000000000000..9c4f200a6c17363f35834614b1f5f5503cfd61e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/callbacks.py @@ -0,0 +1,57 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import copy + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.callbacks.every_n_draw_sample import EveryNDrawSample +from cosmos_policy._src.predict2.configs.common.defaults.callbacks import ( + BASIC_CALLBACKS, + SPEED_CALLBACKS, + WANDB_CALLBACK, +) + +_basic_callback = copy.deepcopy(BASIC_CALLBACKS) + +DEBUG_CALLBACKS = dict() +LONG_RUNNING_CALLBACKS = dict() + +VIZ_ONLINE_SAMPLING_CALLBACKS = dict( + every_n_sample_reg=L(EveryNDrawSample)( + every_n=5000, + save_s3=True, + do_x0_prediction=False, + ), + every_n_sample_ema=L(EveryNDrawSample)( + every_n=5000, + is_ema=True, + save_s3=True, + do_x0_prediction=False, + ), +) + + +def register_callbacks(): + cs = ConfigStore.instance() + cs.store(group="callbacks", package="trainer.callbacks", name="basic", node=_basic_callback) + cs.store(group="callbacks", package="trainer.callbacks", name="wandb", node=WANDB_CALLBACK) + cs.store(group="callbacks", package="trainer.callbacks", name="debug", node=DEBUG_CALLBACKS) + cs.store( + group="callbacks", package="trainer.callbacks", name="viz_online_sampling", node=VIZ_ONLINE_SAMPLING_CALLBACKS + ) + cs.store(group="callbacks", package="trainer.callbacks", name="long", node=LONG_RUNNING_CALLBACKS) + cs.store(group="callbacks", package="trainer.callbacks", name="cluster_speed", node=SPEED_CALLBACKS) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/conditioner.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/conditioner.py new file mode 100644 index 0000000000000000000000000000000000000000..83283fdc3bd07bd1228039460472accabfed0c85 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/conditioner.py @@ -0,0 +1,378 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import random +from dataclasses import dataclass +from typing import Dict, Optional + +import torch +from einops import rearrange +from hydra.core.config_store import ConfigStore +from torch.distributed import get_process_group_ranks + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.utils.context_parallel import broadcast_split_tensor, find_split +from cosmos_policy._src.predict2.conditioner import ( + BooleanFlag, + GeneralConditioner, + ReMapkey, + Text2WorldCondition, + TextAttr, + TextAttrEmptyStringDrop, +) +from cosmos_policy._src.predict2.models.video2world_wan2pt1_model import WAN2PT1_I2V_COND_LATENT_KEY +from cosmos_policy._src.predict2.networks.clip import Wan2pt1CLIPEmb + + +@dataclass(frozen=True) +class Video2WorldCondition(Text2WorldCondition): + use_video_condition: bool = False + # the following two attributes are used to set the video condition; during training, inference + gt_frames: Optional[torch.Tensor] = None + condition_video_input_mask_B_C_T_H_W: Optional[torch.Tensor] = None + + def set_video_condition( + self, + gt_frames: torch.Tensor, + random_min_num_conditional_frames: int, + random_max_num_conditional_frames: int, + num_conditional_frames: Optional[int] = None, + conditional_frames_probs: Optional[Dict[int, float]] = None, + ) -> "Video2WorldCondition": + """ + Sets the video conditioning frames for video-to-video generation. + + This method creates a conditioning mask for the input video frames that determines + which frames will be used as context frames for generating new frames. The method + handles both image batches (T=1) and video batches (T>1) differently. + + Args: + gt_frames: A tensor of ground truth frames with shape [B, C, T, H, W], where: + B = batch size + C = number of channels + T = number of frames + H = height + W = width + + random_min_num_conditional_frames: Minimum number of frames to use for conditioning + when randomly selecting a number of conditioning frames. + + random_max_num_conditional_frames: Maximum number of frames to use for conditioning + when randomly selecting a number of conditioning frames. + + num_conditional_frames: Optional; If provided, all examples in the batch will use + exactly this many frames for conditioning. If None, a random number of frames + between random_min_num_conditional_frames and random_max_num_conditional_frames + will be selected for each example in the batch. + + conditional_frames_probs: Optional; Dictionary mapping number of frames to probabilities. + If provided, overrides the random_min/max_num_conditional_frames with weighted sampling. + Example: {0: 0.5, 1: 0.25, 2: 0.25} for 50% chance of 0 frames, 25% for 1, 25% for 2. + + Returns: + A new Video2WorldCondition object with the gt_frames and conditioning mask set. + The conditioning mask (condition_video_input_mask_B_C_T_H_W) is a binary tensor + of shape [B, 1, T, H, W] where 1 indicates frames used for conditioning and 0 + indicates frames to be generated. + + Notes: + - For image batches (T=1), no conditioning frames are used (num_conditional_frames_B = 0). + - For video batches: + - If num_conditional_frames is provided, all examples use that fixed number of frames. + - Otherwise, each example randomly uses between random_min_num_conditional_frames and + random_max_num_conditional_frames frames. + - The mask marks the first N frames as conditioning frames (set to 1) for each example. + """ + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = gt_frames + + # condition_video_input_mask_B_C_T_H_W + B, _, T, H, W = gt_frames.shape + condition_video_input_mask_B_C_T_H_W = torch.zeros( + B, 1, T, H, W, dtype=gt_frames.dtype, device=gt_frames.device + ) + if T == 1: # handle image batch + num_conditional_frames_B = torch.zeros(B, dtype=torch.int32) + else: # handle video batch + if num_conditional_frames is not None: + if isinstance(num_conditional_frames, torch.Tensor): + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames.cpu() + else: + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames + elif conditional_frames_probs is not None: + # Use weighted sampling based on provided probabilities + frames_options = list(conditional_frames_probs.keys()) + weights = list(conditional_frames_probs.values()) + num_conditional_frames_B = torch.tensor( + random.choices(frames_options, weights=weights, k=B), dtype=torch.int32 + ) + else: + num_conditional_frames_B = torch.randint( + random_min_num_conditional_frames, random_max_num_conditional_frames + 1, size=(B,) + ) + for idx in range(B): + condition_video_input_mask_B_C_T_H_W[idx, :, : num_conditional_frames_B[idx], :, :] += 1 + + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + return type(self)(**kwargs) + + def edit_for_inference( + self, is_cfg_conditional: bool = True, num_conditional_frames: int = 1 + ) -> "Video2WorldCondition": + _condition = self.set_video_condition( + gt_frames=self.gt_frames, + random_min_num_conditional_frames=0, + random_max_num_conditional_frames=0, + num_conditional_frames=num_conditional_frames, + ) + if not is_cfg_conditional: + # Do not use classifier free guidance on conditional frames. + # YB found that it leads to worse results. + _condition.use_video_condition.fill_(True) + return _condition + + def broadcast(self, process_group: torch.distributed.ProcessGroup) -> "Video2WorldCondition": + if self.is_broadcasted: + return self + # extra efforts + gt_frames = self.gt_frames + condition_video_input_mask_B_C_T_H_W = self.condition_video_input_mask_B_C_T_H_W + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = None + kwargs["condition_video_input_mask_B_C_T_H_W"] = None + new_condition = Text2WorldCondition.broadcast( + type(self)(**kwargs), + process_group, + ) + + kwargs = new_condition.to_dict(skip_underscore=False) + _, _, T, _, _ = gt_frames.shape + if process_group is not None: + cp_ranks = get_process_group_ranks(process_group) + cp_size = len(cp_ranks) + use_spatial_split = ( + cp_size > condition_video_input_mask_B_C_T_H_W.shape[2] + or condition_video_input_mask_B_C_T_H_W.shape[2] % cp_size != 0 + ) + after_split_shape = ( + find_split(condition_video_input_mask_B_C_T_H_W.shape, cp_size) if use_spatial_split else None + ) + + if T > 1 and process_group.size() > 1: + if use_spatial_split: + condition_video_input_mask_B_C_T_H_W = rearrange( + condition_video_input_mask_B_C_T_H_W, "b c t h w -> b c (t h w)" + ) + gt_frames = rearrange(gt_frames, "b c t h w -> b c (t h w)") + gt_frames = broadcast_split_tensor(gt_frames, seq_dim=2, process_group=process_group) + condition_video_input_mask_B_C_T_H_W = broadcast_split_tensor( + condition_video_input_mask_B_C_T_H_W, seq_dim=2, process_group=process_group + ) + if use_spatial_split: + condition_video_input_mask_B_C_T_H_W = rearrange( + condition_video_input_mask_B_C_T_H_W, + "b c (t h w) -> b c t h w", + t=after_split_shape[0], + h=after_split_shape[1], + ) + gt_frames = rearrange( + gt_frames, "b c (t h w) -> b c t h w", t=after_split_shape[0], h=after_split_shape[1] + ) + kwargs["gt_frames"] = gt_frames + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + return type(self)(**kwargs) + + +class Video2WorldConditionV2(Video2WorldCondition): + """ + compared to Video2WorldCondition, this class apply zero frames when use_video_condition is False~(unconditional generation in cfg) + in the case, we do zero-out conditional frames in the video condition + """ + + def set_video_condition( + self, + gt_frames: torch.Tensor, + random_min_num_conditional_frames: int, + random_max_num_conditional_frames: int, + num_conditional_frames: Optional[int] = None, + conditional_frames_probs: Optional[Dict[int, float]] = None, + ) -> "Video2WorldConditionV2": + num_conditional_frames = 0 if not self.use_video_condition else num_conditional_frames + return super().set_video_condition( + gt_frames=gt_frames, + random_min_num_conditional_frames=random_min_num_conditional_frames, + random_max_num_conditional_frames=random_max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + conditional_frames_probs=conditional_frames_probs, + ) + + def edit_for_inference( + self, is_cfg_conditional: bool = True, num_conditional_frames: int = 1 + ) -> "Video2WorldConditionV2": + del is_cfg_conditional + _condition = super().set_video_condition( + gt_frames=self.gt_frames, + random_min_num_conditional_frames=0, + random_max_num_conditional_frames=0, + num_conditional_frames=num_conditional_frames, + ) + return _condition + + +class Video2WorldConditioner(GeneralConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> Video2WorldCondition: + output = super()._forward(batch, override_dropout_rate) + return Video2WorldCondition(**output) + + +class Video2WorldConditionerV2(GeneralConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> Video2WorldConditionV2: + output = super()._forward(batch, override_dropout_rate) + return Video2WorldConditionV2(**output) + + +_SHARED_CONFIG = dict( + fps=L(ReMapkey)( + input_key="fps", + output_key="fps", + dropout_rate=0.0, + dtype=None, + ), + padding_mask=L(ReMapkey)( + input_key="padding_mask", + output_key="padding_mask", + dropout_rate=0.0, + dtype=None, + ), + text=L(TextAttr)( + input_key=["t5_text_embeddings"], + dropout_rate=0.2, + use_empty_string=False, + ), + use_video_condition=L(BooleanFlag)( + input_key="fps", + output_key="use_video_condition", + dropout_rate=0.2, + ), +) + +VideoPredictionConditioner: LazyDict = L(Video2WorldConditioner)( + **_SHARED_CONFIG, +) + +VideoPredictionConditionerV2: LazyDict = L(Video2WorldConditionerV2)( + **_SHARED_CONFIG, +) + + +@dataclass(frozen=True) +class VideoPredictionWan2pt1Condition(Text2WorldCondition): + frame_cond_crossattn_emb_B_L_D: Optional[torch.Tensor] = None + y_B_C_T_H_W: Optional[torch.Tensor] = None # image condition + # latent_condition: Optional[torch.Tensor] = None # latent condition + + def broadcast(self, process_group: torch.distributed.ProcessGroup) -> "Video2WorldCondition": + """Broadcasts and splits the condition across the checkpoint parallelism group. + For most condition, such asT2VCondition, we do not need split. + + Args: + process_group: The process group for broadcast and split + + Returns: + A new BaseCondition instance with the broadcasted and split condition. + """ + if self.is_broadcasted: + return self + + y_B_C_T_H_W = self.y_B_C_T_H_W + kwargs = self.to_dict(skip_underscore=False) + kwargs["y_B_C_T_H_W"] = None + new_condition = Text2WorldCondition.broadcast( + type(self)(**kwargs), + process_group, + ) + kwargs = new_condition.to_dict(skip_underscore=False) + if process_group is not None: + y_B_C_T_H_W = broadcast_split_tensor(y_B_C_T_H_W, seq_dim=2, process_group=process_group) + kwargs["y_B_C_T_H_W"] = y_B_C_T_H_W + return type(self)(**kwargs) + + +class VideoPredictionWan2pt1Conditioner(GeneralConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> VideoPredictionWan2pt1Condition: + output = super()._forward(batch, override_dropout_rate) + return VideoPredictionWan2pt1Condition(**output) + + +VideoConditionerFpsPaddingEmptyStringDrppConfig: LazyDict = L(VideoPredictionWan2pt1Conditioner)( + text=L(TextAttrEmptyStringDrop)( + input_key=["t5_text_embeddings"], + dropout_rate=0.2, + ), + fps=L(ReMapkey)( + input_key="fps", + output_key="fps", + dropout_rate=0.0, + dtype=None, + ), + padding_mask=L(ReMapkey)( + input_key="padding_mask", + output_key="padding_mask", + dropout_rate=0.0, + dtype=None, + ), + wanclip=L(Wan2pt1CLIPEmb)( + input_key=["images", "video", WAN2PT1_I2V_COND_LATENT_KEY], + dropout_rate=0.0, + dtype="bfloat16", + ), +) + + +def register_conditioner(): + cs = ConfigStore.instance() + cs.store( + group="conditioner", + package="model.config.conditioner", + name="video_prediction_conditioner", + node=VideoPredictionConditioner, + ) + + cs.store( + group="conditioner", + package="model.config.conditioner", + name="video_prediction_conditioner_v2", + node=VideoPredictionConditionerV2, + ) + + cs.store( + group="conditioner", + package="model.config.conditioner", + name="wan2pt1_video_prediction_conditioner_empty_string_drop", + node=VideoConditionerFpsPaddingEmptyStringDrppConfig, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/model.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/model.py new file mode 100644 index 0000000000000000000000000000000000000000..63c4851465ca9ce40b36227f77d354a28340e216 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/model.py @@ -0,0 +1,82 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.predict2.models.text2world_wan2pt1_model import Text2WorldModelWan2pt1Config +from cosmos_policy._src.predict2.models.video2world_model import Video2WorldConfig, Video2WorldModel +from cosmos_policy._src.predict2.models.video2world_model_rectified_flow import ( + Video2WorldModelRectifiedFlow, + Video2WorldModelRectifiedFlowConfig, +) +from cosmos_policy._src.predict2.models.video2world_wan2pt1_model import I2VWan2pt1Model + +DDP_CONFIG = dict( + trainer=dict( + distributed_parallelism="ddp", + ), + model=L(Video2WorldModel)( + config=Video2WorldConfig(), + _recursive_=False, + ), +) + +FSDP_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(Video2WorldModel)( + config=Video2WorldConfig( + fsdp_shard_size=8, + ), + _recursive_=False, + ), +) + + +FSDP_WAN2PT1_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(I2VWan2pt1Model)( + config=Text2WorldModelWan2pt1Config( + fsdp_shard_size=8, + state_t=24, + ), + _recursive_=False, + ), +) + +FSDP_RECTIFIED_FLOW_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(Video2WorldModelRectifiedFlow)( + config=Video2WorldModelRectifiedFlowConfig( + fsdp_shard_size=8, + state_t=24, + ), + _recursive_=False, + ), +) + + +def register_model(): + cs = ConfigStore.instance() + cs.store(group="model", package="_global_", name="ddp", node=DDP_CONFIG) + cs.store(group="model", package="_global_", name="fsdp", node=FSDP_CONFIG) + cs.store(group="model", package="_global_", name="fsdp_wan2pt1", node=FSDP_WAN2PT1_CONFIG) + cs.store(group="model", package="_global_", name="fsdp_rectified_flow", node=FSDP_RECTIFIED_FLOW_CONFIG) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/net.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/net.py new file mode 100644 index 0000000000000000000000000000000000000000..54e7d6443eda7a79f7a0ef97f1ee083175a9b919 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/defaults/net.py @@ -0,0 +1,110 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import copy + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.networks.minimal_v1_lvg_dit import MinimalV1LVGDiT +from cosmos_policy._src.predict2.networks.minimal_v4_dit import SACConfig +from cosmos_policy._src.predict2.networks.wan2pt1 import WanModel + +WAN2PT1_1PT3B: LazyDict = L(WanModel)( + dim=1536, + eps=1e-06, + ffn_dim=8960, + freq_dim=256, + in_dim=36, # 16 (VAE channels) + 20 (image conditioning) + model_type="i2v", + num_heads=12, + num_layers=30, + out_dim=16, + text_len=512, + cp_comm_type="p2p", + sac_config=L(SACConfig)(mode="block_wise"), + attention_backend="minimal_a2a", +) + +WAN2PT1_14B: LazyDict = L(WanModel)( + dim=5120, + eps=1e-06, + ffn_dim=13824, + freq_dim=256, + in_dim=36, + model_type="i2v", + num_heads=40, + num_layers=40, + out_dim=16, + text_len=512, + cp_comm_type="p2p", + sac_config=L(SACConfig)(mode="block_wise"), + attention_backend="minimal_a2a", +) + +COSMOS_V1_7B_NET_MININET: LazyDict = L(MinimalV1LVGDiT)( + max_img_h=240, + max_img_w=240, + max_frames=128, + in_channels=16, + out_channels=16, + patch_spatial=2, + patch_temporal=1, + model_channels=4096, + num_blocks=28, + num_heads=32, + concat_padding_mask=True, + pos_emb_cls="rope3d", + pos_emb_learnable=True, + pos_emb_interpolation="crop", + use_adaln_lora=True, + adaln_lora_dim=256, + atten_backend="minimal_a2a", + extra_per_block_abs_pos_emb=True, + rope_h_extrapolation_ratio=1.0, + rope_w_extrapolation_ratio=1.0, + rope_t_extrapolation_ratio=2.0, + sac_config=SACConfig(), +) +COSMOS_V1_2B_NET_MININET = copy.deepcopy(COSMOS_V1_7B_NET_MININET) +COSMOS_V1_2B_NET_MININET.model_channels = 2048 +COSMOS_V1_2B_NET_MININET.num_heads = 16 +COSMOS_V1_2B_NET_MININET.num_blocks = 28 +COSMOS_V1_2B_NET_MININET.extra_per_block_abs_pos_emb = False +COSMOS_V1_2B_NET_MININET.rope_t_extrapolation_ratio = 1.0 + +COSMOS_V1_14B_NET_MININET = copy.deepcopy(COSMOS_V1_7B_NET_MININET) +COSMOS_V1_14B_NET_MININET.model_channels = 5120 +COSMOS_V1_14B_NET_MININET.num_heads = 40 +COSMOS_V1_14B_NET_MININET.num_blocks = 36 +COSMOS_V1_14B_NET_MININET.extra_per_block_abs_pos_emb = False +COSMOS_V1_14B_NET_MININET.rope_t_extrapolation_ratio = 1.0 + +mini_net = copy.deepcopy(COSMOS_V1_7B_NET_MININET) +mini_net.model_channels = 1024 +mini_net.num_heads = 8 +mini_net.num_blocks = 2 +mini_net.rope_t_extrapolation_ratio = 1.0 + + +def register_net(): + cs = ConfigStore.instance() + cs.store(group="net", package="model.config.net", name="wan2pt1_1pt3B", node=WAN2PT1_1PT3B) + cs.store(group="net", package="model.config.net", name="wan2pt1_14B", node=WAN2PT1_14B) + cs.store(group="net", package="model.config.net", name="mini_net", node=mini_net) + cs.store(group="net", package="model.config.net", name="cosmos_v1_2B", node=COSMOS_V1_2B_NET_MININET) + cs.store(group="net", package="model.config.net", name="cosmos_v1_7B", node=COSMOS_V1_7B_NET_MININET) + cs.store(group="net", package="model.config.net", name="cosmos_v1_14B", node=COSMOS_V1_14B_NET_MININET) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_14b_reason_1p1.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_14b_reason_1p1.py new file mode 100644 index 0000000000000000000000000000000000000000..740411469af33f878f1304d6abf11aa0ae0730b6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_14b_reason_1p1.py @@ -0,0 +1,363 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +import math + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.configs.video2world.experiment.reason_embeddings.stage3_14B_index_3 import ( + I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR, +) +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import get_cached_replay_dataloader +from cosmos_policy._src.predict2.datasets.dataset_provider import get_image_dataset, get_video_dataset +from cosmos_policy._src.predict2.datasets.joint_dataloader import IterativeJointDataLoader +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy +from cosmos_policy._src.predict2.text_encoders.text_encoder import EmbeddingConcatStrategy + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=25, + logging_iter=2, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1, + ), + every_n_sample_ema=dict( + every_n=1, + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/vid2vid/config.py -- experiment=Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor + +# diff with resumed config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/vid2vid/config.py -- experiment=Stage-a_pt_3-Vid2Vid-Index-5-Size-14B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond +""" +I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_gcp", + }, + {"override /checkpoint": "gcp"}, + {"override /tokenizer": "wan2pt1_tokenizer_gcp"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_gcp", + ), + checkpoint=dict( + save_iter=500, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_qwen_concat_v2/checkpoints/iter_000038000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + min_num_conditional_frames=0, # choose either 1 (img2vid) or 2 (vid2vid) latent frames + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + scaling="rectified_flow", + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720_aggressive", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + s3_credential_path="credentials/gcp_checkpoint.secret", + ), + ), + ), + scheduler=dict( + f_max=[0.3], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + trainer=dict( + max_iter=200_000, + logging_iter=100, + straggler_detection=dict( + enabled=False, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME2 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_gcp", + }, + "_self_", + ], + job=dict( + group=I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_gcp_resume2", + ), + checkpoint=dict( + save_iter=500, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_gcp/checkpoints/iter_000011500", + load_training_state=True, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_IMPROVED_PRETRAINING_DATA = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16['job']['name']}", + { + "override /data_train": None, + }, + "_self_", + ], + job=dict( + group=I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_improved_pretraining_data_gcp", + ), + dataloader_train=L(IterativeJointDataLoader)( + dataloaders={ + "image_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_pretrain_and_synthetic_photoreal_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="qwen2p5_7b_v4", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=3, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + ), + ratio=1, + ), + "image_data_prompt_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_synthetic_filtered_combined_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="prompts", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=3, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + ), + ratio=1, + ), + "video_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name="cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_video_whole", + object_store="gcp", + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + max_fps_thres=38, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + use_native_fps=True, + embedding_type=None, + is_train=True, + chunk_size=256, + ), + batch_size=1, + num_workers=8, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + use_cache=False, + ), + ratio=2, + ), + } + ), + checkpoint=dict( + save_iter=250, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_gcp_resume2/checkpoints/iter_000043750", + load_training_state=True, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +cs = ConfigStore.instance() +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16, + *build_debug_runs(I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16), + ], + [ + I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME2, + *build_debug_runs(I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME2), + ], + [ + I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_IMPROVED_PRETRAINING_DATA, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_IMPROVED_PRETRAINING_DATA + ), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_14b_reason_1p1_rectified_flow.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_14b_reason_1p1_rectified_flow.py new file mode 100644 index 0000000000000000000000000000000000000000..ee297309ee80c41da0fe294e9e4848a7de7c34cb --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_14b_reason_1p1_rectified_flow.py @@ -0,0 +1,704 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import get_cached_replay_dataloader +from cosmos_policy._src.predict2.datasets.dataset_provider import get_image_dataset, get_video_dataset +from cosmos_policy._src.predict2.datasets.joint_dataloader import IterativeJointDataLoader +from cosmos_policy._src.predict2.text_encoders.text_encoder import EmbeddingConcatStrategy + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=1000, + logging_iter=50, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000000000000, + ), + every_n_sample_ema=dict( + every_n=1000000000000, + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_IMPROVED_PRETRAINING_DATA_RECTIFIED_FLOW_GCP = LazyDict( + dict( + defaults=[ + {"override /data_train": None}, + {"override /model": "fsdp_rectified_flow"}, + {"override /net": "cosmos_v1_14B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "adamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "gcp"}, + {"override /tokenizer": "wan2pt1_tokenizer_gcp"}, + "_self_", + ], + job=dict( + group="official_runs_text2world", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_improved_pretraining_data_rectified_flow_gcp", + ), + optimizer=dict( + lr=2 ** (-14.5), + weight_decay=0.2, + betas=[0.9, 0.999], + ), + scheduler=dict( + f_max=[0.3], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + model=dict( + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + fsdp_shard_size=32, + resolution="720", + state_t=24, + shift=5, + use_dynamic_shift=False, + train_time_weight="reweighting", + train_time_distribution="logitnormal", + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + timestep_scale=0.001, + sac_config=dict( + mode="predict2_14b_720_aggressive", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + use_wan_fp32_strategy=True, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.1, + use_empty_string=False, + ), + ), + tokenizer=dict( + temporal_window=16, + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + s3_credential_path="credentials/gcp_checkpoint.secret", + ), + ) + ), + checkpoint=dict( + save_iter=250, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_improved_pretraining_data_gcp/checkpoints/iter_000055750", + load_training_state=False, + strict_resume=True, + ), + model_parallel=dict( + context_parallel_size=2, + ), + trainer=dict( + max_iter=200_000, + logging_iter=100, + straggler_detection=dict( + enabled=False, + max_diff=1.5, + ), + callbacks=dict( + grad_clip=dict( + clip_norm=0.1, + ), + manual_gc=dict( + every_n=200, + ), + every_n_sample_reg=dict( + every_n=1000000000000, + ), + every_n_sample_ema=dict( + every_n=1000000000000, + ), + ), + ), + dataloader_train=L(IterativeJointDataLoader)( + dataloaders={ + "image_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_pretrain_and_synthetic_photoreal_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="qwen2p5_7b_v4", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=3, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + ), + ratio=1, + ), + "image_data_prompt_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_synthetic_filtered_combined_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="prompts", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=3, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + ), + ratio=1, + ), + "video_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name="cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_video_whole", + object_store="gcp", + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + use_native_fps=True, + embedding_type=None, + is_train=True, + chunk_size=256, + ), + batch_size=1, + num_workers=4, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + use_cache=False, + ), + ratio=2, + ), + } + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + + +# Rectified flow run with original training data with >= 720p resolution +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5 = LazyDict( + dict( + defaults=[ + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + {"override /model": "fsdp_rectified_flow"}, + {"override /net": "cosmos_v1_14B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "adamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_text2world", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5", + ), + optimizer=dict( + lr=2 ** (-14.5), + weight_decay=0.001, + betas=[0.9, 0.999], + ), + scheduler=dict( + f_max=[0.3], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + model=dict( + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + fsdp_shard_size=32, + resolution="720", + state_t=24, + shift=5, + use_dynamic_shift=False, + train_time_weight="reweighting", + train_time_distribution="logitnormal", + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + timestep_scale=0.001, + sac_config=dict( + mode="predict2_14b_720_aggressive", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + use_wan_fp32_strategy=True, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + use_empty_string=False, + ), + ), + tokenizer=dict( + temporal_window=16, + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + s3_credential_path="credentials/s3_checkpoint.secret", + ), + ) + ), + checkpoint=dict( + save_iter=250, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_qwen_concat_v2/checkpoints/iter_000031000", + load_training_state=False, + strict_resume=True, + ), + model_parallel=dict( + context_parallel_size=8, + ), + trainer=dict( + max_iter=200_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + callbacks=dict( + grad_clip=dict( + clip_norm=0.1, + ), + manual_gc=dict( + every_n=200, + ), + every_n_sample_reg=dict( + every_n=5000, + ), + every_n_sample_ema=dict( + every_n=5000, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + use_native_fps=True, + ), + ), + ratio=3, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + +# Config for shift 2 +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5['job']['name']}", + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5[ + "job" + ]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift2", + ), + model=dict( + config=dict( + shift=2, + ), + ), + ), + flags={"allow_objects": True}, +) + +# Config for shift 7 +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT7 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5['job']['name']}", + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5[ + "job" + ]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift7", + ), + model=dict( + config=dict( + shift=7, + ), + ), + ), + flags={"allow_objects": True}, +) + +# Config for shift 5 with high sigma strategy +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5['job']['name']}", + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5[ + "job" + ]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma", + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + ), + flags={"allow_objects": True}, +) + +# 4K cooldown config - S3 +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_4K_DATA_S3 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_4K_20250812_s3" + }, + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5[ + "job" + ]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_4K_data_s3", + ), + checkpoint=dict( + save_iter=500, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_qwen_concat_v2/checkpoints/iter_000031000", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + + +# 4K cooldown config - GCP +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_4K_DATA_GCP = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_4K_20250812_gcp" + }, + {"override /tokenizer": "wan2pt1_tokenizer_gcp"}, + {"override /checkpoint": "s3"}, + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5[ + "job" + ]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_4K_data_gcp", + ), + model=dict( + config=dict( + net=dict( + sac_config=dict( + mode="predict2_14b_720_aggressive", + ), + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + s3_credential_path="credentials/gcp_checkpoint.secret", + ), + ), + ), + trainer=dict( + straggler_detection=dict( + enabled=False, + ), + ), + model_parallel=dict( + context_parallel_size=4, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + object_store="gcp", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + object_store="gcp", + ), + ), + ratio=3, + ), + ), + ), + checkpoint=dict( + save_iter=250, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_qwen_concat_v2/checkpoints/iter_000031000", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +# Config for shift 7 - GCP +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT7_GCP = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_4K_DATA_GCP['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_gcp" + }, + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_4K_DATA_GCP[ + "job" + ]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift7_gcp", + ), + model=dict( + config=dict( + shift=7, + ), + ), + ), + flags={"allow_objects": True}, +) + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_IMPROVED_PRETRAINING_DATA_RECTIFIED_FLOW_GCP, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_IMPROVED_PRETRAINING_DATA_RECTIFIED_FLOW_GCP + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5 + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT2, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT2 + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT7, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT7 + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_4K_DATA_GCP, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_4K_DATA_GCP + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_4K_DATA_S3, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_4K_DATA_S3 + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT7_GCP, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT7_GCP + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA + ), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_2B_reason_1p1.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_2B_reason_1p1.py new file mode 100644 index 0000000000000000000000000000000000000000..1ccb4a0809d971a901b1e6ee4e1b8e9b29b8a2b9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_2B_reason_1p1.py @@ -0,0 +1,557 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +import math + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.configs.video2world.experiment.reason_embeddings.stage3_2B import ( + I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY, + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2, +) + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=1000, + logging_iter=50, + callbacks=dict( + every_n_sample_reg=dict( + every_n=10, + ), + every_n_sample_ema=dict( + every_n=10, + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_s3", + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted", + ), + model=dict( + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + scaling="rectified_flow", # correct loss weight for rectified flow + conditioner=dict( + text=dict( + use_empty_string=False, + ), + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000065000/", + load_training_state=False, + strict_resume=True, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + use_cache=False, + ), + ratio=3, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +# Finetuning of Predict2-2B-CR1.1 on weekly data release on video only dataset +# This is achieved by setting the ratio of video_data to 1 and image_data to 0 +WEEKYLY_FINETUNING = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250602_dedup_20250825_s3", + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + # name="weekly_data_release_finetuning_64nodes", + # name="weekly_data_release_finetuning_data_v20250805", + name="weekly_finetuning_pretrainvideo_20250602_dedup_20250825", + ), + trainer=dict( + max_iter=50_000, # Shorter training for fine-tuning + logging_iter=20, + callbacks=dict( + iter_speed=dict( + hit_thres=200, # monitor the speed of the first 200 iterations + ), + every_n_sample_reg=dict( + every_n=500, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, # Standard output fps for video generation + ), + every_n_sample_ema=dict( + every_n=500, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, # Standard output fps for video generation + ), + ), + # Disable validation completely + run_validation=False, + validation_iter=999999999, + ), + checkpoint=dict( + save_iter=5_000, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + ratio=0, + ), + video_data=dict( + ratio=1, # video only + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_COOLDOWN_FROM_10K = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v6_20250607_s3" + }, + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_cooldown_from_10K", + ), + trainer=dict( + max_iter=30_000, + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.0], + warm_up_steps=[0], + cycle_lengths=[30_000], + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + use_native_fps=True, + min_fps_thres=14, + max_fps_thres=30, + ), + ), + ), + ), + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +######################################################## +# Cooldown configs + +# distribution matching + loss weight, on general pretraining dataset +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_COOLDOWN_FROM_10K = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v6_20250607_s3" + }, + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_HQ_cooldown_from_10K", + ), + trainer=dict( + max_iter=50_000, + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.0], + warm_up_steps=[0], + cycle_lengths=[50_000], + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16_hq_cooldown_from_57K/checkpoints/iter_000035000/", + load_training_state=True, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_4K_COOLDOWN_FROM_10K = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_4K_20250812_s3" + }, + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_4K_cooldown_from_10K", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16_hq_cooldown_from_57K_4K_videos/checkpoints/iter_000035000/", + load_training_state=True, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V8_COOLDOWN_FROM_10K = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v8_20250822_s3" + }, + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_HQ_V8_cooldown_from_10K", + ), + trainer=dict( + max_iter=50_000, + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.0], + warm_up_steps=[0], + cycle_lengths=[50_000], + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +######################################################## +# Noise analysis configs + + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_NOISE_ANALYSIS_SHIFT2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_noise_analysis_shift2", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=True, + strict_resume=True, + ), + model=dict( + config=dict( + sde=dict( + p_mean=math.log(2.0), + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_NOISE_ANALYSIS_SHIFT5_CONTCHUNK = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_noise_analysis_shift5_contchunk", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=True, + strict_resume=True, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + use_native_fps=True, + min_fps_thres=14, + max_fps_thres=30, + ), + ), + ), + ), + ), + model=dict( + config=dict( + sde=dict( + p_mean=math.log(5.0), + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_NOISE_ANALYSIS_SHIFT7 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_noise_analysis_shift7", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=True, + strict_resume=True, + ), + model=dict( + config=dict( + sde=dict( + p_mean=math.log(7.0), + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +######################################################## +# qwen0.5B configs +""" +torchrun --nproc_per_node=8 --master_port=12341 scripts/train.py --config=cosmos_policy/_src/predict2/configs/video2world/config.py -- experiment=Stage-c_pt_4-qwen05b-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted +torchrun --nproc_per_node=8 --master_port=12341 scripts/train.py --config=cosmos_policy/_src/predict2/configs/video2world/config.py -- experiment=Stage-c_pt_4-qwen05b-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_mock_wo_resume +""" +T2V_QWEN05B_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-qwen05b-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted", + ), + model=dict( + config=dict( + net=dict( + crossattn_proj_in_channels=21504, + ), + text_encoder_class="qwen0.5B", + text_encoder_config=dict( + ckpt_path="s3://bucket/cosmos_reasoning1/pretrained/qwen2p5_0p5b/checkpoints/iter_000000001/model/", + model_config=dict( + model_config=dict( + name_or_path="Qwen/Qwen2.5-0.5B", + model_type="qwen2_5", + tokenizer_type="Qwen/Qwen2.5-0.5B", + ), + tokenizer=dict( + tokenizer_type="Qwen/Qwen2.5-0.5B", + ), + ), + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000065000/", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_COOLDOWN_FROM_10K, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_COOLDOWN_FROM_10K), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_COOLDOWN_FROM_10K, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_COOLDOWN_FROM_10K), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_4K_COOLDOWN_FROM_10K, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_4K_COOLDOWN_FROM_10K), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_NOISE_ANALYSIS_SHIFT2, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_NOISE_ANALYSIS_SHIFT2), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_NOISE_ANALYSIS_SHIFT7, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_NOISE_ANALYSIS_SHIFT7), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_NOISE_ANALYSIS_SHIFT5_CONTCHUNK, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_NOISE_ANALYSIS_SHIFT5_CONTCHUNK + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V8_COOLDOWN_FROM_10K, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V8_COOLDOWN_FROM_10K + ), + ], + [ + T2V_QWEN05B_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16, + *build_debug_runs(T2V_QWEN05B_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16), + ], + [ + WEEKYLY_FINETUNING, + *build_debug_runs(WEEKYLY_FINETUNING), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_2B_reason_1p1_rectified_flow.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_2B_reason_1p1_rectified_flow.py new file mode 100644 index 0000000000000000000000000000000000000000..6c5a534b49dbe6f17d28827b9a511a2d06c06768 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_2B_reason_1p1_rectified_flow.py @@ -0,0 +1,969 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +import functools + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import ( + duplicate_batches, + duplicate_batches_random, + get_cached_replay_dataloader, +) +from cosmos_policy._src.predict2.datasets.dataset_provider import get_image_dataset, get_video_dataset +from cosmos_policy._src.predict2.datasets.joint_dataloader import IterativeJointDataLoader +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy +from cosmos_policy._src.predict2.text_encoders.text_encoder import EmbeddingConcatStrategy + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=1000, + logging_iter=50, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000000000000, + ), + every_n_sample_ema=dict( + every_n=1000000000000, + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_STANDALONE = LazyDict( + dict( + defaults=[ + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + {"override /model": "fsdp"}, + {"override /net": "cosmos_v1_2B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "fusedadamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_standalone", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.001, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100000], + ), + model=dict( + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + loss_scale=10.0, + adjust_video_noise=False, + scaling="rectified_flow", + sigma_data=1.0, + fsdp_shard_size=8, + resolution="720", + state_t=24, + resize_online=True, + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + rectified_flow_loss_weight_uniform=False, + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_2b_720_aggressive", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + use_empty_string=False, + ), + ), + sde=dict( + p_mean=1.6094379124341003, # math.log(5.0) + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + ), + ) + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000065000/", + load_training_state=False, + strict_resume=True, + ), + model_parallel=dict( + context_parallel_size=2, + ), + trainer=dict( + max_iter=100000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + callbacks=dict( + every_n_sample_reg=dict( + every_n=5000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + every_n_sample_ema=dict( + every_n=5000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + use_native_fps=True, + ), + ), + ratio=3, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW = LazyDict( + dict( + defaults=[ + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + {"override /model": "fsdp_rectified_flow"}, + {"override /net": "cosmos_v1_2B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "adamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only", + ), + optimizer=dict( + lr=3e-5, # 2**(-14.5) = 3.0517578125e-05 + weight_decay=1e-3, + betas=[0.9, 0.999], + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[100], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + fsdp_shard_size=8, + resolution="720", + state_t=24, + shift=5, + use_dynamic_shift=False, + train_time_weight="reweighting", + train_time_distribution="logitnormal", + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + timestep_scale=0.001, + sac_config=dict( + mode="predict2_2b_720_aggressive", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + use_wan_fp32_strategy=True, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + use_empty_string=False, # (TODO: hanzim): check + ), + ), + tokenizer=dict( + temporal_window=16, + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + ), + ) + ), + checkpoint=dict( + save_iter=1000, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + ), + model_parallel=dict( + context_parallel_size=2, + ), + trainer=dict( + max_iter=150_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + callbacks=dict( + grad_clip=dict( + clip_norm=0.1, + ), + manual_gc=dict( + every_n=200, + ), + every_n_sample_reg=dict( + every_n=1000000000000, + ), + every_n_sample_ema=dict( + every_n=1000000000000, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + use_native_fps=True, + ), + ), + ratio=3, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + + +# no resume 1, resume 2 is fix of the timestep argmin bug that misses dim=1 +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_RESUME2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_resume2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only/checkpoints/iter_000037000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) +# w/ high sigma +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-rectified_flow_only_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) + +# rf + image data (all res) +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_DEBUG1 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_debug1", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only/checkpoints/iter_000037000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + dataloader_train=L(IterativeJointDataLoader)( + dataloaders={ + "image_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_pretrain_and_synthetic_photoreal_20250805_image_whole", + object_store="s3", + resolution="${model.config.resolution}", + is_train=True, + caption_type="qwen2p5_7b_v4", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=12, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + ), + ratio=1, + ), + "image_data_prompt_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_synthetic_filtered_combined_20250805_image_whole", + object_store="s3", + resolution="${model.config.resolution}", + is_train=True, + caption_type="prompts", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=12, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + ), + ratio=1, + ), + "video_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name="cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_video_whole", + object_store="s3", + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + use_native_fps=True, + embedding_type=None, + is_train=True, + chunk_size=256, + ), + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + num_workers=2, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + ), + ratio=2, + ), + }, + ), + ), + flags={"allow_objects": True}, +) +# rf + gt720p data +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_DEBUG2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_debug2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only/checkpoints/iter_000037000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + dataset_resolution_type="gt720p", + ) + ) + ) + ), + ), + ), + flags={"allow_objects": True}, +) +# rf + new augmentor +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_DEBUG3 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_debug3", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only/checkpoints/iter_000037000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + dataset_resolution_type="all", + augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + ), + num_workers=2, + ) + ) + ), + ), + ), + flags={"allow_objects": True}, +) + + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + {"override /data_train": None}, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_improved", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only/checkpoints/iter_000037000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + dataloader_train=L(IterativeJointDataLoader)( + dataloaders={ + "image_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_pretrain_and_synthetic_photoreal_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="qwen2p5_7b_v4", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=12, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + ), + ratio=1, + ), + "image_data_prompt_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_synthetic_filtered_combined_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="prompts", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=12, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + ), + ratio=1, + ), + "video_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name="cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_video_whole", + object_store="s3", + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + # will use the augmentor to filter out frame drop + # so min and max fps can just use generic ones + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + # does not touch on low res data that will have jittering + dataset_resolution_type="gt720p", + use_native_fps=True, + embedding_type=None, + is_train=True, + chunk_size=256, + ), + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + num_workers=2, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + ), + ratio=2, + ), + }, + ), + ), + flags={"allow_objects": True}, +) + + +T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + {"override /data_train": None}, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_improved2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only/checkpoints/iter_000037000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + dataloader_train=L(IterativeJointDataLoader)( + dataloaders={ + "image_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_pretrain_and_synthetic_photoreal_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="qwen2p5_7b_v4", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=12, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + ), + ratio=1, + ), + "image_data_prompt_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_synthetic_filtered_combined_20250805_image_whole", + object_store="gcp", + resolution="${model.config.resolution}", + is_train=True, + caption_type="prompts", + dataset_resolution_type="gt720p", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=12, + num_workers=8, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + ), + ratio=1, + ), + "video_data": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name="cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_video_whole", + object_store="s3", + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + # does not touch on low res data that will have jittering + dataset_resolution_type="gt720p", + use_native_fps=True, + embedding_type=None, + is_train=True, + chunk_size=256, + ), + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + num_workers=8, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + ), + ratio=2, + ), + }, + ), + ), + flags={"allow_objects": True}, +) + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_RESUME2, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_RESUME2 + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED2, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED2 + ), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_DEBUG1, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_DEBUG1), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_DEBUG2, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_DEBUG2), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_DEBUG3, + *build_debug_runs(T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_DEBUG3), + ], + [ + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_HIGH_SIGMA, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_HIGH_SIGMA + ), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_7b_reason_1p1.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_7b_reason_1p1.py new file mode 100644 index 0000000000000000000000000000000000000000..5a1200addd25609ba6540014559de1385cc1f8c3 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/model_7b_reason_1p1.py @@ -0,0 +1,615 @@ +# Script for training 7B model from scratch + + +# ----------------------------------------------------------------------------- +# Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. +# All rights reserved. +# +# This codebase constitutes NVIDIA proprietary technology and is strictly +# confidential. Any unauthorized reproduction, distribution, or disclosure +# of this code, in whole or in part, outside NVIDIA is strictly prohibited +# without prior written consent. +# +# For inquiries regarding the use of this code in other NVIDIA proprietary +# projects, please contact the Deep Imagination Research Team at +# dir@exchange.nvidia.com. +# ----------------------------------------------------------------------------- + +# Configs for resuming from stage3 training + +import functools + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import ( + duplicate_batches_random, + get_cached_replay_dataloader, +) +from cosmos_policy._src.predict2.datasets.dataset_provider import get_image_dataset, get_video_dataset +from cosmos_policy._src.predict2.datasets.joint_dataloader import IterativeJointDataLoader +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy +from cosmos_policy._src.predict2.text_encoders.text_encoder import EmbeddingConcatStrategy + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=25, + logging_iter=2, + callbacks=dict( + every_n_sample_reg=dict( + every_n=20, + ), + every_n_sample_ema=dict( + every_n=20, + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +def build_gcp_config(job): + # This function contains the config changes needed for GCP usage. + gcp_config = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /checkpoint": "gcp"}, + {"override /tokenizer": "wan2pt1_tokenizer_gcp"}, + "_self_", + ], + model=dict( + config=dict( + text_encoder_config=dict( + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + s3_credential_path="credentials/gcp_checkpoint.secret", + ), + conditioner=dict( + text=dict( + empty_string_embeddings_path="s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt", + credential_path="credentials/gcp_training.secret", + ), + ), + ) + ), + job=dict( + name=( + job["job"]["name"].replace("_s3", "_gcp") + if "_s3" in job["job"]["name"] + else job["job"]["name"] + "_gcp" + ), + ), + ) + + gcp_config["dataloader_train"] = dict(dataloaders=dict()) + for dataset_name in job["dataloader_train"]["dataloaders"].keys(): + gcp_config["dataloader_train"]["dataloaders"][dataset_name] = dict( + dataloader=dict( + dataset=dict( + object_store="gcp", + ), + ), + ) + return [gcp_config] + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=cosmos_policy/_src/predict2/configs/video2world/config.py -- experiment=official_runs_text2world_401_7b_text_to_world_256res_s3 + +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --config=cosmos_policy/_src/predict2/configs/video2world/config.py -- experiment=official_runs_text2world_401_7b_text_to_world_256res_s3 job.group=debug +""" +T2W_7B_MODEL_401_256RES_S3: LazyDict = LazyDict( + dict( + defaults=[ + {"override /data_train": None}, + {"override /model": "fsdp"}, + {"override /net": "cosmos_v1_7B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "fusedadamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_text2world", + name="official_runs_text2world_401_7b_text_to_world_256res_s3", + ), + optimizer=dict( + lr=2 ** (-14), + weight_decay=0.05, + ), + scheduler=dict( + f_max=[1.0], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[300_000], + ), + model=dict( + config=dict( + min_num_conditional_frames=0, # choose either 1 (img2vid) or 2 (vid2vid) latent frames + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + loss_scale=10.0, + adjust_video_noise=True, + scaling="rectified_flow", + sigma_data=1.0, + fsdp_shard_size=8, + resolution="256", + state_t=16, + resize_online=True, + net=dict( + rope_enable_fps_modulation=False, + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.1, + use_empty_string=True, + ), + ), + sde=dict( + p_mean=0.0, + p_std=1.2, + sigma_max=80, + sigma_min=0.0002, + ), + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + text_encoder_class="reason1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ), + ) + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=500_000, + logging_iter=200, + callbacks=dict( + every_n_sample_reg=dict( + every_n=5000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + every_n_sample_ema=dict( + every_n=5000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + ), + ), + dataloader_train=L(IterativeJointDataLoader)( + dataloaders={ + "image_data_qwen_2p5_7b_v4_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_pretrain_and_synthetic_photoreal_20250805_image_whole", + object_store="s3", + resolution="256", + is_train=True, + caption_type="qwen2p5_7b_v4", + dataset_resolution_type="all", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=64, + num_workers=16, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + cache_size=32, + use_cache=False, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1), + ), + ratio=1, + ), + "image_data_prompt_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_synthetic_filtered_combined_20250805_image_whole", + object_store="s3", + resolution="256", + is_train=True, + caption_type="prompts", + dataset_resolution_type="all", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=64, + num_workers=16, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + cache_size=32, + use_cache=False, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1), + ), + ratio=1, + ), + "video_data_cosmos_pretrain": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name="cosmos_pretrainvideo_20250707_dedup_accumulated_and_high_quality_v3_202505_video_whole", + object_store="s3", + resolution="256", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=61, + dataset_resolution_type="all", + use_native_fps=True, + embedding_type=None, + is_train=True, + chunk_size=256, + ), + batch_size=6, + num_workers=8, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=2), + ), + ratio=2, + ), + } + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + + +T2W_7B_MODEL_402_256RES_S3_REASON_1P1 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2W_7B_MODEL_401_256RES_S3['job']['name']}", + "_self_", + ], + job=dict( + group=T2W_7B_MODEL_401_256RES_S3["job"]["group"], + name="official_runs_text2world_402_7b_reason_1p1_text_to_world_256res_s3", + ), + model=dict( + config=dict( + conditioner=dict( + text=dict( + dropout_rate=0.1, + use_empty_string=False, + ), + ), + text_encoder_class="reason1p1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ckpt_path="s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/", + s3_credential_path="credentials/s3_checkpoint.secret", + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/official_runs_text2world_401_7b_text_to_world_256res_gcp/checkpoints/iter_000050000/", + load_training_state=True, + strict_resume=True, + ), + dataloader_train=L(IterativeJointDataLoader)( + dataloaders={ + "image_data_qwen_2p5_7b_v4_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_pretrain_and_synthetic_photoreal_20250805_image_whole", + object_store="s3", + resolution="256", + is_train=True, + caption_type="qwen2p5_7b_v4", + dataset_resolution_type="all", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=64, + num_workers=16, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + cache_size=32, + use_cache=False, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1), + ), + ratio=1, + ), + "image_data_prompt_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_synthetic_filtered_combined_20250805_image_whole", + object_store="s3", + resolution="256", + is_train=True, + caption_type="prompts", + dataset_resolution_type="all", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=64, + num_workers=16, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + cache_size=32, + use_cache=False, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1), + ), + ratio=1, + ), + "video_data_cosmos_pretrain": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name="cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_video_whole", + object_store="s3", + resolution="256", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=61, + dataset_resolution_type="all", + use_native_fps=True, + embedding_type=None, + is_train=True, + chunk_size=256, + ), + batch_size=6, + num_workers=8, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=2), + ), + ratio=2, + ), + } + ), + ), + flags={"allow_objects": True}, +) + +# Model training in aws s3 cluster +T2W_7B_MODEL_403_256RES_S3_REASON_1P1 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2W_7B_MODEL_402_256RES_S3_REASON_1P1['job']['name']}", + "_self_", + ], + job=dict( + group=T2W_7B_MODEL_402_256RES_S3_REASON_1P1["job"]["group"], + name="official_runs_text2world_403_7b_reason_1p1_text_to_world_256res_s3", + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_text2world/official_runs_text2world_402_7b_reason_1p1_text_to_world_256res_s3/checkpoints/iter_000055000/", + load_training_state=True, + strict_resume=True, + ), + dataloader_train=L(IterativeJointDataLoader)( + dataloaders={ + "image_data_qwen_2p5_7b_v4_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_pretrain_and_synthetic_photoreal_20250805_image_whole", + object_store="s3", + resolution="256", + is_train=True, + caption_type="qwen2p5_7b_v4", + dataset_resolution_type="all", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=24, + num_workers=16, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + cache_size=32, + use_cache=False, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1), + ), + ratio=1, + ), + "image_data_prompt_captions": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_image_dataset)( + dataset_name="cosmos_synthetic_filtered_combined_20250805_image_whole", + object_store="s3", + resolution="256", + is_train=True, + caption_type="prompts", + dataset_resolution_type="all", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + batch_size=24, + num_workers=16, + prefetch_factor=4, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="image_dataloader", + cache_size=32, + use_cache=False, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1), + ), + ratio=1, + ), + "video_data_cosmos_pretrain": dict( + dataloader=L(get_cached_replay_dataloader)( + dataset=L(get_video_dataset)( + dataset_name="cosmos_pretrainvideo_20250806_dedup_accumulated_and_high_quality_v3_202505_video_whole", + object_store="s3", + resolution="256", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=61, + dataset_resolution_type="all", + use_native_fps=True, + embedding_type=None, + is_train=True, + chunk_size=256, + ), + batch_size=2, + num_workers=8, + prefetch_factor=2, + sampler=None, + persistent_workers=False, + pin_memory=True, + cache_replay_name="video_dataloader", + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1), + ), + ratio=2, + ), + } + ), + ), + flags={"allow_objects": True}, +) + + +cs = ConfigStore.instance() +for _item, _item_wo_resume, _item_mock_wo_resume, _item_gcp in [ + [ + T2W_7B_MODEL_401_256RES_S3, + *build_debug_runs(T2W_7B_MODEL_401_256RES_S3), + *build_gcp_config(T2W_7B_MODEL_401_256RES_S3), + ], + [ + T2W_7B_MODEL_402_256RES_S3_REASON_1P1, + *build_debug_runs(T2W_7B_MODEL_402_256RES_S3_REASON_1P1), + *build_gcp_config(T2W_7B_MODEL_402_256RES_S3_REASON_1P1), + ], + [ + T2W_7B_MODEL_403_256RES_S3_REASON_1P1, + *build_debug_runs(T2W_7B_MODEL_403_256RES_S3_REASON_1P1), + *build_gcp_config(T2W_7B_MODEL_403_256RES_S3_REASON_1P1), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) + if _item_gcp is not None: + cs.store( + group="experiment", + package="_global_", + name=_item_gcp["job"]["name"], + node=_item_gcp, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/stage3_14B_index_3.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/stage3_14B_index_3.py new file mode 100644 index 0000000000000000000000000000000000000000..224f1236d6f3f052327f4d074c956b75eca7cdbc --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/stage3_14B_index_3.py @@ -0,0 +1,958 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +import functools +import math + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import duplicate_batches, duplicate_batches_random +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy +from cosmos_policy._src.predict2.text_encoders.text_encoder import EmbeddingConcatStrategy + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=25, + logging_iter=2, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1, + ), + every_n_sample_ema=dict( + every_n=1, + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/vid2vid/config.py -- experiment=Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor + +# diff with resumed config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/vid2vid/config.py -- experiment=Stage-a_pt_3-Vid2Vid-Index-5-Size-14B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond +""" +I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR: LazyDict = LazyDict( + dict( + defaults=[ + {"override /data_train": "image_cosmos_pretrain_qwen_20250415_video_cosmos_pretrain_v1_3_20250426_s3"}, + {"override /model": "fsdp"}, + {"override /net": "cosmos_v1_14B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "fusedadamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor", + ), + optimizer=dict( + lr=2 ** (-14.5), + weight_decay=0.2, + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[300_000], + ), + model=dict( + config=dict( + min_num_conditional_frames=1, # choose either 1 (img2vid) or 2 (vid2vid) latent frames + max_num_conditional_frames=2, + loss_scale=10.0, + adjust_video_noise=True, + scaling="rectified_flow", + sigma_data=1.0, + fsdp_shard_size=32, + resolution="480", + state_t=20, + resize_online=True, + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=20.0 / 24, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + ) + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-a_pt_3-Vid2Vid-Index-5-Size-14B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond/checkpoints/iter_000052500", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=4, + ), + trainer=dict( + max_iter=200_000, + logging_iter=200, + callbacks=dict( + every_n_sample_reg=dict( + every_n=500000000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + every_n_sample_ema=dict( + every_n=500000000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=20 // 4, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="480", + ), + ), + ratio="${trainer.grad_accum_iter}", + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1), + dataset=dict( + resolution="480", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + use_native_fps=True, + ), + ), + ratio="${trainer.grad_accum_iter}", + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + + +""" +# print config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/vid2vid/config.py -- experiment=Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40 +""" +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + # "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250707_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_qwen_concat", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40/checkpoints/iter_000018000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + text_encoder_class="reason1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=200_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_NATIVE_FPS: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_native_fps_qwen_concat", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_qwen_concat/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + trainer=dict( + max_iter=200_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +################################################################################ + +# 14B, 720p, 10fps +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_100_SIZE_14B_RES_720_FPS10_T15_HQV5_from_43: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250707_dedup_accumulated_and_high_quality_v3_202505_s3" + }, # @qinsheng + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-100-Size-14B-Res-720-Fps-10-Note-T15_HQV5_from_43_qwen_concat", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-100-Size-14B-Res-720-Fps-10-Note-T15_HQV5_from_43/checkpoints/iter_000010000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=16, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=16.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + text_encoder_class="reason1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=61, + dataset_resolution_type="all", + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +# 14B, 480p, 10fps +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_101_SIZE_14B_RES_480_FPS10_T15_HQV5_from_43: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250707_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-101-Size-14B-Res-480-Fps-10-Note-T15_HQV5_from_43_qwen_concat", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-101-Size-14B-Res-480-Fps-10-Note-T15_HQV5_from_43/checkpoints/iter_000010000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="480", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=16, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=16.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + text_encoder_class="reason1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=61, + dataset_resolution_type="all", + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +# 14B, 480p, 16fps +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_102_SIZE_14B_RES_480_FPS16_T24_HQV5_from_43: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250707_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-102-Size-14B-Res-480-Fps-16-Note-T24_HQV5_from_43_qwen_concat", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-102-Size-14B-Res-480-Fps-16-Note-T24_HQV5_from_43/checkpoints/iter_000010000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="480", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + text_encoder_class="reason1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + trainer=dict( + max_iter=200_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +# Post noise fix +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_NATIVE_FPS_LOSS_REWEIGHTED: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_NATIVE_FPS['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_NATIVE_FPS["job"][ + "group" + ], + name="Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_native_fps_qwen_concat_loss_reweighted", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_native_fps_qwen_concat/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=False, + ), + scheduler=dict( + f_max=[0.3], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[100000], + ), + trainer=dict( + max_iter=100000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + model=dict( + config=dict( + adjust_video_noise=False, + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + rectified_flow_loss_weight_uniform=False, + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_V2: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + # "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250707_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_qwen_concat_v2", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40_native_fps_qwen_concat/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + min_num_conditional_frames=0, # choose either 1 (img2vid) or 2 (vid2vid) latent frames + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + scaling="rectified_flow", + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + text_encoder_class="reason1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + trainer=dict( + max_iter=200_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +cs = ConfigStore.instance() +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_NATIVE_FPS, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_NATIVE_FPS + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_100_SIZE_14B_RES_720_FPS10_T15_HQV5_from_43, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_100_SIZE_14B_RES_720_FPS10_T15_HQV5_from_43), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_101_SIZE_14B_RES_480_FPS10_T15_HQV5_from_43, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_101_SIZE_14B_RES_480_FPS10_T15_HQV5_from_43), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_102_SIZE_14B_RES_480_FPS16_T24_HQV5_from_43, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_102_SIZE_14B_RES_480_FPS16_T24_HQV5_from_43), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_NATIVE_FPS_LOSS_REWEIGHTED, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_NATIVE_FPS_LOSS_REWEIGHTED + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_V2, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40_V2), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/stage3_2B.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/stage3_2B.py new file mode 100644 index 0000000000000000000000000000000000000000..35dfa018310220d3ba0d9829b273804c81a9803f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/reason_embeddings/stage3_2B.py @@ -0,0 +1,1828 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +import functools +import math + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import duplicate_batches, duplicate_batches_random +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy +from cosmos_policy._src.predict2.text_encoders.text_encoder import EmbeddingConcatStrategy + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=1000, + logging_iter=50, + callbacks=dict( + every_n_sample_reg=dict( + every_n=10, + ), + every_n_sample_ema=dict( + every_n=10, + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=cosmos_policy/_src/predict2/configs/video2world/config.py -- experiment=Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4 +""" +I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY: LazyDict = LazyDict( + dict( + defaults=[ + {"override /data_train": "mock"}, + {"override /model": "fsdp"}, + {"override /net": "cosmos_v1_2B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "fusedadamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + ), + optimizer=dict( + lr=2 ** (-14), # 2**(-14) = 6.103515625e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + min_num_conditional_frames=1, # choose either 1 (img2vid) or 2 (vid2vid) latent frames + max_num_conditional_frames=2, + loss_scale=10.0, + adjust_video_noise=True, + scaling="rectified_flow", + sigma_data=1.0, + fsdp_shard_size=8, + resolution="480", + state_t=24, + resize_online=True, + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=1.0, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.3, + ), + ), + ) + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs/Stage-a_pt_3-Index-1-Size-2B-Res-480-Fps-16-Note-qwen_imagecaption/checkpoints/iter_000057500", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=1, + ), + trainer=dict( + max_iter=250_000, + logging_iter=200, + callbacks=dict( + every_n_sample_reg=dict( + every_n=500000000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + every_n_sample_ema=dict( + every_n=500000000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=24, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="${model.config.resolution}", + ), + ), + ratio=0, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=True, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.5), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + caption_type="i2w_qwen2p5_7b_later_frames", + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + {"override /data_train": "mock"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720_aggressive", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + text_encoder_class="reason1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22/checkpoints/iter_000026000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=200_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_RESUME4: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4", + ), + trainer=dict( + max_iter=200_000, + logging_iter=50, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + use_cache=False, + ), + ratio=1, + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat/checkpoints/iter_000010000", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_1E2: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_RESUME4['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_RESUME4["job"][ + "group" + ], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_1e2", + ), + trainer=dict( + max_iter=100_000, + logging_iter=50, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.01, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100_000], + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_1E3: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_RESUME4['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_RESUME4["job"][ + "group" + ], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_1e3", + ), + trainer=dict( + max_iter=100_000, + logging_iter=50, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.001, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100_000], + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma", + ), + trainer=dict( + max_iter=100000, + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.001, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100000], + ), + model=dict( + config=dict( + sde=dict( + p_mean=math.log(5.5), + p_std=1.1, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid_debug/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma/checkpoints/iter_000002500", + load_training_state=True, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted", + ), + trainer=dict( + max_iter=100000, + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.001, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100000], + ), + model=dict( + config=dict( + sde=dict( + p_mean=math.log(5.5), + p_std=1.1, + sigma_max=200, + sigma_min=0.01, + ), + rectified_flow_loss_weight_uniform=False, + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_resume2", + ), + trainer=dict( + max_iter=100000, + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.001, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100000], + ), + model=dict( + config=dict( + adjust_video_noise=False, + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + rectified_flow_loss_weight_uniform=False, + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted/checkpoints/iter_000015000/", + load_training_state=True, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +# distribution matching + loss weight, on general pretraining dataset +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_GENERAL = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_v2_general", + ), + model=dict( + config=dict( + scaling="rectified_flow", # correct loss weight for rectified flow + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_GENERAL_RESUME1 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_GENERAL['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_v2_general_resume1", + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_v2_general/checkpoints/iter_000002500/", + load_training_state=True, + strict_resume=True, + ), + trainer=dict( + straggler_detection=dict( + enabled=False, + ), + logging_iter=20, + ), + model=dict( + config=dict( + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.2, + ), + ), + ) + ), + ), + flags={"allow_objects": True}, +) +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_AV = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2['job']['name']}", + { + "override /data_train": "mock", + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_v2_av", + ), + model=dict( + config=dict( + scaling="rectified_flow", # correct loss weight for rectified flow + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_AV_RESUME1 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_AV['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_v2_av_resume1", + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_v2_av/checkpoints/iter_000007500/", + load_training_state=True, + strict_resume=True, + ), + trainer=dict( + straggler_detection=dict( + enabled=False, + ), + logging_iter=20, + ), + model=dict( + config=dict( + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.2, + ), + ), + ) + ), + ), + flags={"allow_objects": True}, +) +# distribution matching + loss weight + wan fp32, on general pretraining dataset +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V3_GENERAL_CP4 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_v3_general_cp4", + ), + model=dict( + config=dict( + scaling="rectified_flow", # correct loss weight for rectified flow + # use wan fp32 + use_wan_fp32_strategy=True, + net=dict( + use_wan_fp32_strategy=True, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.2, + ), + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + model_parallel=dict( + context_parallel_size=4, + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V3_AV_CP4 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2['job']['name']}", + { + "override /data_train": "mock", + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_v3_av_cp4", + ), + model=dict( + config=dict( + scaling="rectified_flow", # correct loss weight for rectified flow + # use wan fp32 + use_wan_fp32_strategy=True, + net=dict( + use_wan_fp32_strategy=True, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.2, + ), + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + model_parallel=dict( + context_parallel_size=4, + ), + ), + flags={"allow_objects": True}, +) +# distribution matching + loss weight + wan fp32 + empty string embedding, on general pretraining dataset +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V4_GENERAL_CP4 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_v4_general_cp4", + ), + model=dict( + config=dict( + scaling="rectified_flow", # correct loss weight for rectified flow + # use wan fp32 + use_wan_fp32_strategy=True, + net=dict( + use_wan_fp32_strategy=True, + ), + conditioner=dict( + text=dict( + use_empty_string=True, + ), + use_video_condition=dict( + dropout_rate=0.2, + ), + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + model_parallel=dict( + context_parallel_size=4, + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V4_AV_CP4 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2['job']['name']}", + { + "override /data_train": "mock", + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_v4_av_cp4", + ), + model=dict( + config=dict( + scaling="rectified_flow", # correct loss weight for rectified flow + # use wan fp32 + use_wan_fp32_strategy=True, + net=dict( + use_wan_fp32_strategy=True, + ), + conditioner=dict( + text=dict( + use_empty_string=True, + ), + use_video_condition=dict( + dropout_rate=0.2, + ), + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + model_parallel=dict( + context_parallel_size=4, + ), + ), + flags={"allow_objects": True}, +) +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma", + ), + trainer=dict( + max_iter=100000, + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.001, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100000], + ), + model=dict( + config=dict( + sde=dict( + p_mean=math.log(5.5), + p_std=1.1, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid_debug/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma/checkpoints/iter_000002500", + load_training_state=True, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted", + ), + trainer=dict( + max_iter=100000, + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.001, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100000], + ), + model=dict( + config=dict( + sde=dict( + p_mean=math.log(5.5), + p_std=1.1, + sigma_max=200, + sigma_min=0.01, + ), + rectified_flow_loss_weight_uniform=False, + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_resume2", + ), + trainer=dict( + max_iter=100000, + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.001, + ), + scheduler=dict( + f_max=[0.5], + f_min=[0.2], + warm_up_steps=[2_000], + cycle_lengths=[100000], + ), + model=dict( + config=dict( + adjust_video_noise=False, + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + rectified_flow_loss_weight_uniform=False, + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted/checkpoints/iter_000015000/", + load_training_state=True, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +# distribution matching + loss weight + use empty string, on general pretraining dataset +T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_EMPTY_STRING = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_empty_string", + ), + model=dict( + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + scaling="rectified_flow", # correct loss weight for rectified flow + conditioner=dict( + text=dict( + use_empty_string=True, + ), + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_wd_high_sigma_loss_reweighted_resume2/checkpoints/iter_000055000", + load_training_state=False, + strict_resume=True, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + use_cache=False, + ), + ratio=3, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +# distribution matching + loss weight, on general pretraining dataset +T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_EMPTY_STRING['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted", + ), + model=dict( + config=dict( + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + conditioner=dict( + text=dict( + use_empty_string=False, + ), + ), + ), + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000032500/", + ), + ), + flags={"allow_objects": True}, +) + + +################################################################################ +# Configs for different resolutions and fps + +# 2B, 720p, 10fps +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_100_SIZE_2B_RES_720_FPS10_HQ_V5_from_26 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + {"override /data_train": "mock"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-100-Size-2B-Res-720-Fps-10-Note-HQ_V5_from_26_qwen_concat", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=16, + resize_online=True, + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720_aggressive", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=16.0 / 24, + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + text_encoder_class="reason1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-100-Size-2B-Res-720-Fps-10-Note-HQ_V5_from_26/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=200_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=61, + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +# 2B, 480p, 10fps +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_101_SIZE_2B_RES_480_FPS10_HQ_V5_from_26 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + {"override /data_train": "mock"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-101-Size-2B-Res-480-Fps-10-Note-HQ_V5_from_26_qwen_concat", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="480", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=16, + resize_online=True, + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=16.0 / 24, + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + text_encoder_class="reason1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-101-Size-2B-Res-480-Fps-10-Note-HQ_V5_from_26/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=200_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=61, + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +# 2B, 480p, 16fps +I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_102_SIZE_2B_RES_480_FPS16_HQ_V5_from_26 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + {"override /data_train": "mock"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-102-Size-2B-Res-480-Fps-16-Note-HQ_V5_from_26_qwen_concat", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[200_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="480", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + use_crossattn_projection=True, + crossattn_proj_in_channels=100352, + crossattn_emb_channels=1024, + ), + text_encoder_class="reason1_7B", + text_encoder_config=dict( + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + compute_online=True, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-102-Size-2B-Res-480-Fps-16-Note-HQ_V5_from_26/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=200_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type=None, + augmentor_name="image_basic_augmentor_without_embeddings", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type=None, + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +######################################################## +# Post training cooldown configs + +# distribution matching + loss weight, on general pretraining dataset +T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_HQ_COOLDOWN_FROM_57K = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16_hq_cooldown_from_57K", + ), + model=dict( + config=dict( + conditioner=dict( + text=dict( + use_empty_string=False, + ), + ), + ), + ), + trainer=dict( + max_iter=50_000, + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.0], + warm_up_steps=[0], + cycle_lengths=[50_000], + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000057500/", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + +T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_HQ_COOLDOWN_FROM_57K_4K_VIDEOS = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_HQ_COOLDOWN_FROM_57K['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16_hq_cooldown_from_57K_4K_videos", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000057500/", + load_training_state=False, + strict_resume=True, + ), + ), + flags={"allow_objects": True}, +) + + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_RESUME4, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_RESUME4), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_100_SIZE_2B_RES_720_FPS10_HQ_V5_from_26, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_100_SIZE_2B_RES_720_FPS10_HQ_V5_from_26), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_101_SIZE_2B_RES_480_FPS10_HQ_V5_from_26, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_101_SIZE_2B_RES_480_FPS10_HQ_V5_from_26), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_102_SIZE_2B_RES_480_FPS16_HQ_V5_from_26, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_102_SIZE_2B_RES_480_FPS16_HQ_V5_from_26), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_1E3, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_1E3), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_1E2, + *build_debug_runs(I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_1E2), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2 + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_GENERAL, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_GENERAL + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_GENERAL_RESUME1, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_GENERAL_RESUME1 + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_AV, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_AV + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_AV_RESUME1, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V2_AV_RESUME1 + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V3_GENERAL_CP4, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V3_GENERAL_CP4 + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V3_AV_CP4, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V3_AV_CP4 + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V4_GENERAL_CP4, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V4_GENERAL_CP4 + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V4_AV_CP4, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_V4_AV_CP4 + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED + ), + ], + [ + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2, + *build_debug_runs( + I2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_RESUME2 + ), + ], + [ + T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_EMPTY_STRING, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED_EMPTY_STRING + ), + ], + [ + T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED + ), + ], + [ + T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_HQ_COOLDOWN_FROM_57K, + *build_debug_runs(T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_HQ_COOLDOWN_FROM_57K), + ], + [ + T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_HQ_COOLDOWN_FROM_57K_4K_VIDEOS, + *build_debug_runs( + T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_HQ_COOLDOWN_FROM_57K_4K_VIDEOS + ), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/sparse_14B.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/sparse_14B.py new file mode 100644 index 0000000000000000000000000000000000000000..0737ac3ae38fc3a7a273c4a525a25b14d7c2c3b4 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/sparse_14B.py @@ -0,0 +1,969 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training with sparse attention + +import math + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy + +#################################################################################################################### +# NATTEN / Sparse Attention configurations for 14B MinimalDiTV4 +#################################################################################################################### + +NATTEN_PARAMETERS_14B_COMB01 = [ + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 0 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 1 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 2 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 3 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 4 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 5 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 6 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 7 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 8 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 9 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 10 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 11 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 12 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 13 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 14 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 15 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 16 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 17 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 18 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 19 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 20 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 21 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 22 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 23 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 24 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 25 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 26 + None, # layer 27 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 28 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 29 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 30 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 31 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 32 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 33 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 34 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 35 +] + +NATTEN_PARAMETERS_14B_COMB02 = [ + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 0 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 1 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 2 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 3 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 4 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 5 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 6 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 7 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 8 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 9 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 10 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 11 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 12 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 13 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 14 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 15 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 16 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 17 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 18 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 19 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 20 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 21 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 22 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 23 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 24 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 25 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 26 + None, # layer 27 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 28 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 29 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 30 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 31 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 32 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 33 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 34 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 35 +] + +#################################################################################################################### + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=25, + logging_iter=2, + callbacks=dict( + every_n_sample_reg=dict( + every_n=12, + ), + every_n_sample_ema=dict( + every_n=12, + ), + reg_model_image2video_sora_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_sora_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + reg_model_image2video_vbench_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_vbench_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +I2V_STAGE_C_PT_4_INDEX_200_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_from_40_SparseAttn_9Dense: LazyDict = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_1_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-200-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_40_sparse-attn_9dense", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-38-Size-14B-Res-720-Fps-16-Note-T24_HQV3_from_35", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + n_dense_blocks=9, + natten_parameters={"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_201_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_43_SparseAttn_Comb01_1dense: LazyDict = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-201-Size-14B-Res-720-Fps-16-Note-T24_HQV2_1_from_43_sparse-attn_comb01-1dense", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38/checkpoints/iter_000015000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_14B_COMB01, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_202_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_43_SparseAttn_Comb02_1dense: LazyDict = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-202-Size-14B-Res-720-Fps-16-Note-T24_HQV2_1_from_43_sparse-attn_comb02-1dense", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38/checkpoints/iter_000015000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_14B_COMB02, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_203_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_from_40_SparseAttn_Comb02_1dense: LazyDict = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_1_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-203-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_40_sparse-attn_comb02-1dense", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-38-Size-14B-Res-720-Fps-16-Note-T24_HQV3_from_35", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_14B_COMB02, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_204_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_203_SparseAttn_Comb02_1dense: LazyDict = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-204-Size-14B-Res-720-Fps-16-Note-T24_HQV2_1_from_203_sparse-attn_comb02-1dense", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid_gna/Stage-c_pt_4-Index-203-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_40_sparse-attn_comb02-1dense/checkpoints/iter_000031000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_14B_COMB02, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[50_001], + ), + trainer=dict( + max_iter=50_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +# Final Configuration +# Alternative LR schedule +I2V_STAGE_C_PT_4_INDEX_205_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_203_SparseAttn_Comb02_1dense_altLR: LazyDict = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-205-Size-14B-Res-720-Fps-16-Note-T24_HQV2_1_from_203_sparse-attn_comb02-1dense-altLR", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid_gna/Stage-c_pt_4-Index-203-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_40_sparse-attn_comb02-1dense/checkpoints/iter_000031000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_14B_COMB02, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +# 10 FPS fine-tune +# Copied from index 100 (used to train official 10fps checkpoint), added NATTEN/sparsity, +# swapped initial checkpoint with the final 14B w/ sparsity. +I2V_STAGE_C_PT_4_INDEX_230_SIZE_14B_RES_720_FPS10_T15_HQV5_from_205_SparseAttn_Comb02_1dense: LazyDict = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-230-Size-14B-Res-720-Fps-10-Note-T15_HQV5_from_205_sparse-attn_comb02-1dense", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid_gna/Stage-c_pt_4-Index-205-Size-14B-Res-720-Fps-16-Note-T24_HQV2_1_from_203_sparse-attn_comb02-1dense-altLR/checkpoints/iter_000018000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=16, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=16.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_14B_COMB02, + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=61, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + I2V_STAGE_C_PT_4_INDEX_200_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_from_40_SparseAttn_9Dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_200_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_from_40_SparseAttn_9Dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_201_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_43_SparseAttn_Comb01_1dense, + *build_debug_runs( + I2V_STAGE_C_PT_4_INDEX_201_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_43_SparseAttn_Comb01_1dense + ), + ], + [ + I2V_STAGE_C_PT_4_INDEX_202_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_43_SparseAttn_Comb02_1dense, + *build_debug_runs( + I2V_STAGE_C_PT_4_INDEX_202_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_43_SparseAttn_Comb02_1dense + ), + ], + [ + I2V_STAGE_C_PT_4_INDEX_203_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_from_40_SparseAttn_Comb02_1dense, + *build_debug_runs( + I2V_STAGE_C_PT_4_INDEX_203_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_from_40_SparseAttn_Comb02_1dense + ), + ], + [ + I2V_STAGE_C_PT_4_INDEX_204_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_203_SparseAttn_Comb02_1dense, + *build_debug_runs( + I2V_STAGE_C_PT_4_INDEX_204_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_203_SparseAttn_Comb02_1dense + ), + ], + [ + I2V_STAGE_C_PT_4_INDEX_205_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_203_SparseAttn_Comb02_1dense_altLR, + *build_debug_runs( + I2V_STAGE_C_PT_4_INDEX_205_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_203_SparseAttn_Comb02_1dense_altLR + ), + ], + [ + I2V_STAGE_C_PT_4_INDEX_230_SIZE_14B_RES_720_FPS10_T15_HQV5_from_205_SparseAttn_Comb02_1dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_230_SIZE_14B_RES_720_FPS10_T15_HQV5_from_205_SparseAttn_Comb02_1dense), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/sparse_2B.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/sparse_2B.py new file mode 100644 index 0000000000000000000000000000000000000000..bc003ef259f4df117ab67880f9f8fc73ebcd2bda --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/sparse_2B.py @@ -0,0 +1,2247 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training with sparse attention + +import functools +import math + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import duplicate_batches_random +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy + +#################################################################################################################### +# NATTEN / Sparse Attention configurations for 2B MinimalDiTV4 +#################################################################################################################### + +NATTEN_PARAMETERS_2B_COMB01 = [ + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 0, 90% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 1, 50% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 2, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 3, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 4, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 5, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 6, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 7, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 8, 90% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 9, 50% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 10, 90% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 11, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 12, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 13, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 14, 50% + None, # layer 15, SA + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 16, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 17, 50% + None, # layer 18, SA + None, # layer 19, SA + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 20, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 21, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 22, 50% + None, # layer 23, SA + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 24, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 25, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 26, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 27, 50% +] + +# Final chosen config for Predict2 2B +NATTEN_PARAMETERS_2B_COMB02 = [ + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 0 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 1 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 2 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 3 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 4 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 5 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 6 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 7 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 8 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 9 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 10 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 11 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 12 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 13 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 14 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 15 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 16 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 17 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 18 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 19 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 20 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 21 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 22 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 23 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 24 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 25 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 26 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 27 +] + +NATTEN_PARAMETERS_2B_COMB03 = [ + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 0 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 1 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 2 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 3 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 4 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 5 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 6 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 7 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 8 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 9 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 10 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 11 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 12 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 13 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 14 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 15 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 16 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 17 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 18 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 19 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 20 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 21 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 22 + None, # layer 23 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 24 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 25 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 26 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 27 +] + +NATTEN_PARAMETERS_2B_COMB04 = [ + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 0 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 1 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 2 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 3 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 4 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 5 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 6 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 7 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 8 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 9 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 10 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 11 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 12 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 13 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 14 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 15 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 16 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 17 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 18 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 19 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 20 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 21 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 22 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 23 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 24 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 25 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 26 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 27 +] + +NATTEN_PARAMETERS_2B_COMB05 = [ + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # layer 0 + {"window_size": (-1, 4, 24), "stride": (1, 1, 8), "dilation": (1, 11, 1), "base_size": (-1, 44, 80)}, # layer 1 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 2 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 3 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 4 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 5 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 6 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 7 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 8 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 9 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # layer 10 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 11 + None, # layer 12 + None, # layer 13 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 14 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 15 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 16 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 17 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 18 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 19 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 20 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 21 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 22 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 23 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 24 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 25 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 26 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # layer 27 +] + +#################################################################################################################### + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=25, + logging_iter=2, + callbacks=dict( + every_n_sample_reg=dict( + every_n=12, + ), + every_n_sample_ema=dict( + every_n=12, + ), + reg_model_image2video_sora_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_sora_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + reg_model_image2video_vbench_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_vbench_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +I2V_STAGE_C_PT_4_INDEX_200_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_7Dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-200-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_7dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=7, + natten_parameters={"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_201_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_6Dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-201-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_6dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=6, + natten_parameters={"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_202_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_4Dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-202-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_4dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=4, + natten_parameters={"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_203_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_9Dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-203-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_9dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=9, + natten_parameters={"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_204_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_12Dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-204-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_12dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=12, + natten_parameters={"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_205_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_7Dense_NA = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-205-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_7dense-na", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=7, + natten_parameters={"window_size": (-1, 12, 24), "stride": (1, 1, 1), "base_size": (-1, 44, 80)}, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_206_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb01_4dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-206-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_comb01-4dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_2B_COMB01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_207_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb02_0dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-207-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_comb02-0dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_2B_COMB02, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_208_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb03_1dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-208-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_comb03-1dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_2B_COMB03, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_209_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb04_0dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-209-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_comb04-0dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_2B_COMB04, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_210_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb05_2dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-210-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_comb05-2dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_2B_COMB05, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_211_SIZE_2B_RES_720_FPS16_HQ_V6_from_207_SparseAttn_Comb02_0dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v6_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-211-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_207_sparse-attn_comb02-0dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_2B_COMB02, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid_gna/Stage-c_pt_4-Index-207-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_comb02-0dense/checkpoints/iter_000032000", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=48_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +# Final Configuration +I2V_STAGE_C_PT_4_INDEX_212_SIZE_2B_RES_720_FPS16_HQ_V6_from_207_SparseAttn_Comb02_0dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v6_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-212-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_207_sparse-attn_comb02-0dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[60_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_2B_COMB02, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid_gna/Stage-c_pt_4-Index-207-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_comb02-0dense/checkpoints/iter_000032000", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=60_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_213_SIZE_2B_RES_720_FPS16_HQ_V6_from_207_SparseAttn_Comb02_0dense_altLR = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v6_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-213-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_207_sparse-attn_comb02-0dense-altLR", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_2B_COMB02, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid_gna/Stage-c_pt_4-Index-207-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_22_sparse-attn_comb02-0dense/checkpoints/iter_000032000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=26_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +# 10 FPS fine-tune +# Copied from index 100 (used to train official 10fps checkpoint), added NATTEN/sparsity, +# swapped initial checkpoint with the final 14B w/ sparsity. +I2V_STAGE_C_PT_4_INDEX_230_SIZE_2B_RES_720_FPS10_HQ_V5_from_212_SparseAttn_Comb02_0dense = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-230-Size-2B-Res-720-Fps-10-Note-HQ_V5_from_212_sparse-attn_comb02-0dense", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=16, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=16.0 / 24, + n_dense_blocks=0, + natten_parameters=NATTEN_PARAMETERS_2B_COMB02, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid_gna/Stage-c_pt_4-Index-212-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_207_sparse-attn_comb02-0dense/checkpoints/iter_000050000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=26_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=61, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + I2V_STAGE_C_PT_4_INDEX_200_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_7Dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_200_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_7Dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_201_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_6Dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_201_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_6Dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_202_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_4Dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_202_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_4Dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_203_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_9Dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_203_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_9Dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_204_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_12Dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_204_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_12Dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_205_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_7Dense_NA, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_205_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_7Dense_NA), + ], + [ + I2V_STAGE_C_PT_4_INDEX_206_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb01_4dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_206_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb01_4dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_207_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb02_0dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_207_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb02_0dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_208_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb03_1dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_208_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb03_1dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_209_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb04_0dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_209_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb04_0dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_210_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb05_2dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_210_SIZE_2B_RES_720_FPS16_HQ_V3_from_22_SparseAttn_Comb05_2dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_211_SIZE_2B_RES_720_FPS16_HQ_V6_from_207_SparseAttn_Comb02_0dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_211_SIZE_2B_RES_720_FPS16_HQ_V6_from_207_SparseAttn_Comb02_0dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_212_SIZE_2B_RES_720_FPS16_HQ_V6_from_207_SparseAttn_Comb02_0dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_212_SIZE_2B_RES_720_FPS16_HQ_V6_from_207_SparseAttn_Comb02_0dense), + ], + [ + I2V_STAGE_C_PT_4_INDEX_213_SIZE_2B_RES_720_FPS16_HQ_V6_from_207_SparseAttn_Comb02_0dense_altLR, + *build_debug_runs( + I2V_STAGE_C_PT_4_INDEX_213_SIZE_2B_RES_720_FPS16_HQ_V6_from_207_SparseAttn_Comb02_0dense_altLR + ), + ], + [ + I2V_STAGE_C_PT_4_INDEX_230_SIZE_2B_RES_720_FPS10_HQ_V5_from_212_SparseAttn_Comb02_0dense, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_230_SIZE_2B_RES_720_FPS10_HQ_V5_from_212_SparseAttn_Comb02_0dense), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/stage3.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/stage3.py new file mode 100644 index 0000000000000000000000000000000000000000..2bf4b1bdf7660330a094dc07fd3b34103c796a92 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/stage3.py @@ -0,0 +1,521 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +import functools + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import duplicate_batches, duplicate_batches_random + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=25, + logging_iter=2, + callbacks=dict( + every_n_sample_reg=dict( + every_n=12, + ), + every_n_sample_ema=dict( + every_n=12, + ), + reg_model_image2video_sora_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_sora_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + reg_model_image2video_vbench_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_vbench_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +Text2World_STAGE_A_PT_3_INDEX_1_SIZE_2B_RES_480_FPS16_QWEN_IMAGE: LazyDict = LazyDict( + dict( + defaults=[ + {"override /data_train": "image_cosmos_pretrain_qwen_20250415_video_cosmos_pretrain_20241219_s3"}, + {"override /model": "fsdp"}, + {"override /net": "cosmos_v1_2B"}, + {"override /conditioner": "add_fps_padding_mask"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "fusedadamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs", + name="Stage-a_pt_3-Index-1-Size-2B-Res-480-Fps-16-Note-qwen_imagecaption", + ), + optimizer=dict( + lr=2 ** (-14), # 2**(-14) = 6.103515625e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + loss_scale=10.0, + adjust_video_noise=True, + scaling="rectified_flow", + sigma_data=1.0, + fsdp_shard_size=8, + resolution="480", + state_t=24, + resize_online=True, + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=1.0, + ), + ) + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/ablation_2B_0321_tokenizer_data/ablation_2B_0321_tokenizer_data_204pt1_wan2pt1tokenizer_resizeT24_Res480V250322-241219/checkpoints/iter_000150000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=150_000, + logging_iter=200, + callbacks=dict( + every_n_sample_reg=dict( + every_n=5_000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + every_n_sample_ema=dict( + every_n=5_000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=24, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="480", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.1), + dataset=dict( + resolution="480", + num_video_frames=93, + video_decoder_name="chunked_video_decoder_with_fixed_fps", + min_fps_thres=16, + max_fps_thres=45, + ), + ), + ratio=1, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + +Text2World_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{Text2World_STAGE_A_PT_3_INDEX_1_SIZE_2B_RES_480_FPS16_QWEN_IMAGE['job']['name']}", + {"override /net": "cosmos_v1_14B"}, + "_self_", + ], + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/ablation_2B_0321_tokenizer_data/ablation_2B_0321_tokenizer_data_218pt1_wan2pt1tokenizer_resizeT20_Res480V250322-241219_14B_perframe_noise/checkpoints/iter_000110000", + load_training_state=False, + strict_resume=False, + ), + job=dict( + group=Text2World_STAGE_A_PT_3_INDEX_1_SIZE_2B_RES_480_FPS16_QWEN_IMAGE["job"]["group"], + name="Stage-a_pt_3-Index-4-Size-14B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise", + ), + model_parallel=dict( + context_parallel_size=4, + ), + model=dict( + config=dict( + state_t=20, + net=dict( + rope_t_extrapolation_ratio=20.0 / 24, + ), + resize_online=True, + fsdp_shard_size=32, + ), + ), + optimizer=dict( + lr=2 ** (-14.5), + weight_decay=0.2, + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[300_000], + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=20 // 4, + num_workers=4, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict(resolution="480"), + ), + ratio="${trainer.grad_accum_iter}", + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="480", + num_video_frames=100, + video_decoder_name="chunked_video_decoder_with_fixed_fps", + min_fps_thres=16, + max_fps_thres=45, + ), + ), + ratio="${trainer.grad_accum_iter}", + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +# Vid2vid config + +V2V_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE_JOINT_2FRAMECOND: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{Text2World_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE['job']['name']}", + {"override /conditioner": "video_prediction_conditioner"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + "video2world_val_sampling_image2video_vbench", + "video2world_val_sampling_image2video_sora", + ] + }, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-a_pt_3-Video2World-Index-4-Size-14B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond", + ), + trainer=dict( + max_iter=200_000, + grad_accum_iter=2, + callbacks=dict( + reg_model_image2video_sora_val_sampling=dict( + every_n=5_000, + use_negative_prompt=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_sora_val_sampling=dict( + every_n=5_000, + use_negative_prompt=True, + latent_video_length="${model.config.state_t}", + ), + reg_model_image2video_vbench_val_sampling=dict( + every_n=5_000, + use_negative_prompt=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_vbench_val_sampling=dict( + every_n=5_000, + use_negative_prompt=True, + latent_video_length="${model.config.state_t}", + ), + ), + ), + model=dict( + config=dict( + min_num_conditional_frames=1, # choose either 1 (img2vid) or 2 (video2world) latent frames + max_num_conditional_frames=2, + ) + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs/Stage-a_pt_3-Index-4-Size-14B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise/checkpoints/iter_000035000", + ), + ), + flags={"allow_objects": True}, +) + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-a_pt_3-Video2World-Index-5-Size-14B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond +""" +V2V_STAGE_A_PT_3_INDEX_5_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE_JOINT_2FRAMECOND: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{V2V_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE_JOINT_2FRAMECOND['job']['name']}", + {"override /conditioner": "video_prediction_conditioner_v2"}, + "_self_", + ], + job=dict( + group=V2V_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE_JOINT_2FRAMECOND["job"]["group"], + name="Stage-a_pt_3-Video2World-Index-5-Size-14B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond", + ), + trainer=dict( + max_iter=200_000, + grad_accum_iter=1, + ), + ) +) + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-a_pt_3-Video2World-Index-1-Size-2B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond +""" +V2V_STAGE_A_PT_3_INDEX_1_SIZE_2B_RES_480_FPS_16_IMAGE_JOINT_2FRAMECOND: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{V2V_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE_JOINT_2FRAMECOND['job']['name']}", + {"override /net": "cosmos_v1_2B"}, + "_self_", + ], + job=dict( + group=V2V_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE_JOINT_2FRAMECOND["job"]["group"], + name="Stage-a_pt_3-Video2World-Index-1-Size-2B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond", + ), + optimizer=dict( + lr=2 ** (-14), # 2**(-14) = 6.103515625e-05 + weight_decay=0.1, + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs/Stage-a_pt_3-Index-1-Size-2B-Res-480-Fps-16-Note-qwen_imagecaption/checkpoints/iter_000057500", + load_training_state=False, + ), + model_parallel=dict( + context_parallel_size=1, + ), + model=dict( + config=dict( + loss_scale=10.0, + adjust_video_noise=True, + scaling="rectified_flow", + sigma_data=1.0, + fsdp_shard_size=8, + resolution="480", + state_t=24, + resize_online=True, + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=1.0, + ), + ) + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=24, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="480", + ), + ), + ratio="${trainer.grad_accum_iter}", + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.1), + dataset=dict( + resolution="480", + num_video_frames=93, + video_decoder_name="chunked_video_decoder_with_fixed_fps", + min_fps_thres=16, + max_fps_thres=45, + ), + ), + ratio="${trainer.grad_accum_iter}", + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + +V2V_STAGE_A_PT_3_INDEX_2_SIZE_2B_RES_480_FPS_16_IMAGE_JOINT_2FRAMECOND_GRADACCM_1: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{V2V_STAGE_A_PT_3_INDEX_1_SIZE_2B_RES_480_FPS_16_IMAGE_JOINT_2FRAMECOND['job']['name']}", + "_self_", + ], + job=dict( + group=V2V_STAGE_A_PT_3_INDEX_1_SIZE_2B_RES_480_FPS_16_IMAGE_JOINT_2FRAMECOND["job"]["group"], + name="Stage-a_pt_3-Video2World-Index-2-Size-2B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond_gradaccm_1", + ), + trainer=dict( + grad_accum_iter=1, + ), + ) +) + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + Text2World_STAGE_A_PT_3_INDEX_1_SIZE_2B_RES_480_FPS16_QWEN_IMAGE, + *build_debug_runs(Text2World_STAGE_A_PT_3_INDEX_1_SIZE_2B_RES_480_FPS16_QWEN_IMAGE), + ], + [ + Text2World_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE, + *build_debug_runs(Text2World_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE), + ], + [ + V2V_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE_JOINT_2FRAMECOND, + *build_debug_runs(V2V_STAGE_A_PT_3_INDEX_4_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE_JOINT_2FRAMECOND), + ], + [ + V2V_STAGE_A_PT_3_INDEX_5_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE_JOINT_2FRAMECOND, + *build_debug_runs(V2V_STAGE_A_PT_3_INDEX_5_SIZE_14B_RES_480_FPS_16_QWEN_IMAGE_JOINT_2FRAMECOND), + ], + [ + V2V_STAGE_A_PT_3_INDEX_1_SIZE_2B_RES_480_FPS_16_IMAGE_JOINT_2FRAMECOND, + *build_debug_runs(V2V_STAGE_A_PT_3_INDEX_1_SIZE_2B_RES_480_FPS_16_IMAGE_JOINT_2FRAMECOND), + ], + [ + V2V_STAGE_A_PT_3_INDEX_2_SIZE_2B_RES_480_FPS_16_IMAGE_JOINT_2FRAMECOND_GRADACCM_1, + *build_debug_runs(V2V_STAGE_A_PT_3_INDEX_2_SIZE_2B_RES_480_FPS_16_IMAGE_JOINT_2FRAMECOND_GRADACCM_1), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/stage3_2B_ablation.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/stage3_2B_ablation.py new file mode 100644 index 0000000000000000000000000000000000000000..84e9886d2f5027fca038cc022f6a9210287c5646 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/stage3_2B_ablation.py @@ -0,0 +1,3235 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +import functools +import math + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import duplicate_batches, duplicate_batches_random +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=25, + logging_iter=2, + callbacks=dict( + every_n_sample_reg=dict( + every_n=12, + ), + every_n_sample_ema=dict( + every_n=12, + ), + reg_model_image2video_sora_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_sora_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + reg_model_image2video_vbench_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_vbench_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames +""" +I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY: LazyDict = LazyDict( + dict( + defaults=[ + {"override /data_train": "image_cosmos_pretrain_qwen_20250415_video_cosmos_pretrain_v1_3_20250426_s3"}, + {"override /model": "fsdp"}, + {"override /net": "cosmos_v1_2B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "fusedadamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + "video2world_val_sampling_image2video_vbench", + "video2world_val_sampling_image2video_sora", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-3-Size-2B-Res-480-Fps-16-Note-qwen_video_only_later_frames", + ), + optimizer=dict( + lr=2 ** (-14), # 2**(-14) = 6.103515625e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + min_num_conditional_frames=1, # choose either 1 (img2vid) or 2 (video2world) latent frames + max_num_conditional_frames=2, + loss_scale=10.0, + adjust_video_noise=True, + scaling="rectified_flow", + sigma_data=1.0, + fsdp_shard_size=8, + resolution="480", + state_t=24, + resize_online=True, + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=1.0, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.3, + ), + ), + ) + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs/Stage-a_pt_3-Index-1-Size-2B-Res-480-Fps-16-Note-qwen_imagecaption/checkpoints/iter_000057500", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=1, + ), + trainer=dict( + max_iter=150_000, + logging_iter=200, + callbacks=dict( + every_n_sample_reg=dict( + every_n=5_000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + every_n_sample_ema=dict( + every_n=5_000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + reg_model_image2video_sora_val_sampling=dict( + every_n=5_000, + use_negative_prompt=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_sora_val_sampling=dict( + every_n=5_000, + use_negative_prompt=True, + latent_video_length="${model.config.state_t}", + ), + reg_model_image2video_vbench_val_sampling=dict( + every_n=5_000, + use_negative_prompt=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_vbench_val_sampling=dict( + every_n=5_000, + use_negative_prompt=True, + latent_video_length="${model.config.state_t}", + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=24, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="${model.config.resolution}", + ), + ), + ratio=0, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=True, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.5), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + caption_type="i2w_qwen2p5_7b_later_frames", + ), + ), + ratio=1, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_4_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY_FULL_PRMOPT = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-4-Size-2B-Res-480-Fps-16-Note-qwen_video_only_full_frames", + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + augmentor_name="video_basic_augmentor_v2", + ), + ), + ), + ), + ), + ) +) + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-5-Size-2B-Res-480-Fps-16-Note-qwen_joint_full_frames +""" +I2V_STAGE_C_PT_4_INDEX_5_SIZE_2B_RES_480_FPS16_JOINT_FULL_PRMOPT = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-5-Size-2B-Res-480-Fps-16-Note-qwen_joint_full_frames", + ), + model=dict( + config=dict( + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.2, + ), + text=dict( + dropout_rate=0.2, + ), + ), + ) + ), + trainer=dict( + grad_accum_iter=1, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=24, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict(resolution="480"), + ), + ratio="${trainer.grad_accum_iter}", + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + num_workers=8, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + augmentor_name="video_basic_augmentor_v2", + ), + ), + ratio="${trainer.grad_accum_iter}", + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-6-Size-2B-Res-480-Fps-16-Note-qwen_joint_full_frames +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-6-Size-2B-Res-480-Fps-16-Note-qwen_joint_full_frames_mock_wo_resume +""" +I2V_STAGE_C_PT_4_INDEX_6_SIZE_2B_RES_480_FPS16_JOINT_FULL_PRMOPT = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-6-Size-2B-Res-480-Fps-16-Note-qwen_joint_full_frames", + ), + trainer=dict( + grad_accum_iter=1, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=24, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict(resolution="480"), + ), + ratio="${trainer.grad_accum_iter}", + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + num_workers=8, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + augmentor_name="video_basic_augmentor_v2", + ), + ), + ratio="${trainer.grad_accum_iter}", + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +# dryrun dataloader +PYTHONPATH=$(pwd) torchrun --nproc_per_node=2 --master_port=12341 projects/cosmos/diffusion/v2/scripts/dataloader_e2e_test_cli.py --niter 5 --dump_vis_data --dump_item --dump_meta --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-8-Size-2B-Res-480p-Fps-16-Note-new_data + +# run it locally with mock data +torchrun --nproc_per_node=8 --master_port=12342 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-8-Size-2B-Res-480p-Fps-16-Note-new_data_mock_wo_resume" +""" +I2V_STAGE_C_PT_4_INDEX_8_SIZE_2B_RES_480p_FPS16_NEW_DATA = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250513_video_cosmos_pretrain_v2_and_high_quality_v0_uniform_dist_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-8-Size-2B-Res-480p-Fps-16-Note-new_data", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + resolution="480p", + state_t=20, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + ), + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-6-Size-2B-Res-480-Fps-16-Note-qwen_joint_full_frames/checkpoints/iter_000102500", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=30_000, + logging_iter=200, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size="${model.config.state_t}", + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_9_SIZE_2B_RES_480p_FPS16_05_20_pretrain2pt2_robo_wan = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrain_v2_2_and_high_quality_v0_robotics_and_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-9-Size-2B-Res-480p-Fps-16-Note-05_20_pretrain2pt2_robo_wan", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + resolution="480p", + state_t=20, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-8-Size-2B-Res-480p-Fps-16-Note-new_data/checkpoints/iter_000026000", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size="${model.config.state_t}", + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +""" +# print config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-10-Size-2B-Res-480p-Fps-16-Note-05_22_pretrain2pt2_hq_wan" + +# debug: train with mock data and wo_resume +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-10-Size-2B-Res-480p-Fps-16-Note-05_22_pretrain2pt2_hq_wan_wo_resume" +""" +I2V_STAGE_C_PT_4_INDEX_10_SIZE_2B_RES_480p_FPS16_05_22_pretrain2pt2_hq_wan = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrain_v2_2_0522_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-10-Size-2B-Res-480p-Fps-16-Note-05_22_pretrain2pt2_hq_wan", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + resolution="480p", + state_t=20, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + net=dict( + sac_config=dict( + mode="mm_only", + ) + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-9-Size-2B-Res-480p-Fps-16-Note-05_20_pretrain2pt2_robo_wan/checkpoints/iter_000065000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size="${model.config.state_t}", + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +# print config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-11-Size-2B-Res-720-Fps-16-Note-05_22_accumulated_hq_wan_wo_resume" + +# debug: train with mock data and wo_resume +NVTE_FUSED_ATTN=0 torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-11-Size-2B-Res-720-Fps-16-Note-05_22_accumulated_hq_wan_mock_wo_resume" ckpt_type=dummy +""" +I2V_STAGE_C_PT_4_INDEX_11_SIZE_2B_RES_720_FPS16_05_22_accumulated_hq_wan = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250522_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-11-Size-2B-Res-720-Fps-16-Note-05_22_accumulated_hq_wan", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + resolution="720", + state_t=20, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=1.0, + ), + ), + ), + model_parallel=dict( + context_parallel_size=2, + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-10-Size-2B-Res-480p-Fps-16-Note-05_22_pretrain2pt2_hq_wan/checkpoints/iter_000035000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=10, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_12_SIZE_2B_RES_720_FPS16_05_24_accumulated_hq_wan = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250524_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-12-Size-2B-Res-720-Fps-16-Note-05_24_accumulated_hq_wan", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + resolution="720", + state_t=20, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=1.0, + ), + ), + ), + model_parallel=dict( + context_parallel_size=2, + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-11-Size-2B-Res-720-Fps-16-Note-05_22_accumulated_hq_wan/checkpoints/iter_000042500", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=10, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_14_SIZE_2B_RES_720_FPS16_05_27_accumulated_hq_wan = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250527_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-14-Size-2B-Res-720-Fps-16-Note-05_27_accumulated_hq_wan", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.8], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + resolution="720", + state_t=20, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=1.0, + ), + ), + ), + model_parallel=dict( + context_parallel_size=2, + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-12-Size-2B-Res-720-Fps-16-Note-05_24_accumulated_hq_wan/checkpoints/iter_000012500", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=10, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +""" +Sparse Attn + +# dryrun dataloader +PYTHONPATH=$(pwd) torchrun \ + --nproc_per_node=2 \ + --master_port=12341 \ + projects/cosmos/diffusion/v2/scripts/dataloader_e2e_test_cli.py \ + --niter 5 \ + --dump_vis_data \ + --dump_item --dump_meta \ + --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- \ + experiment=Stage-c_pt_4-Index-38-Size-2B-Res-480p-Fps-16-Note-05_20_pretrain2pt2_robo_wan-sparse_attn-7dense-adaptive-Tx12x24-s1x4x8 + +# run it locally with mock data +PYTHONPATH=$(pwd) torchrun \ + --nproc_per_node=8 \ + --master_port=12342 \ + -m scripts.train \ + --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- \ + experiment="Stage-c_pt_4-Index-38-Size-2B-Res-480p-Fps-16-Note-05_20_pretrain2pt2_robo_wan-sparse_attn-7dense-adaptive-Tx12x24-s1x4x8" +""" + + +I2V_STAGE_C_PT_4_INDEX_38_SIZE_2B_RES_480p_FPS16_05_20_pretrain2pt2_robo_wan_SPARSE_ATTN_7DENSE_ADAPTIVE_Tx12x24_s1x4x8 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrain_v2_2_and_high_quality_v0_robotics_and_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-38-Size-2B-Res-480p-Fps-16-Note-05_20_pretrain2pt2_robo_wan-sparse_attn-7dense-adaptive-Tx12x24-s1x4x8", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.99], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + resolution="480p", + state_t=20, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + net=dict( + n_dense_blocks=7, + natten_parameters={"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-8-Size-2B-Res-480p-Fps-16-Note-new_data/checkpoints/iter_000026000", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size="${model.config.state_t}", + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +# locally run +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-15-Size-2B-Res-256-Fps-16-Note-05_27_accumulated_hq_wan_mock_wo_resume" ckpt_type=dummy +# dryrun dataloader +PYTHONPATH=$(pwd) torchrun --nproc_per_node=2 --master_port=12341 projects/cosmos/diffusion/v2/scripts/dataloader_e2e_test_cli.py --niter 5 --dump_vis_data --dump_item --dump_meta --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-15-Size-2B-Res-256-Fps-16-Note-05_27_accumulated_hq_wan dataloader_train.dataloaders.image_data.ratio=0 +""" +I2V_STAGE_C_PT_4_INDEX_15_SIZE_2B_RES_256_FPS16_05_27_accumulated_hq_wan = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250527_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-15-Size-2B-Res-256-Fps-16-Note-05_27_accumulated_hq_wan", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[1.0], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + resolution="256", + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="mm_only", + ), + rope_h_extrapolation_ratio=1.0, + rope_w_extrapolation_ratio=1.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + model_parallel=dict( + context_parallel_size=1, + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-14-Size-2B-Res-720-Fps-16-Note-05_27_accumulated_hq_wan/checkpoints/iter_000032500", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=96, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=5, + use_cache=True, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +# locally run + mock data + no resume +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-16-Size-2B-Res-480p-Fps-16-Note-05_28_accumulated_hq_wan_mock_wo_resume ckpt_type=dummy +""" +I2V_STAGE_C_PT_4_INDEX_16_SIZE_2B_RES_480p_FPS16_05_28_accumulated_hq_wan = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250528_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-16-Size-2B-Res-480p-Fps-16-Note-05_28_accumulated_hq_wan", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[1.0], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + resolution="480p", + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="mm_only", + ), + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + model_parallel=dict( + context_parallel_size=1, + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-15-Size-2B-Res-256-Fps-16-Note-05_27_accumulated_hq_wan/checkpoints/iter_000052500", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=24, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_17_SIZE_2B_RES_480p_FPS16_06_02_accumulated_hq_wan = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250602_dedup_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-17-Size-2B-Res-480p-Fps-16-Note-06_02_accumulated_hq_wan", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[1.0], + f_min=[0.4], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model=dict( + config=dict( + resolution="480p", + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="mm_only", + ), + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + model_parallel=dict( + context_parallel_size=1, + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-16-Size-2B-Res-480p-Fps-16-Note-05_28_accumulated_hq_wan/checkpoints/iter_000087500", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=24, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +# print config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-18-Size-2B-Res-720-Fps-16-Note-06_02_accumulated_hq_wan + +test locally +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-18-Size-2B-Res-720-Fps-16-Note-06_02_accumulated_hq_wan_mock_wo_resume ckpt_type=dummy +""" +I2V_STAGE_C_PT_4_INDEX_18_SIZE_2B_RES_720_FPS16_06_02_accumulated_hq_wan = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250602_dedup_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-18-Size-2B-Res-720-Fps-16-Note-06_02_accumulated_hq_wan", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-17-Size-2B-Res-480p-Fps-16-Note-06_02_accumulated_hq_wan", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_19_SIZE_2B_RES_720_FPS16_06_02_accumulated_hq_wan_from_17_tune_sigma_extrahigh = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250602_dedup_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-19-Size-2B-Res-720-Fps-16-Note-06_02_accumulated_hq_wan_from_17_tune_sigma_extrahigh", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.UNIFORM80_2000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-17-Size-2B-Res-480p-Fps-16-Note-06_02_accumulated_hq_wan", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +# run local debug +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19_mock_wo_resume" ckpt_type=dummy model.config.net.num_blocks=2 +""" +I2V_STAGE_C_PT_4_INDEX_20_SIZE_2B_RES_720_FPS16_06_04_accumulated_hq_wan_from_19 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250604_dedup_accumulated_and_high_quality_v2_wan_synthetic_v0_202505_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.08, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=1.0, + p_std=1.5, + sigma_max=500, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-19-Size-2B-Res-720-Fps-16-Note-06_02_accumulated_hq_wan_from_17_tune_sigma_extrahigh", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_21_SIZE_2B_RES_720_FPS16_HQ_V2_from_20 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v2_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-21-Size-2B-Res-720-Fps-16-Note-HQ_V2_from_20", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_22_SIZE_2B_RES_720_FPS16_HQ_V3_from_20 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-22-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_20", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[400_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-2B-Res-720-Fps-16-Note-06_04_accumulated_hq_wan_from_19/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_23_SIZE_2B_RES_720_FPS16_HQ_V3_1080_from_22 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-23-Size-2B-Res-1080-Fps-16-Note-HQ_V3_1080_from_22", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-22-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_20/checkpoints/iter_000020000", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt1080p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_24_SIZE_2B_RES_720_FPS16_HQ_V4_1080_from_22 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v4_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-24-Size-2B-Res-1080-Fps-16-Note-HQ_V4_1080_from_22", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-22-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_20/checkpoints/iter_000020000", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt1080p", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_25_SIZE_2B_RES_720_FPS16_HQ_V5_from_22 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-25-Size-2B-Res-720-Fps-16-Note-HQ_V5_from_22", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-22-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_20/checkpoints/iter_000020000", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=26_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v6_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-22-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_20/checkpoints/iter_000020000", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=26_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_27_SIZE_2B_RES_720_FPS16_HQ_V6_FIX_DATA_from_22 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v6_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-27-Size-2B-Res-720-Fps-16-Note-HQ_V6_FIX_DATA_from_22", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.01], + warm_up_steps=[2_000], + cycle_lengths=[30_001], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-22-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_20/checkpoints/iter_000020000", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + max_iter=30_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_100_SIZE_2B_RES_720_FPS10_HQ_V5_from_26 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-100-Size-2B-Res-720-Fps-10-Note-HQ_V5_from_26", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=16, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=16.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22/checkpoints/iter_000026000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=26_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=61, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_101_SIZE_2B_RES_480_FPS10_HQ_V5_from_26 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-101-Size-2B-Res-480-Fps-10-Note-HQ_V5_from_26", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="480", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=16, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=16.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22/checkpoints/iter_000026000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=26_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=61, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) +I2V_STAGE_C_PT_4_INDEX_102_SIZE_2B_RES_480_FPS16_HQ_V5_from_26 = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY["job"]["group"], + name="Stage-c_pt_4-Index-102-Size-2B-Res-480-Fps-16-Note-HQ_V5_from_26", + ), + optimizer=dict( + lr=2 ** (-14.5), # 2**(-14.5) = 3.0517578125e-05 + weight_decay=0.1, + ), + scheduler=dict( + f_max=[0.6], + f_min=[0.3], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + config=dict( + resolution="480", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + text_encoder_class="T5", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_2b_720", + ), + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22/checkpoints/iter_000026000", + load_training_state=False, + strict_resume=False, + ), + trainer=dict( + max_iter=26_000, + logging_iter=200, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=12, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.8), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="all", + num_video_frames=93, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +cs = ConfigStore.instance() + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY), + ], + [ + I2V_STAGE_C_PT_4_INDEX_4_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY_FULL_PRMOPT, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_4_SIZE_2B_RES_480_FPS16_QWEN_VIDEO_ONLY_FULL_PRMOPT), + ], + [ + I2V_STAGE_C_PT_4_INDEX_5_SIZE_2B_RES_480_FPS16_JOINT_FULL_PRMOPT, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_5_SIZE_2B_RES_480_FPS16_JOINT_FULL_PRMOPT), + ], + [ + I2V_STAGE_C_PT_4_INDEX_6_SIZE_2B_RES_480_FPS16_JOINT_FULL_PRMOPT, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_6_SIZE_2B_RES_480_FPS16_JOINT_FULL_PRMOPT), + ], + [ + I2V_STAGE_C_PT_4_INDEX_8_SIZE_2B_RES_480p_FPS16_NEW_DATA, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_8_SIZE_2B_RES_480p_FPS16_NEW_DATA), + ], + [ + I2V_STAGE_C_PT_4_INDEX_9_SIZE_2B_RES_480p_FPS16_05_20_pretrain2pt2_robo_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_9_SIZE_2B_RES_480p_FPS16_05_20_pretrain2pt2_robo_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_10_SIZE_2B_RES_480p_FPS16_05_22_pretrain2pt2_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_10_SIZE_2B_RES_480p_FPS16_05_22_pretrain2pt2_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_11_SIZE_2B_RES_720_FPS16_05_22_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_11_SIZE_2B_RES_720_FPS16_05_22_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_12_SIZE_2B_RES_720_FPS16_05_24_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_12_SIZE_2B_RES_720_FPS16_05_24_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_14_SIZE_2B_RES_720_FPS16_05_27_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_14_SIZE_2B_RES_720_FPS16_05_27_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_38_SIZE_2B_RES_480p_FPS16_05_20_pretrain2pt2_robo_wan_SPARSE_ATTN_7DENSE_ADAPTIVE_Tx12x24_s1x4x8, + *build_debug_runs( + I2V_STAGE_C_PT_4_INDEX_38_SIZE_2B_RES_480p_FPS16_05_20_pretrain2pt2_robo_wan_SPARSE_ATTN_7DENSE_ADAPTIVE_Tx12x24_s1x4x8 + ), + ], + [ + I2V_STAGE_C_PT_4_INDEX_15_SIZE_2B_RES_256_FPS16_05_27_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_15_SIZE_2B_RES_256_FPS16_05_27_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_16_SIZE_2B_RES_480p_FPS16_05_28_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_16_SIZE_2B_RES_480p_FPS16_05_28_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_17_SIZE_2B_RES_480p_FPS16_06_02_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_17_SIZE_2B_RES_480p_FPS16_06_02_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_18_SIZE_2B_RES_720_FPS16_06_02_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_18_SIZE_2B_RES_720_FPS16_06_02_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_19_SIZE_2B_RES_720_FPS16_06_02_accumulated_hq_wan_from_17_tune_sigma_extrahigh, + *build_debug_runs( + I2V_STAGE_C_PT_4_INDEX_19_SIZE_2B_RES_720_FPS16_06_02_accumulated_hq_wan_from_17_tune_sigma_extrahigh + ), + ], + [ + I2V_STAGE_C_PT_4_INDEX_20_SIZE_2B_RES_720_FPS16_06_04_accumulated_hq_wan_from_19, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_20_SIZE_2B_RES_720_FPS16_06_04_accumulated_hq_wan_from_19), + ], + [ + I2V_STAGE_C_PT_4_INDEX_21_SIZE_2B_RES_720_FPS16_HQ_V2_from_20, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_21_SIZE_2B_RES_720_FPS16_HQ_V2_from_20), + ], + [ + I2V_STAGE_C_PT_4_INDEX_22_SIZE_2B_RES_720_FPS16_HQ_V3_from_20, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_22_SIZE_2B_RES_720_FPS16_HQ_V3_from_20), + ], + [ + I2V_STAGE_C_PT_4_INDEX_23_SIZE_2B_RES_720_FPS16_HQ_V3_1080_from_22, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_23_SIZE_2B_RES_720_FPS16_HQ_V3_1080_from_22), + ], + [ + I2V_STAGE_C_PT_4_INDEX_24_SIZE_2B_RES_720_FPS16_HQ_V4_1080_from_22, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_24_SIZE_2B_RES_720_FPS16_HQ_V4_1080_from_22), + ], + [ + I2V_STAGE_C_PT_4_INDEX_25_SIZE_2B_RES_720_FPS16_HQ_V5_from_22, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_25_SIZE_2B_RES_720_FPS16_HQ_V5_from_22), + ], + [ + I2V_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22), + ], + [ + I2V_STAGE_C_PT_4_INDEX_27_SIZE_2B_RES_720_FPS16_HQ_V6_FIX_DATA_from_22, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_27_SIZE_2B_RES_720_FPS16_HQ_V6_FIX_DATA_from_22), + ], + # variants + [ + I2V_STAGE_C_PT_4_INDEX_100_SIZE_2B_RES_720_FPS10_HQ_V5_from_26, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_100_SIZE_2B_RES_720_FPS10_HQ_V5_from_26), + ], + [ + I2V_STAGE_C_PT_4_INDEX_101_SIZE_2B_RES_480_FPS10_HQ_V5_from_26, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_101_SIZE_2B_RES_480_FPS10_HQ_V5_from_26), + ], + [ + I2V_STAGE_C_PT_4_INDEX_102_SIZE_2B_RES_480_FPS16_HQ_V5_from_26, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_102_SIZE_2B_RES_480_FPS16_HQ_V5_from_26), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/state3_14B_index_3.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/state3_14B_index_3.py new file mode 100644 index 0000000000000000000000000000000000000000..32774420a9850d0bb9f80602bebe3faed5ac3f7b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/resume_text2world/state3_14B_index_3.py @@ -0,0 +1,4326 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Configs for resuming from stage3 training + +import functools +import math + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import duplicate_batches, duplicate_batches_random +from cosmos_policy._src.predict2.models.video2world_model import HighSigmaStrategy + +_TRAINER_DEBUG_CONFIG = dict( + max_iter=25, + logging_iter=2, + callbacks=dict( + every_n_sample_reg=dict( + every_n=12, + ), + every_n_sample_ema=dict( + every_n=12, + ), + reg_model_image2video_sora_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_sora_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + reg_model_image2video_vbench_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ema_model_image2video_vbench_val_sampling=dict( + every_n=13, + is_debug=True, + latent_video_length="${model.config.state_t}", + ), + ), +) +_CKPT_DEBUG_CONFIG = dict( + save_iter=10, + load_path="", + load_training_state=False, + strict_resume=False, +) + + +def build_debug_runs(job): + wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + mock_wo_resume = dict( + defaults=[ + f"/experiment/{job['job']['name']}", + {"override /data_train": "mock"}, + "_self_", + ], + job=dict( + group=job["job"]["group"] + "_debug", + name=f"{job['job']['name']}_MOCK_WO_RESUME" + "_${now:%Y-%m-%d}_${now:%H-%M-%S}", + ), + trainer=_TRAINER_DEBUG_CONFIG, + checkpoint=_CKPT_DEBUG_CONFIG, + ) + + return [wo_resume, mock_wo_resume] + + +""" +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor + +# diff with resumed config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-a_pt_3-Video2World-Index-5-Size-14B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond +""" +I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR: LazyDict = LazyDict( + dict( + defaults=[ + {"override /data_train": "image_cosmos_pretrain_qwen_20250415_video_cosmos_pretrain_v1_3_20250426_s3"}, + {"override /model": "fsdp"}, + {"override /net": "cosmos_v1_14B"}, + {"override /conditioner": "video_prediction_conditioner"}, + {"override /ckpt_type": "dcp"}, + {"override /optimizer": "fusedadamw"}, + { + "override /callbacks": [ + "basic", + "viz_online_sampling", + "wandb", + "cluster_speed", + ] + }, + {"override /checkpoint": "s3"}, + {"override /tokenizer": "wan2pt1_tokenizer"}, + "_self_", + ], + job=dict( + group="official_runs_video2world", + name="Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor", + ), + optimizer=dict( + lr=2 ** (-14.5), + weight_decay=0.2, + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[300_000], + ), + model=dict( + config=dict( + min_num_conditional_frames=1, # choose either 1 (img2vid) or 2 (video2world) latent frames + max_num_conditional_frames=2, + loss_scale=10.0, + adjust_video_noise=True, + scaling="rectified_flow", + sigma_data=1.0, + fsdp_shard_size=32, + resolution="480", + state_t=20, + resize_online=True, + net=dict( + rope_enable_fps_modulation=False, + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=20.0 / 24, + ), + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + ) + ), + checkpoint=dict( + save_iter=2_500, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-a_pt_3-Video2World-Index-5-Size-14B-Res-480-Fps-16-Note-qwen_imagecaption_sync_noise_joint_2framecond/checkpoints/iter_000052500", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=4, + ), + trainer=dict( + max_iter=150_000, + logging_iter=200, + callbacks=dict( + every_n_sample_reg=dict( + every_n=5_000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + every_n_sample_ema=dict( + every_n=5_000, + do_x0_prediction=False, + guidance=[0, 3, 7], + fps=16, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=20 // 4, + num_workers=6, + use_cache=False, + cache_size=8, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches, n=1), + dataset=dict( + resolution="480", + ), + ), + ratio="${trainer.grad_accum_iter}", + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1), + dataset=dict( + resolution="480", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + ), + ), + ratio="${trainer.grad_accum_iter}", + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_6_SIZE_14B_RES_480_SHORT_DURATION_64N: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-6-Size-14B-Res-480-Fps-16-Note-video_data1pt3_short_duration_64n", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor/checkpoints/iter_000060000", + load_training_state=False, + strict_resume=False, + ), + scheduler=dict( + f_max=[0.2], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[50_000], + ), + trainer=dict( + max_iter=20_000, + logging_iter=200, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + augmentor_name="video_basic_augmentor_v3_full_frames", + ), + ), + ), + ), + ), + ) +) + + +I2V_STAGE_C_PT_4_INDEX_10_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_CONTINUE_64N: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-10-Size-14B-Res-480-Fps-16-Note-video_data1pt3_continue_64n", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor/checkpoints/iter_000060000", + load_training_state=False, + strict_resume=False, + ), + scheduler=dict( + f_max=[0.2], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[50_000], + ), + trainer=dict( + max_iter=20_000, + logging_iter=200, + ), + ) +) + + +I2V_STAGE_C_PT_4_INDEX_11_SIZE_14B_RES_480_VIDEO202505_64N: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + {"override /data_train": "image_cosmos_pretrain_qwen_20250415_video_cosmos_pretrainvideo_202505_s3"}, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-11-Size-14B-Res-480-Fps-16-Note-video202505_64n", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-3-Size-14B-Res-480-Fps-16-Note-video_data1pt3_augmentor/checkpoints/iter_000060000", + load_training_state=False, + strict_resume=False, + ), + scheduler=dict( + f_max=[0.2], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[50_000], + ), + trainer=dict( + max_iter=20_000, + logging_iter=200, + ), + ) +) + +""" +# dryrun dataloader +PYTHONPATH=$(pwd) torchrun --nproc_per_node=8 --master_port=12341 projects/cosmos/diffusion/v2/scripts/dataloader_e2e_test_cli.py --niter 5 --dump_vis_data --dump_item --dump_meta --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-20-Size-14B-Res-480-Fps-16-Note-05_20_pretrain2pt2_robo_wan +""" +I2V_STAGE_C_PT_4_INDEX_20_SIZE_14B_RES_480_FPS16_05_20_pretrain2pt2_robo_wan: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrain_v2_2_and_high_quality_v0_robotics_and_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-20-Size-14B-Res-480-Fps-16-Note-05_20_pretrain2pt2_robo_wan", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-10-Size-14B-Res-480-Fps-16-Note-video_data1pt3_continue_64n/checkpoints/iter_000020000", + load_training_state=False, + strict_resume=False, + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[50_000], + ), + trainer=dict( + max_iter=20_000, + logging_iter=200, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ) +) + +I2V_STAGE_C_PT_4_INDEX_21_SIZE_14B_RES_480p_FPS16_05_22_pretrain2pt2_hq_wan: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrain_v2_2_0522_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-21-Size-14B-Res-480p-Fps-16-Note-05_22_pretrain2pt2_hq_wan", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-20-Size-14B-Res-480-Fps-16-Note-05_20_pretrain2pt2_robo_wan/checkpoints/iter_000020000", + load_training_state=True, + strict_resume=False, + ), + model=dict( + config=dict( + resolution="480p", + state_t=20, + resize_online=True, + net=dict( + sac_config=dict( + mode="mm_only", + ) + ), + ), + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[50_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=200, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ) +) + +""" +# dryrun config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-22-Size-14B-Res-720-Fps-16-Note-T24_05_22_accumulated_hq_wan +""" +I2V_STAGE_C_PT_4_INDEX_22_SIZE_14B_RES_720_FPS16_T24_05_22_accumulated_hq_wan: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250522_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-22-Size-14B-Res-720-Fps-16-Note-T24_05_22_accumulated_hq_wan", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-21-Size-14B-Res-480p-Fps-16-Note-05_22_pretrain2pt2_hq_wan/checkpoints/iter_000050000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_14b_720", + ) + ), + ), + ), + scheduler=dict( + f_max=[0.2], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[50_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ) +) + +I2V_STAGE_C_PT_4_INDEX_23_SIZE_14B_RES_720_FPS16_T24_05_22_accumulated_hq_wan_rerun: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_22_SIZE_14B_RES_720_FPS16_T24_05_22_accumulated_hq_wan['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_22_SIZE_14B_RES_720_FPS16_T24_05_22_accumulated_hq_wan["job"]["group"], + name="Stage-c_pt_4-Index-23-Size-14B-Res-720-Fps-16-Note-T24_05_22_accumulated_hq_wan_rerun", + ), + checkpoint=dict( + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-22-Size-14B-Res-720-Fps-16-Note-T24_05_22_accumulated_hq_wan/checkpoints/iter_000008000", + load_training_state=True, + strict_resume=False, + ), + trainer=dict( + straggler_detection=dict( + enabled=False, + max_diff=1.5, + ) + ), + ) +) + +I2V_STAGE_C_PT_4_INDEX_24_SIZE_14B_RES_720_FPS16_T24_05_27_accumulated_hq_wan: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250527_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-24-Size-14B-Res-720-Fps-16-Note-T24_05_27_accumulated_hq_wan", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-23-Size-14B-Res-720-Fps-16-Note-T24_05_22_accumulated_hq_wan_rerun/checkpoints/iter_000019000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + net=dict( + sac_config=dict( + mode="predict2_14b_720", + ) + ), + ), + ), + scheduler=dict( + f_max=[0.2], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[50_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ) +) + +""" +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-25-Size-14B-Res-256-Fps-16-Note-T24_05_27_accumulated_hq_wan_mock_wo_resume" ckpt_type=dummy +""" +I2V_STAGE_C_PT_4_INDEX_25_SIZE_14B_RES_256_FPS16_T24_05_27_accumulated_hq_wan: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250527_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-25-Size-14B-Res-256-Fps-16-Note-T24_05_27_accumulated_hq_wan", + ), + checkpoint=dict( + save_iter=5_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-24-Size-14B-Res-720-Fps-16-Note-T24_05_27_accumulated_hq_wan/checkpoints/iter_000011000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=1, + ), + model=dict( + config=dict( + resolution="256", + denoise_replace_gt_frames=False, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + net=dict( + rope_h_extrapolation_ratio=1.0, + rope_w_extrapolation_ratio=1.0, + rope_t_extrapolation_ratio=1.0, + sac_config=dict( + mode="mm_only", + ), + ), + ), + ), + scheduler=dict( + f_max=[1.0], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[100_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=24, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + cache_size=16, + concat_size=1, + cache_augment_fn=functools.partial(duplicate_batches_random, n=1.2), + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-26-Size-14B-Res-256-Fps-16-Note-T24_05_27_accumulated_hq_wan_rerun_mock_wo_resume" ckpt_type=dummy +""" +I2V_STAGE_C_PT_4_INDEX_26_SIZE_14B_RES_256_FPS16_T24_05_27_accumulated_hq_wan_rerun: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_25_SIZE_14B_RES_256_FPS16_T24_05_27_accumulated_hq_wan['job']['name']}", + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_25_SIZE_14B_RES_256_FPS16_T24_05_27_accumulated_hq_wan["job"]["group"], + name="Stage-c_pt_4-Index-26-Size-14B-Res-256-Fps-16-Note-T24_05_27_accumulated_hq_wan_rerun", + ), + model=dict( + config=dict( + denoise_replace_gt_frames=True, + ) + ), + checkpoint=dict( + save_iter=2_500, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-25-Size-14B-Res-256-Fps-16-Note-T24_05_27_accumulated_hq_wan_64N/checkpoints/iter_000005000", + load_training_state=True, + strict_resume=False, + ), + ) +) + +""" +# print config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-27-Size-14B-Res-480p-Fps-16-Note-T24_05_28_accumulated_hq_wan" + +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-27-Size-14B-Res-480p-Fps-16-Note-T24_05_28_accumulated_hq_wan_mock_wo_resume" ckpt_type=dummy +""" +I2V_STAGE_C_PT_4_INDEX_27_SIZE_14B_RES_480p_FPS16_T24_05_28_accumulated_hq_wan: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250528_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-27-Size-14B-Res-480p-Fps-16-Note-T24_05_28_accumulated_hq_wan", + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-26-Size-14B-Res-256-Fps-16-Note-T24_05_27_accumulated_hq_wan_rerun/checkpoints/iter_000047500", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=4, + ), + model=dict( + config=dict( + resolution="480p", + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + net=dict( + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="mm_only", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.4], # tune it on the fly. 1.0 is too agressive. loss is not stable. + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=6, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_28_SIZE_14B_RES_480p_FPS16_T24_05_28_accumulated_hq_wan_tune_pmean_from_index27_24k: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250528_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-28-Size-14B-Res-480p-Fps-16-Note-T24_05_28_accumulated_hq_wan_tune_pmean_from_index27_24k", + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-27-Size-14B-Res-480p-Fps-16-Note-T24_05_28_accumulated_hq_wan/checkpoints/iter_000024000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=4, + ), + model=dict( + config=dict( + resolution="480p", + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=0.7, + p_std=1.0, + ), + net=dict( + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="mm_only", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.4], # tune it on the fly. 1.0 is too agressive. loss is not stable. + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=6, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_29_SIZE_14B_RES_480p_FPS16_T24_06_01_accumulated_hq_wan: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250601_dedup_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-29-Size-14B-Res-480p-Fps-16-Note-T24_06_01_accumulated_hq_wan", + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-28-Size-14B-Res-480p-Fps-16-Note-T24_05_28_accumulated_hq_wan_tune_pmean_from_index27_24k/checkpoints/iter_000038000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=4, + ), + model=dict( + config=dict( + resolution="480p", + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=0.0, + p_std=1.0, + ), + net=dict( + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="mm_only", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.4], # tune it on the fly. 1.0 is too agressive. loss is not stable. + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=6, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_30_SIZE_14B_RES_480p_FPS16_T24_HQ_V0: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-30-Size-14B-Res-480p-Fps-16-Note-T24_HQ_V0_FROM_28", + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-28-Size-14B-Res-480p-Fps-16-Note-T24_05_28_accumulated_hq_wan_tune_pmean_from_index27_24k/checkpoints/iter_000038000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=4, + ), + model=dict( + config=dict( + resolution="480p", + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=0.0, + p_std=1.0, + ), + net=dict( + rope_h_extrapolation_ratio=2.0, + rope_w_extrapolation_ratio=2.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="mm_only", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.4], # tune it on the fly. 1.0 is too agressive. loss is not stable. + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=6, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_31_SIZE_14B_RES_720p_FPS16_T24_HQ_V0: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-31-Size-14B-Res-720p-Fps-16-Note-T24_HQ_V0_FROM_30", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-30-Size-14B-Res-480p-Fps-16-Note-T24_HQ_V0_FROM_28/checkpoints/iter_000008000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=0.0, + p_std=1.0, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_32_SIZE_14B_RES_480p_FPS16_T24_06_02_data_from_29: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250602_dedup_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-32-Size-14B-Res-480p-Fps-16-Note-T24_06_02_data_from_29", + ), + checkpoint=dict( + save_iter=2_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-29-Size-14B-Res-480p-Fps-16-Note-T24_06_01_accumulated_hq_wan", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=0.0, + p_std=1.0, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_33_SIZE_14B_RES_720_FPS16_T24_06_02_data_tune_sigma_extrahigh_from_32: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250602_dedup_accumulated_and_high_quality_v1_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-33-Size-14B-Res-720-Fps-16-Note-T24_06_02_data_tune_sigma_extrahigh_from_32", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-32-Size-14B-Res-480p-Fps-16-Note-T24_06_02_data_from_29", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.UNIFORM80_2000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=1.0, + p_std=1.0, + sigma_max=1000, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_34_SIZE_14B_RES_720p_FPS16_T24_HQ_V0_FROM_31: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-34-Size-14B-Res-720p-Fps-16-Note-T24_HQ_V0_FROM_31", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-31-Size-14B-Res-720p-Fps-16-Note-T24_HQ_V0_FROM_30", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + denoise_replace_gt_frames=True, + high_sigma_strategy=str(HighSigmaStrategy.UNIFORM80_2000), + high_sigma_ratio=0.05, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=0.0, + p_std=1.0, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +# run local debug +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-35-Size-14B-Res-720-Fps-16-Note-T24_06_04_data_tune_sigma_from_33_mock_wo_resume" ckpt_type=dummy model.config.net.num_blocks=2 +""" +I2V_STAGE_C_PT_4_INDEX_35_SIZE_14B_RES_720_FPS16_T24_06_04_data_tune_sigma_from_33: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250604_dedup_accumulated_and_high_quality_v2_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-35-Size-14B-Res-720-Fps-16-Note-T24_06_04_data_tune_sigma_from_33", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-33-Size-14B-Res-720-Fps-16-Note-T24_06_02_data_tune_sigma_extrahigh_from_32", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.08, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=1.0, + p_std=1.5, + sigma_max=500, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_36_SIZE_14B_RES_720_FPS16_T24_06_04_data_tune_sigma_from_33: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250604_dedup_accumulated_and_high_quality_v2_wan_synthetic_v0_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-36-Size-14B-Res-720-Fps-16-Note-T24_06_04_data_tune_sigma_from_33", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-33-Size-14B-Res-720-Fps-16-Note-T24_06_02_data_tune_sigma_extrahigh_from_32/checkpoints/iter_000010000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.20], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_37_SIZE_14B_RES_720_FPS16_T24_HQV2_from_35: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v2_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-37-Size-14B-Res-720-Fps-16-Note-T24_HQV2_from_35", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-35-Size-14B-Res-720-Fps-16-Note-T24_06_04_data_tune_sigma_from_33/checkpoints/iter_000024000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +# local debug +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-38-Size-14B-Res-720-Fps-16-Note-T24_HQV3_from_35_mock_wo_resume" ckpt_type=dummy model.config.net.num_blocks=2 +""" +I2V_STAGE_C_PT_4_INDEX_38_SIZE_14B_RES_720_FPS16_T24_HQV3_from_35: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-38-Size-14B-Res-720-Fps-16-Note-T24_HQV3_from_35", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-35-Size-14B-Res-720-Fps-16-Note-T24_06_04_data_tune_sigma_from_33/checkpoints/iter_000024000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_39_SIZE_14B_RES_720_FPS16_T24_HQV2pt1_from_37: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v2_1_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-39-Size-14B-Res-720-Fps-16-Note-T24_HQV2pt1_from_37", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-37-Size-14B-Res-720-Fps-16-Note-T24_HQV2_from_35", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +""" +# local debug +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38_mock_wo_resume" ckpt_type=dummy model.config.net.num_blocks=2 +""" +I2V_STAGE_C_PT_4_INDEX_40_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_from_38: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_1_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-38-Size-14B-Res-720-Fps-16-Note-T24_HQV3_from_35", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[40_000], + ), + trainer=dict( + max_iter=100_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_41_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_1080_from_40: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v3_1_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-41-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_1080_from_40", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38/checkpoints/iter_000015000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt1080p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_42_SIZE_14B_RES_720_FPS16_T24_HQV4_1080_from_40: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v4_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-42-Size-14B-Res-720-Fps-16-Note-T24_HQV4_1080_from_40", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38/checkpoints/iter_000015000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt1080p", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +""" +# print config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40 +""" +I2V_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38/checkpoints/iter_000015000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_44_SIZE_14B_RES_720_FPS16_T24_HQV6_from_40: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v6_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-44-Size-14B-Res-720-Fps-16-Note-T24_HQV6_from_40", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38/checkpoints/iter_000015000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +# print config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-45-Size-14B-Res-720-Fps-16-Note-T24_HQV2_1_from_35" +# dryrun dataloader +PYTHONPATH=$(pwd) torchrun --nproc_per_node=8 --master_port=12341 projects/cosmos/diffusion/v2/scripts/dataloader_e2e_test_cli.py --niter 5 --dump_vis_data --dump_item --dump_meta --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment=Stage-c_pt_4-Index-45-Size-14B-Res-720-Fps-16-Note-T24_HQV2_1_from_35 +""" +I2V_STAGE_C_PT_4_INDEX_45_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_35: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v2_1_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-45-Size-14B-Res-720-Fps-16-Note-T24_HQV2_1_from_35", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-37-Size-14B-Res-720-Fps-16-Note-T24_HQV2_from_35", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[30_001], + ), + trainer=dict( + max_iter=30_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_46_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_35: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v2_1_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-46-Size-14B-Res-720-Fps-16-Note-T24_HQV2_1_from_35", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-37-Size-14B-Res-720-Fps-16-Note-T24_HQV2_from_35", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[30_001], + ), + trainer=dict( + max_iter=30_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_47_SIZE_14B_RES_720_FPS16_T24_HQV5_FIX_DATA_from_40: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-47-Size-14B-Res-720-Fps-16-Note-T24_HQV5_FIX_DATA_from_40", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38/checkpoints/iter_000015000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.01], + warm_up_steps=[2_000], + cycle_lengths=[25_001], + ), + trainer=dict( + max_iter=25_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + use_native_fps=True, + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_48_SIZE_14B_RES_720_FPS16_T24_HQV5_VIDOE_ONLY_from_40: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-48-Size-14B-Res-720-Fps-16-Note-T24_HQV5_VIDOE_ONLY_from_40", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38/checkpoints/iter_000015000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=0, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_49_SIZE_14B_RES_720_FPS16_T24_HQV5_ROPE_TUNE_from_40: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-49-Size-14B-Res-720-Fps-16-Note-T24_HQV5_ROPE_TUNE_from_40", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38/checkpoints/iter_000015000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=1.0, + rope_w_extrapolation_ratio=1.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.01], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=20_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +""" +# print +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-50-Size-14B-Res-720-Fps-16-Note-T24_HQV5_SHIFT24_from_40" + +local train 14B with shift24 +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-50-Size-14B-Res-720-Fps-16-Note-T24_HQV5_SHIFT24_from_40_mock_wo_resume" model.config.net.num_blocks=10 +""" +I2V_STAGE_C_PT_4_INDEX_50_SIZE_14B_RES_720_FPS16_T24_HQV5_SHIFT24_from_40: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-50-Size-14B-Res-720-Fps-16-Note-T24_HQV5_SHIFT24_from_40", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-40-Size-14B-Res-720-Fps-16-Note-T24_HQV3pt1_from_38/checkpoints/iter_000015000", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.SHIFT24), + high_sigma_ratio=0.02, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.01], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=20_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +""" +test locally +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-51-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_from_43_mock_wo_resume" model.config.net.num_blocks=10 +""" +I2V_STAGE_C_PT_4_INDEX_51_SIZE_14B_RES_720_FPS16_T24_DATA_0612_from_43: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250612_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-51-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_from_43", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.5, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.01], + warm_up_steps=[2_000], + cycle_lengths=[50_001], + ), + trainer=dict( + max_iter=50_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=3, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +""" +# local run it +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-52-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_from_51_mock_wo_resume" model.config.net.num_blocks=10 +""" +I2V_STAGE_C_PT_4_INDEX_52_SIZE_14B_RES_720_FPS16_T24_DATA_0612_from_51: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250612_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-52-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_from_51", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-51-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_from_43", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.BALANCED_TWO_HEADS_V1), + high_sigma_ratio=0.05, + low_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.6, + sigma_max=120, + sigma_min=0.2, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.01], + warm_up_steps=[2_000], + cycle_lengths=[70_001], + ), + trainer=dict( + max_iter=70_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=3, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +""" +# print config +torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --dryrun --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-53-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_HARDCODED_20STEPS_from_51" + +# local run it +torchrun --nproc_per_node=4 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="Stage-c_pt_4-Index-53-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_HARDCODED_20STEPS_from_51_mock_wo_resume" model.config.net.num_blocks=10 model_parallel.context_parallel_size=4 ckpt_type=dummy +""" +I2V_STAGE_C_PT_4_INDEX_53_SIZE_14B_RES_720_FPS16_T24_DATA_0612_HARDCODED_20STEPS_from_51: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250612_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-53-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_HARDCODED_20STEPS_from_51", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-51-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_from_43", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.HARDCODED_20steps), + high_sigma_ratio=0.05, + low_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.6, + sigma_max=120, + sigma_min=0.2, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.01], + warm_up_steps=[2_000], + cycle_lengths=[70_001], + ), + trainer=dict( + max_iter=70_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + callbacks=dict( + every_n_sample_ema=dict( + num_sampling_step=20, + ), + every_n_sample_reg=dict( + num_sampling_step=20, + ), + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=3, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_54_SIZE_14B_RES_720_FPS16_T24_DATA_0612_from_52: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_pretrainvideo_20250612_dedup_accumulated_and_high_quality_v3_202505_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-54-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_from_52", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-52-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_from_51", + load_training_state=True, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.BALANCED_TWO_HEADS_V1), + high_sigma_ratio=0.05, + low_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.6, + sigma_max=120, + sigma_min=0.2, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.3], + f_min=[0.01], + warm_up_steps=[2_000], + cycle_lengths=[150_001], + ), + trainer=dict( + max_iter=150_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=3, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_55_SIZE_14B_RES_720_FPS16_T24_DATA_HQ_V5_from_54: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-55-Size-14B-Res-720-Fps-16-Note-T24_DATA_HQ_V5_from_54", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-54-Size-14B-Res-720-Fps-16-Note-T24_DATA_0612_from_52", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.BALANCED_TWO_HEADS_V1), + high_sigma_ratio=0.05, + low_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.6, + sigma_max=120, + sigma_min=0.2, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.3], + f_min=[0.01], + warm_up_steps=[1_000], + cycle_lengths=[30_001], + ), + trainer=dict( + max_iter=30_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="gt720p", + ), + ), + ratio=3, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + + +""" +torchrun --nproc_per_node=8 --master_port=12341 -m scripts.train --config=projects/cosmos/diffusion/v2/configs/video2world/config.py -- experiment="I2V_STAGE_C_PT_4_INDEX_00_SIZE_14B_RES_480_FPS16_PROFILING" +""" +I2V_STAGE_C_PT_4_INDEX_00_SIZE_14B_RES_480_FPS16_PROFILING: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "mock", + }, + {"override /ckpt_type": "dummy"}, + { + "override /callbacks": [ + "basic", + "wandb", + "cluster_speed", + ] + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="I2V_STAGE_C_PT_4_INDEX_00_SIZE_14B_RES_480_FPS16_PROFILING", + ), + checkpoint=dict( + save_iter=1_000, + load_path="", + ), + trainer=dict( + max_iter=2_00, + logging_iter=20, + ), + model_parallel=dict( + context_parallel_size=4, + ), + model=dict( + config=dict( + state_t=20, # 24 + fsdp_shard_size=32, + resolution="480", + ) + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=5, + dataset=dict( + t5_dim=1024, + resolution="${model.config.resolution}", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + dataset=dict( + t5_dim=1024, + resolution="${model.config.resolution}", + ), + ), + ratio=1, + ), + ), + ), + upload_reproducible_setup=True, + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_100_SIZE_14B_RES_720_FPS10_T15_HQV5_from_43: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, # @qinsheng + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-100-Size-14B-Res-720-Fps-10-Note-T15_HQV5_from_43", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40/checkpoints/iter_000018000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="720", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=16, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=16.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=61, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_101_SIZE_14B_RES_480_FPS10_T15_HQV5_from_43: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-101-Size-14B-Res-480-Fps-10-Note-T15_HQV5_from_43", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40/checkpoints/iter_000018000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="480", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=16, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=16.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=61, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) +I2V_STAGE_C_PT_4_INDEX_102_SIZE_14B_RES_480_FPS16_T24_HQV5_from_43: LazyDict = LazyDict( + dict( + defaults=[ + f"/experiment/{I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v5_20250607_s3" + }, + "_self_", + ], + job=dict( + group=I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR["job"]["group"], + name="Stage-c_pt_4-Index-102-Size-14B-Res-480-Fps-16-Note-T24_HQV5_from_43", + ), + checkpoint=dict( + save_iter=1_000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-T24_HQV5_from_40/checkpoints/iter_000018000", + load_training_state=False, + strict_resume=False, + ), + model_parallel=dict( + context_parallel_size=8, + ), + model=dict( + config=dict( + resolution="480", + high_sigma_strategy=str(HighSigmaStrategy.LOGUNIFORM200_100000), + high_sigma_ratio=0.05, + denoise_replace_gt_frames=True, + state_t=24, + resize_online=True, + tokenizer=dict( + temporal_window=16, + ), + sde=dict( + p_mean=math.log(4.0), + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + ), + net=dict( + rope_h_extrapolation_ratio=3.0, + rope_w_extrapolation_ratio=3.0, + rope_t_extrapolation_ratio=24.0 / 24, + sac_config=dict( + mode="predict2_14b_720", + ), + ), + ), + ), + scheduler=dict( + f_max=[0.25], + f_min=[0.1], + warm_up_steps=[2_000], + cycle_lengths=[20_001], + ), + trainer=dict( + max_iter=18_000, + logging_iter=100, + straggler_detection=dict( + enabled=True, + max_diff=1.5, + ), + ), + dataloader_train=dict( + dataloaders=dict( + image_data=dict( + dataloader=dict( + batch_size=3, + dataset=dict( + resolution="${model.config.resolution}", + dataset_resolution_type="gt720p", + caption_type="qwen2p5_7b_v4", + embedding_type="t5_xxl", + ), + ), + ratio=1, + ), + video_data=dict( + dataloader=dict( + batch_size=1, + use_cache=False, + dataset=dict( + resolution="${model.config.resolution}", + video_decoder_name="video_naive_bytes", + augmentor_name="video_basic_augmentor_v2", + embedding_type="t5_xxl", + max_fps_thres=60, + min_fps_thres=10, + caption_type="t2w_qwen2p5_7b", + num_video_frames=93, + dataset_resolution_type="all", + ), + ), + ratio=1, + ), + ), + ), + ), + flags={"allow_objects": True}, +) +cs = ConfigStore.instance() +cs.store( + group="experiment", + package="_global_", + name=f"{I2V_STAGE_C_PT_4_INDEX_00_SIZE_14B_RES_480_FPS16_PROFILING['job']['name']}", + node=I2V_STAGE_C_PT_4_INDEX_00_SIZE_14B_RES_480_FPS16_PROFILING, +) + +for _item, _item_wo_resume, _item_mock_wo_resume in [ + [ + I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_3_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_AUGMENTOR), + ], + [ + I2V_STAGE_C_PT_4_INDEX_6_SIZE_14B_RES_480_SHORT_DURATION_64N, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_6_SIZE_14B_RES_480_SHORT_DURATION_64N), + ], + [ + I2V_STAGE_C_PT_4_INDEX_10_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_CONTINUE_64N, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_10_SIZE_14B_RES_480_NEW_VIDEO_DATA1PT3_CONTINUE_64N), + ], + [ + I2V_STAGE_C_PT_4_INDEX_11_SIZE_14B_RES_480_VIDEO202505_64N, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_11_SIZE_14B_RES_480_VIDEO202505_64N), + ], + [ + I2V_STAGE_C_PT_4_INDEX_20_SIZE_14B_RES_480_FPS16_05_20_pretrain2pt2_robo_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_20_SIZE_14B_RES_480_FPS16_05_20_pretrain2pt2_robo_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_21_SIZE_14B_RES_480p_FPS16_05_22_pretrain2pt2_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_21_SIZE_14B_RES_480p_FPS16_05_22_pretrain2pt2_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_22_SIZE_14B_RES_720_FPS16_T24_05_22_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_22_SIZE_14B_RES_720_FPS16_T24_05_22_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_23_SIZE_14B_RES_720_FPS16_T24_05_22_accumulated_hq_wan_rerun, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_23_SIZE_14B_RES_720_FPS16_T24_05_22_accumulated_hq_wan_rerun), + ], + [ + I2V_STAGE_C_PT_4_INDEX_24_SIZE_14B_RES_720_FPS16_T24_05_27_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_24_SIZE_14B_RES_720_FPS16_T24_05_27_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_25_SIZE_14B_RES_256_FPS16_T24_05_27_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_25_SIZE_14B_RES_256_FPS16_T24_05_27_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_26_SIZE_14B_RES_256_FPS16_T24_05_27_accumulated_hq_wan_rerun, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_26_SIZE_14B_RES_256_FPS16_T24_05_27_accumulated_hq_wan_rerun), + ], + [ + I2V_STAGE_C_PT_4_INDEX_27_SIZE_14B_RES_480p_FPS16_T24_05_28_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_27_SIZE_14B_RES_480p_FPS16_T24_05_28_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_28_SIZE_14B_RES_480p_FPS16_T24_05_28_accumulated_hq_wan_tune_pmean_from_index27_24k, + *build_debug_runs( + I2V_STAGE_C_PT_4_INDEX_28_SIZE_14B_RES_480p_FPS16_T24_05_28_accumulated_hq_wan_tune_pmean_from_index27_24k + ), + ], + [ + I2V_STAGE_C_PT_4_INDEX_29_SIZE_14B_RES_480p_FPS16_T24_06_01_accumulated_hq_wan, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_29_SIZE_14B_RES_480p_FPS16_T24_06_01_accumulated_hq_wan), + ], + [ + I2V_STAGE_C_PT_4_INDEX_30_SIZE_14B_RES_480p_FPS16_T24_HQ_V0, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_30_SIZE_14B_RES_480p_FPS16_T24_HQ_V0), + ], + [ + I2V_STAGE_C_PT_4_INDEX_31_SIZE_14B_RES_720p_FPS16_T24_HQ_V0, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_31_SIZE_14B_RES_720p_FPS16_T24_HQ_V0), + ], + [ + I2V_STAGE_C_PT_4_INDEX_32_SIZE_14B_RES_480p_FPS16_T24_06_02_data_from_29, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_32_SIZE_14B_RES_480p_FPS16_T24_06_02_data_from_29), + ], + [ + I2V_STAGE_C_PT_4_INDEX_33_SIZE_14B_RES_720_FPS16_T24_06_02_data_tune_sigma_extrahigh_from_32, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_33_SIZE_14B_RES_720_FPS16_T24_06_02_data_tune_sigma_extrahigh_from_32), + ], + [ + I2V_STAGE_C_PT_4_INDEX_34_SIZE_14B_RES_720p_FPS16_T24_HQ_V0_FROM_31, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_34_SIZE_14B_RES_720p_FPS16_T24_HQ_V0_FROM_31), + ], + [ + I2V_STAGE_C_PT_4_INDEX_35_SIZE_14B_RES_720_FPS16_T24_06_04_data_tune_sigma_from_33, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_35_SIZE_14B_RES_720_FPS16_T24_06_04_data_tune_sigma_from_33), + ], + [ + I2V_STAGE_C_PT_4_INDEX_36_SIZE_14B_RES_720_FPS16_T24_06_04_data_tune_sigma_from_33, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_36_SIZE_14B_RES_720_FPS16_T24_06_04_data_tune_sigma_from_33), + ], + [ + I2V_STAGE_C_PT_4_INDEX_37_SIZE_14B_RES_720_FPS16_T24_HQV2_from_35, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_37_SIZE_14B_RES_720_FPS16_T24_HQV2_from_35), + ], + [ + I2V_STAGE_C_PT_4_INDEX_38_SIZE_14B_RES_720_FPS16_T24_HQV3_from_35, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_38_SIZE_14B_RES_720_FPS16_T24_HQV3_from_35), + ], + [ + I2V_STAGE_C_PT_4_INDEX_39_SIZE_14B_RES_720_FPS16_T24_HQV2pt1_from_37, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_39_SIZE_14B_RES_720_FPS16_T24_HQV2pt1_from_37), + ], + [ + I2V_STAGE_C_PT_4_INDEX_40_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_from_38, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_40_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_from_38), + ], + [ + I2V_STAGE_C_PT_4_INDEX_41_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_1080_from_40, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_41_SIZE_14B_RES_720_FPS16_T24_HQV3pt1_1080_from_40), + ], + [ + I2V_STAGE_C_PT_4_INDEX_42_SIZE_14B_RES_720_FPS16_T24_HQV4_1080_from_40, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_42_SIZE_14B_RES_720_FPS16_T24_HQV4_1080_from_40), + ], + [ + I2V_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_T24_HQV5_from_40), + ], + [ + I2V_STAGE_C_PT_4_INDEX_44_SIZE_14B_RES_720_FPS16_T24_HQV6_from_40, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_44_SIZE_14B_RES_720_FPS16_T24_HQV6_from_40), + ], + [ + I2V_STAGE_C_PT_4_INDEX_45_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_35, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_45_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_35), + ], + [ + I2V_STAGE_C_PT_4_INDEX_46_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_35, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_46_SIZE_14B_RES_720_FPS16_T24_HQV2_1_from_35), + ], + [ + I2V_STAGE_C_PT_4_INDEX_47_SIZE_14B_RES_720_FPS16_T24_HQV5_FIX_DATA_from_40, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_47_SIZE_14B_RES_720_FPS16_T24_HQV5_FIX_DATA_from_40), + ], + [ + I2V_STAGE_C_PT_4_INDEX_48_SIZE_14B_RES_720_FPS16_T24_HQV5_VIDOE_ONLY_from_40, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_48_SIZE_14B_RES_720_FPS16_T24_HQV5_VIDOE_ONLY_from_40), + ], + [ + I2V_STAGE_C_PT_4_INDEX_49_SIZE_14B_RES_720_FPS16_T24_HQV5_ROPE_TUNE_from_40, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_49_SIZE_14B_RES_720_FPS16_T24_HQV5_ROPE_TUNE_from_40), + ], + [ + I2V_STAGE_C_PT_4_INDEX_50_SIZE_14B_RES_720_FPS16_T24_HQV5_SHIFT24_from_40, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_50_SIZE_14B_RES_720_FPS16_T24_HQV5_SHIFT24_from_40), + ], + [ + I2V_STAGE_C_PT_4_INDEX_51_SIZE_14B_RES_720_FPS16_T24_DATA_0612_from_43, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_51_SIZE_14B_RES_720_FPS16_T24_DATA_0612_from_43), + ], + [ + I2V_STAGE_C_PT_4_INDEX_52_SIZE_14B_RES_720_FPS16_T24_DATA_0612_from_51, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_52_SIZE_14B_RES_720_FPS16_T24_DATA_0612_from_51), + ], + [ + I2V_STAGE_C_PT_4_INDEX_53_SIZE_14B_RES_720_FPS16_T24_DATA_0612_HARDCODED_20STEPS_from_51, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_53_SIZE_14B_RES_720_FPS16_T24_DATA_0612_HARDCODED_20STEPS_from_51), + ], + [ + I2V_STAGE_C_PT_4_INDEX_54_SIZE_14B_RES_720_FPS16_T24_DATA_0612_from_52, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_54_SIZE_14B_RES_720_FPS16_T24_DATA_0612_from_52), + ], + [ + I2V_STAGE_C_PT_4_INDEX_55_SIZE_14B_RES_720_FPS16_T24_DATA_HQ_V5_from_54, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_55_SIZE_14B_RES_720_FPS16_T24_DATA_HQ_V5_from_54), + ], + # variants + [ + I2V_STAGE_C_PT_4_INDEX_100_SIZE_14B_RES_720_FPS10_T15_HQV5_from_43, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_100_SIZE_14B_RES_720_FPS10_T15_HQV5_from_43), + ], + [ + I2V_STAGE_C_PT_4_INDEX_101_SIZE_14B_RES_480_FPS10_T15_HQV5_from_43, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_101_SIZE_14B_RES_480_FPS10_T15_HQV5_from_43), + ], + [ + I2V_STAGE_C_PT_4_INDEX_102_SIZE_14B_RES_480_FPS16_T24_HQV5_from_43, + *build_debug_runs(I2V_STAGE_C_PT_4_INDEX_102_SIZE_14B_RES_480_FPS16_T24_HQV5_from_43), + ], +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + if _item_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_wo_resume", + node=_item_wo_resume, + ) + if _item_mock_wo_resume is not None: + cs.store( + group="experiment", + package="_global_", + name=f"{_item['job']['name']}_mock_wo_resume", + node=_item_mock_wo_resume, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_14B_RF.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_14B_RF.py new file mode 100644 index 0000000000000000000000000000000000000000..7dc51159504bbd98953a1c802365245ac7de214f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_14B_RF.py @@ -0,0 +1,455 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from copy import deepcopy + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.configs.video2world.experiment.reason_embeddings.model_14b_reason_1p1_rectified_flow import ( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA, +) + +STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_FACE_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_face_focused_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-rf_face_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma/checkpoints/iter_000012500", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000, + ), + every_n_sample_ema=dict( + every_n=1000, + ), + ), + ), + # dataloader_train=dict( + # dataloaders=dict( + # video_data=dict( + # dataloader=dict( + # dataset=dict( + # augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + # ) + # ) + # ) + # ) + # ) + ), + flags={"allow_objects": True}, +) + +STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_CROWDED_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_crowded_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-rf_crowded_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma/checkpoints/iter_000012500", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000, + ), + every_n_sample_ema=dict( + every_n=1000, + ), + ), + ), + # dataloader_train=dict( + # dataloaders=dict( + # video_data=dict( + # dataloader=dict( + # dataset=dict( + # augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + # ) + # ) + # ) + # ) + # ) + ), + flags={"allow_objects": True}, +) + +STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_HIGH_MOTION_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_high_motion_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-rf_high_motion_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma/checkpoints/iter_000012500", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000, + ), + every_n_sample_ema=dict( + every_n=1000, + ), + ), + ), + # dataloader_train=dict( + # dataloaders=dict( + # video_data=dict( + # dataloader=dict( + # dataset=dict( + # augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + # ) + # ) + # ) + # ) + # ) + ), + flags={"allow_objects": True}, +) + + +STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_ROBOTICS_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_robotics_only_20250717_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-rf_robotics_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma/checkpoints/iter_000012500", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000, + ), + every_n_sample_ema=dict( + every_n=1000, + ), + ), + ), + # dataloader_train=dict( + # dataloaders=dict( + # video_data=dict( + # dataloader=dict( + # dataset=dict( + # augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + # ) + # ) + # ) + # ) + # ) + ), + flags={"allow_objects": True}, +) + +STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_AV_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_av_only_20250717_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-rf_av_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma/checkpoints/iter_000012500", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000, + ), + every_n_sample_ema=dict( + every_n=1000, + ), + ), + ), + # dataloader_train=dict( + # dataloaders=dict( + # video_data=dict( + # dataloader=dict( + # dataset=dict( + # augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + # ) + # ) + # ) + # ) + # ) + ), + flags={"allow_objects": True}, +) + +STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_PHYSICAL_AI_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_physical_ai_20250812_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-rf_physical_ai_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma/checkpoints/iter_000012500", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000, + ), + every_n_sample_ema=dict( + every_n=1000, + ), + ), + ), + # dataloader_train=dict( + # dataloaders=dict( + # video_data=dict( + # dataloader=dict( + # dataset=dict( + # augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + # ) + # ) + # ) + # ) + # ) + ), + flags={"allow_objects": True}, +) + +STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_4K_COOLDOWN_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_RESUME_FROM_REASON1P1_RECTIFIED_FLOW_SHIFT5_HIGH_SIGMA['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_4K_20250812_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-43-Size-14B-Res-720-Fps-16-Note-rf_4k_cooldown_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-43-Size-14B-Res-720-Fps-16_resume_from_reason1p1_rectified_flow_shift5_high_sigma/checkpoints/iter_000012500", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000, + ), + every_n_sample_ema=dict( + every_n=1000, + ), + ), + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.0], + warm_up_steps=[0], + cycle_lengths=[50_000], + ), + # dataloader_train=dict( + # dataloaders=dict( + # video_data=dict( + # dataloader=dict( + # dataset=dict( + # augmentor_name="noframedrop_nocameramove_video_augmentor_v1", + # ) + # ) + # ) + # ) + # ) + ), + flags={"allow_objects": True}, +) + + +variants = [ + ( + "variant1", # smaller lr + { + "scheduler.f_max": [0.4], + "scheduler.f_min": [0.1], + }, + ), +] + + +def apply_kv_to_config(config, key, value): + """ + The key is dot seperated, e.g. "model.config.sde.p_mean" + Creates and returns a new config with only the specified key modified. + When the full key path is in the config, the value is updated. + Otherwise, the key is added to the config. + """ + # Create a deep copy of the config + new_config = deepcopy(config) + + parts = key.split(".") + current = new_config + + # Navigate to the parent of the final attribute, creating missing parts as needed + for i, part in enumerate(parts[:-1]): + if isinstance(current, (dict, LazyDict)): + if part not in current: + # Create a new LazyDict for missing intermediate keys + current[part] = dict() + current = current[part] + else: + # For object attributes + if not hasattr(current, part): + # Create a new LazyDict for missing intermediate attributes + setattr(current, part, dict()) + current = getattr(current, part) + + # Set the value on the final attribute + final_key = parts[-1] + if isinstance(current, (dict, LazyDict)): + current[final_key] = value + else: + setattr(current, final_key, value) + + return new_config + + +cs = ConfigStore.instance() + +for _item in [ + STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_FACE_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_CROWDED_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_HIGH_MOTION_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_ROBOTICS_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_AV_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_PHYSICAL_AI_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_43_SIZE_14B_RES_720_FPS16_4K_COOLDOWN_RF_HIGH_SIGMA, +]: + if isinstance(_item, LazyDict): + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + for variant_name, variant_config in variants: + _item_variant = deepcopy(_item) + # Apply all overrides from variant_config + # import pdb; pdb.set_trace() + for key, value in variant_config.items(): + _item_variant = apply_kv_to_config(_item_variant, key, value) + + # Update the job name to include the variant + _item_variant["job"]["name"] = f"{_item['job']['name']}_{variant_name}" + + # Store the variant configuration + cs.store(group="experiment", package="_global_", name=f"{_item_variant['job']['name']}", node=_item_variant) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B.py new file mode 100644 index 0000000000000000000000000000000000000000..25e0aefa74c92e5196fe6ba4b7bf39f9cb244110 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B.py @@ -0,0 +1,870 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math +from copy import deepcopy + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.configs.video2world.experiment.reason_embeddings.model_2B_reason_1p1 import ( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16, +) +from cosmos_policy._src.predict2.configs.video2world.experiment.reason_embeddings.stage3_2B import ( + T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED, +) + +I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_face_focused_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-3-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_face_focused", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + resolution="720", + scaling="rectified_flow", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) +I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED_FORMAL = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_face_focused_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-3-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_face_focused_formal", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + resolution="720", + scaling="rectified_flow", + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_2_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_crowded_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_crowded", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + resolution="720", + scaling="rectified_flow", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) +I2V_STAGE_C_PT_4_INDEX_2_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED_FORMAL = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_crowded_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_crowded_formal", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + resolution="720", + scaling="rectified_flow", + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_4_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_HIGH_MOTION = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_high_motion_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-4-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_high_motion", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + resolution="720", + scaling="rectified_flow", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) +I2V_STAGE_C_PT_4_INDEX_4_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_HIGH_MOTION_FORMAL = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_high_motion_20250607_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-4-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_high_motion_formal", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + resolution="720", + scaling="rectified_flow", + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_101_SIZE_2B_RES_720_FPS16_ROBOTICS = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_robotics_only_20250717_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-101-Size-2B-Res-720-Fps-16-Note-HQ_V7_robotics", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + resolution="720", + scaling="rectified_flow", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) +I2V_STAGE_C_PT_4_INDEX_101_SIZE_2B_RES_720_FPS16_ROBOTICS_FORMAL = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_robotics_only_20250717_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-101-Size-2B-Res-720-Fps-16-Note-HQ_V7_robotics_formal", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + resolution="720", + scaling="rectified_flow", + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) + +I2V_STAGE_C_PT_4_INDEX_201_SIZE_2B_RES_720_FPS16_FISHEYE = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + {"override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_fisheye_20250806_s3"}, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-201-Size-2B-Res-720-Fps-16-Note-HQ_V7_fisheye", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + resolution="720", + scaling="rectified_flow", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) + + +I2V_STAGE_C_PT_4_INDEX_301_SIZE_2B_RES_720_FPS16_AV = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_av_only_20250717_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-301-Size-2B-Res-720-Fps-16-Note-HQ_V7_av", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + resolution="720", + scaling="rectified_flow", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) +I2V_STAGE_C_PT_4_INDEX_301_SIZE_2B_RES_720_FPS16_AV_FORMAL = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_av_only_20250717_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-301-Size-2B-Res-720-Fps-16-Note-HQ_V7_av_formal", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + resolution="720", + scaling="rectified_flow", + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000010000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) +I2V_STAGE_C_PT_4_INDEX_401_SIZE_2B_RES_720_FPS16_PHYSICAL_AI = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_physical_ai_20250812_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-401-Size-2B-Res-720-Fps-16-Note-HQ_V7_physical_ai", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + resolution="720", + scaling="rectified_flow", + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) +I2V_STAGE_C_PT_4_INDEX_401_SIZE_2B_RES_720_FPS16_PHYSICAL_AI_FORMAL = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_HQ_V6_from_22_WD_HIGH_SIGMA_LOSS_REWEIGHTED['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_physical_ai_20250812_s3" + }, + "_self_", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-401-Size-2B-Res-720-Fps-16-Note-HQ_V7_physical_ai_formal", + ), + model_parallel=dict( + context_parallel_size=2, + ), + model=dict( + # The base config already have those, but still explicitly set them here for clarity + config=dict( + resolution="720", + scaling="rectified_flow", + min_num_conditional_frames=0, + max_num_conditional_frames=2, + conditional_frames_probs={0: 0.5, 1: 0.25, 2: 0.25}, + conditioner=dict( + use_video_condition=dict( + dropout_rate=0.0, + ), + text=dict( + dropout_rate=0.2, + ), + ), + sde=dict( + p_mean=math.log(5.0), + p_std=1.0, + sigma_max=200, + sigma_min=0.01, + ), + ), + ), + checkpoint=dict( + save_iter=2_000, + # final lr ~0.00002 - 0.000025 + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + save_to_object_store=dict( + enabled=True, + ), + load_from_object_store=dict( + enabled=True, + ), + ), + trainer=dict( + max_iter=200_000, + logging_iter=20, + straggler_detection=dict( + enabled=False, + ), + ), + ), + flags={"allow_objects": True}, +) + +variants = [ + ( + "variant1", # smaller lr + { + "scheduler.f_max": [0.4], + "scheduler.f_min": [0.1], + }, + ), +] + + +def apply_kv_to_config(config, key, value): + """ + The key is dot seperated, e.g. "model.config.sde.p_mean" + Creates and returns a new config with only the specified key modified. + When the full key path is in the config, the value is updated. + Otherwise, the key is added to the config. + """ + # Create a deep copy of the config + new_config = deepcopy(config) + + parts = key.split(".") + current = new_config + + # Navigate to the parent of the final attribute, creating missing parts as needed + for i, part in enumerate(parts[:-1]): + if isinstance(current, (dict, LazyDict)): + if part not in current: + # Create a new LazyDict for missing intermediate keys + current[part] = dict() + current = current[part] + else: + # For object attributes + if not hasattr(current, part): + # Create a new LazyDict for missing intermediate attributes + setattr(current, part, dict()) + current = getattr(current, part) + + # Set the value on the final attribute + final_key = parts[-1] + if isinstance(current, (dict, LazyDict)): + current[final_key] = value + else: + setattr(current, final_key, value) + + return new_config + + +cs = ConfigStore.instance() + +for _item in [ + I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED, + I2V_STAGE_C_PT_4_INDEX_2_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED, + I2V_STAGE_C_PT_4_INDEX_101_SIZE_2B_RES_720_FPS16_ROBOTICS, + I2V_STAGE_C_PT_4_INDEX_201_SIZE_2B_RES_720_FPS16_FISHEYE, + I2V_STAGE_C_PT_4_INDEX_301_SIZE_2B_RES_720_FPS16_AV, + I2V_STAGE_C_PT_4_INDEX_4_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_HIGH_MOTION, + I2V_STAGE_C_PT_4_INDEX_401_SIZE_2B_RES_720_FPS16_PHYSICAL_AI, + I2V_STAGE_C_PT_4_INDEX_3_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED_FORMAL, + I2V_STAGE_C_PT_4_INDEX_2_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED_FORMAL, + I2V_STAGE_C_PT_4_INDEX_101_SIZE_2B_RES_720_FPS16_ROBOTICS_FORMAL, + I2V_STAGE_C_PT_4_INDEX_301_SIZE_2B_RES_720_FPS16_AV_FORMAL, + I2V_STAGE_C_PT_4_INDEX_4_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_HIGH_MOTION_FORMAL, + I2V_STAGE_C_PT_4_INDEX_401_SIZE_2B_RES_720_FPS16_PHYSICAL_AI_FORMAL, +]: + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + for variant_name, variant_config in variants: + _item_variant = deepcopy(_item) + # Apply all overrides from variant_config + # import pdb; pdb.set_trace() + for key, value in variant_config.items(): + _item_variant = apply_kv_to_config(_item_variant, key, value) + + # Update the job name to include the variant + _item_variant["job"]["name"] = f"{_item['job']['name']}_{variant_name}" + + # Store the variant configuration + cs.store(group="experiment", package="_global_", name=f"{_item_variant['job']['name']}", node=_item_variant) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py new file mode 100644 index 0000000000000000000000000000000000000000..a5f9887e8e945ff210a66a152c210074ba894d4e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py @@ -0,0 +1,866 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from copy import deepcopy + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.predict2.configs.video2world.experiment.reason_embeddings.model_2B_reason_1p1_rectified_flow import ( + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW, + T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED, +) + +# continual training w/ base RF config +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED_RF_ONLY2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_human_only_face_focused_20250607_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_only2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-3-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_face_focused_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +# continual training w/ base RF config + high sigma +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_human_only_face_focused_20250607_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-3-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_face_focused_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +# continual training w/ base RF config + high sigma + start w/ 23k base + fix conditional frame timestep discrependcy (MR 5033) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED_RF_HIGH_SIGMA2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_face_focused_20250607_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_resume2/checkpoints/iter_000023000", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000, + ), + every_n_sample_ema=dict( + every_n=1000, + ), + ), + ), + ), + flags={"allow_objects": True}, +) +# continual training w/ base RF config + new prompt image data + no drop frame augmentor + gt720p resolution +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED_RF = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-3-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_face_focused_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + dataset_name="cosmos_posttraining_hq_v7_human_only_face_focused_20250607_video_whole", + ) + ) + ) + ) + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED_RF_ONLY2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_human_only_crowded_20250607_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_only2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_crowded_formal/checkpoints/iter_000032000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_human_only_crowded_20250607_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_crowded_formal/checkpoints/iter_000032000/", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED_RF_HIGH_SIGMA2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_human_only_crowded_20250607_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_resume2/checkpoints/iter_000023000", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000, + ), + every_n_sample_ema=dict( + every_n=1000, + ), + ), + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED_RF = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_crowded_formal/checkpoints/iter_000032000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + dataset_name="cosmos_posttraining_hq_v7_human_only_crowded_20250607_video_whole", + ) + ) + ) + ) + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_HIGH_MOTION_RF_ONLY2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_human_only_high_motion_20250607_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_only2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-4-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_high_motion_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_HIGH_MOTION_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_human_only_high_motion_20250607_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-4-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_high_motion_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_HIGH_MOTION_RF = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-4-Size-2B-Res-720-Fps-16-Note-HQ_V7_human_only_high_motion_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + dataset_name="cosmos_posttraining_hq_v7_human_only_high_motion_20250607_video_whole", + ) + ) + ) + ) + ), + ), + flags={"allow_objects": True}, +) + +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_ROBOTICS_RF_ONLY2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_robotics_only_20250717_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_only2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-101-Size-2B-Res-720-Fps-16-Note-HQ_V7_robotics_formal/checkpoints/iter_000028000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) + +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_ROBOTICS_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_robotics_only_20250717_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-101-Size-2B-Res-720-Fps-16-Note-HQ_V7_robotics_formal/checkpoints/iter_000028000/", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_ROBOTICS_RF_HIGH_SIGMA2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250702_video_cosmos_posttraining_hq_v7_robotics_only_20250717_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted_1_1_rectified_flow_only_resume2/checkpoints/iter_000023000", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + callbacks=dict( + every_n_sample_reg=dict( + every_n=1000, + ), + every_n_sample_ema=dict( + every_n=1000, + ), + ), + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_ROBOTICS_RF = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-101-Size-2B-Res-720-Fps-16-Note-HQ_V7_robotics_formal/checkpoints/iter_000028000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + dataset_name="cosmos_posttraining_hq_v7_robotics_only_20250717_video_whole", + ) + ) + ) + ) + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_AV_RF_ONLY2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_av_only_20250717_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_only2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-301-Size-2B-Res-720-Fps-16-Note-HQ_V7_av_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_AV_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_av_only_20250717_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_high_sigma", + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-301-Size-2B-Res-720-Fps-16-Note-HQ_V7_av_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_AV_RF = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-301-Size-2B-Res-720-Fps-16-Note-HQ_V7_av_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + dataset_name="cosmos_posttraining_hq_v7_av_only_20250717_video_whole", + ) + ) + ) + ) + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_PHYSICAL_AI_RF_ONLY2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_physical_ai_20250812_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_only2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-401-Size-2B-Res-720-Fps-16-Note-HQ_V7_physical_ai_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_PHYSICAL_AI_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_physical_ai_20250812_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-401-Size-2B-Res-720-Fps-16-Note-HQ_V7_physical_ai_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_PHYSICAL_AI_RF = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-Index-401-Size-2B-Res-720-Fps-16-Note-HQ_V7_physical_ai_formal/checkpoints/iter_000030000/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + dataset_name="cosmos_posttraining_hq_v7_physical_ai_20250812_video_whole", + ) + ) + ) + ) + ), + ), + flags={"allow_objects": True}, +) + +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_4KCOOLDOWN_RF_ONLY2 = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_4K_20250812_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_only2", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_4K_cooldown_from_10K/checkpoints/iter_000047500/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.0], + warm_up_steps=[0], + cycle_lengths=[50_000], + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_4KCOOLDOWN_RF_HIGH_SIGMA = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + { + "override /data_train": "image_cosmos_pretrain_and_synthetic_20250520_video_cosmos_posttraining_hq_v7_4K_20250812_s3" + }, + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_high_sigma", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_4K_cooldown_from_10K/checkpoints/iter_000047500/", + load_training_state=False, + strict_resume=True, + ), + model=dict( + config=dict( + use_high_sigma_strategy=True, + ), + ), + trainer=dict( + logging_iter=20, + ), + scheduler=dict( + f_max=[0.4], + f_min=[0.0], + warm_up_steps=[0], + cycle_lengths=[50_000], + ), + ), + flags={"allow_objects": True}, +) +STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_4KCOOLDOWN_RF = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_IMPROVED['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown", + ), + checkpoint=dict( + save_iter=1000, + load_path="cosmos_diffusion_v2/official_runs_text2world/Stage-c_pt_4-reason_embeddings-v1p1-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_4K_cooldown_from_10K/checkpoints/iter_000047500/", + load_training_state=False, + strict_resume=True, + ), + trainer=dict( + logging_iter=20, + ), + dataloader_train=dict( + dataloaders=dict( + video_data=dict( + dataloader=dict( + dataset=dict( + dataset_name="cosmos_posttraining_hq_v7_4K_20250812_video_whole", + ) + ) + ) + ) + ), + ), + flags={"allow_objects": True}, +) + +# This is not for training, but a config to load the edm ckpt (e.g. the merged predict 2.5, or edm sft ckpt etc) +# and add the needed overrides to load the edm ckpt and run rf inference +STAGE_C_PT_4_INDEX_2_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_WITH_EDM_CKPT = LazyDict( + dict( + defaults=[ + f"/experiment/{T2V_REASON_EMBEDDINGS_V1P1_STAGE_C_PT_4_INDEX_26_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW['job']['name']}", + ], + job=dict( + group="official_runs_vid2vid", + name="Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt", + ), + model=dict( + config=dict( + conditional_frame_timestep=0.1, + use_kerras_sigma_at_inference=True, + ) + ), + ) +) + +variants = [ + ( + "variant1", # smaller lr + { + "scheduler.f_max": [0.4], + "scheduler.f_min": [0.1], + }, + ), +] + + +def apply_kv_to_config(config, key, value): + """ + The key is dot seperated, e.g. "model.config.sde.p_mean" + Creates and returns a new config with only the specified key modified. + When the full key path is in the config, the value is updated. + Otherwise, the key is added to the config. + """ + # Create a deep copy of the config + new_config = deepcopy(config) + + parts = key.split(".") + current = new_config + + # Navigate to the parent of the final attribute, creating missing parts as needed + for i, part in enumerate(parts[:-1]): + if isinstance(current, (dict, LazyDict)): + if part not in current: + # Create a new LazyDict for missing intermediate keys + current[part] = dict() + current = current[part] + else: + # For object attributes + if not hasattr(current, part): + # Create a new LazyDict for missing intermediate attributes + setattr(current, part, dict()) + current = getattr(current, part) + + # Set the value on the final attribute + final_key = parts[-1] + if isinstance(current, (dict, LazyDict)): + current[final_key] = value + else: + setattr(current, final_key, value) + + return new_config + + +cs = ConfigStore.instance() + +for _item in [ + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED_RF, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED_RF, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_HIGH_MOTION_RF, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_ROBOTICS_RF, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_AV_RF, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_PHYSICAL_AI_RF, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_4KCOOLDOWN_RF, + # new set of exp that only do RF, 8/29 + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED_RF_ONLY2, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED_RF_ONLY2, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_HIGH_MOTION_RF_ONLY2, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_ROBOTICS_RF_ONLY2, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_AV_RF_ONLY2, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_PHYSICAL_AI_RF_ONLY2, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_4KCOOLDOWN_RF_ONLY2, + # with high sigma + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_HIGH_MOTION_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_ROBOTICS_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_AV_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_PHYSICAL_AI_RF_HIGH_SIGMA, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_4KCOOLDOWN_RF_HIGH_SIGMA, + # new high sigma w/ 23k base + fix conditional frame timestep discrependcy (MR 5033) + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_FACE_FOCUSED_RF_HIGH_SIGMA2, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_HUMAN_ONLY_CROWDED_RF_HIGH_SIGMA2, + STAGE_C_PT_4_INDEX_1_SIZE_2B_RES_720_FPS16_ROBOTICS_RF_HIGH_SIGMA2, + STAGE_C_PT_4_INDEX_2_SIZE_2B_RES_720_FPS16_RECTIFIED_FLOW_WITH_EDM_CKPT, +]: + log.info(f"Storing {_item['job']['name']}") + cs.store(group="experiment", package="_global_", name=f"{_item['job']['name']}", node=_item) + for variant_name, variant_config in variants: + _item_variant = deepcopy(_item) + # Apply all overrides from variant_config + # import pdb; pdb.set_trace() + for key, value in variant_config.items(): + _item_variant = apply_kv_to_config(_item_variant, key, value) + + # Update the job name to include the variant + _item_variant["job"]["name"] = f"{_item['job']['name']}_{variant_name}" + + # Store the variant configuration + cs.store(group="experiment", package="_global_", name=f"{_item_variant['job']['name']}", node=_item_variant) diff --git a/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentor_provider.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentor_provider.py new file mode 100644 index 0000000000000000000000000000000000000000..931d231b5c0eff93101aebd381f872240299e555 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentor_provider.py @@ -0,0 +1,600 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Optional + +import cosmos_policy._src.imaginaire.datasets.webdataset.augmentors.image.normalize as normalize +import cosmos_policy._src.imaginaire.datasets.webdataset.augmentors.image.padding as padding +import cosmos_policy._src.imaginaire.datasets.webdataset.augmentors.image.resize as resize +import cosmos_policy._src.predict2.datasets.augmentors.append_fps_frames_for_image as append_fps_frames_for_image +import cosmos_policy._src.predict2.datasets.augmentors.caption_filter as caption_filter +import cosmos_policy._src.predict2.datasets.augmentors.merge_datadict as merge_datadict +import cosmos_policy._src.predict2.datasets.augmentors.text_transforms_for_image as text_transforms_for_image +import cosmos_policy._src.predict2.datasets.augmentors.text_transforms_for_video as text_transforms_for_video +import cosmos_policy._src.predict2.datasets.augmentors.video_parsing as video_parsing +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.predict2.datasets.utils import IMAGE_RES_SIZE_INFO, VIDEO_RES_SIZE_INFO + +AUGMENTOR_OPTIONS = {} + +CAMERA_MOVEMENT_PHRASES = [ + # Panning + "camera pan", + "camera pans", + "camera slowly pan", + "camera slowly pans", + "camera quickly pans", + "camera fast pans", + "panning shot", + "panning camera", + "slow pan", + "quick pan", + "fast pan", + "pan across", + "pan around", + "pan shot", + "panoramic shot", + # Tracking / Dolly + "camera moves", + "camera slowly moves", + "camera quickly moves", + "moving camera", + "tracking shot", + "tracking camera", + "dolly shot", + "dolly in", + "dolly out", + "camera follows", + "camera tracks", + "tracking movement", + # Sweeps / Rotations + "sweeping camera", + "camera sweep", + "rotating camera", + "camera rotation", + "camera rotates", + "camera circles around", + # Tilts + "camera tilt", + "camera tilts", + "camera slowly tilts", + "tilting camera", + "tilt up", + "tilt down", + # Zooms + "camera zoom", + "camera zooms", + "zooming camera", + "zoom in", + "zoom out", + # Handheld / Shake + "handheld camera", + "handheld shot", + "shaky camera", + "camera shake", + "shaky shot", + "handheld movement", +] + + +def augmentor_register(key): + log.info(f"registering {key}...") + + def decorator(func): + AUGMENTOR_OPTIONS[key] = func + return func + + return decorator + + +def get_video_text_transform( + caption_type: str, + embedding_type: Optional[str] = "t5_xxl", + long_caption_ratio: int = 7, + medium_caption_ratio: int = 2, + short_caption_ratio: int = 1, + user_caption_ratio: int = 90, + num_video_frames: int = -1, +): + del num_video_frames + if caption_type == "vila_caption": + video_text_transform = L(text_transforms_for_video.TextTransformForVideo)( + input_keys=[], + args={ + "captions_key": "metas", + "embeddings_key": embedding_type, + "caption_windows_key": "windows", + "caption_type": "vila_caption", + "embedding_caption_type": "vila_caption", + "t5_tokens": {"num": 512}, + "is_mask_all_ones": True, + }, + ) + elif caption_type == "t2w_qwen2p5_7b": + log.info( + f"caption_type: {caption_type}, long_caption_ratio: {long_caption_ratio}, medium_caption_ratio: {medium_caption_ratio}, short_caption_ratio: {short_caption_ratio}, user_caption_ratio: {user_caption_ratio}" + ) + video_text_transform = L(text_transforms_for_video.TextTransformForVideo)( + input_keys=[], + args={ + "captions_key": "metas", + "embeddings_key": embedding_type, + "caption_windows_key": "t2w_windows", + "caption_type": "qwen2p5_7b_caption", + "embedding_caption_type": "t2w_qwen2p5_7b", + "t5_tokens": {"num": 512}, + "is_mask_all_ones": True, + "caption_probs": { + "long": long_caption_ratio, + "medium": medium_caption_ratio, + "short": short_caption_ratio, + "user": user_caption_ratio, + }, + }, + ) + elif caption_type == "i2w_qwen2p5_7b_later_frames": + video_text_transform = L(text_transforms_for_video.TextTransformForVideo)( + input_keys=[], + args={ + "captions_key": "metas", + "embeddings_key": embedding_type, + "caption_windows_key": "i2w_windows_later_frames", + "caption_type": "qwen2p5_7b_caption", + "embedding_caption_type": "i2w_qwen2p5_7b_later_frames", + "t5_tokens": {"num": 512}, + "is_mask_all_ones": True, + "caption_probs": { + "long": long_caption_ratio, + "medium": medium_caption_ratio, + "short": short_caption_ratio, + "user": user_caption_ratio, + }, + }, + ) + elif caption_type == "t2w_qwen3_vl_30b_a3b": + log.info( + f"caption_type: {caption_type}, long_caption_ratio: {long_caption_ratio}, medium_caption_ratio: {medium_caption_ratio}, short_caption_ratio: {short_caption_ratio}, user_caption_ratio: {user_caption_ratio}" + ) + video_text_transform = L(text_transforms_for_video.TextTransformForVideo)( + input_keys=[], + args={ + "captions_key": "metas", + "embeddings_key": embedding_type, + "caption_windows_key": "t2w_windows", + "caption_type": "qwen3_vl_30b_a3b_caption", + "embedding_caption_type": caption_type, + "t5_tokens": {"num": 512}, + "is_mask_all_ones": True, + "caption_probs": { + "long": long_caption_ratio, + "medium": medium_caption_ratio, + "short": short_caption_ratio, + "user": user_caption_ratio, + }, + }, + ) + elif caption_type == "i2w_qwen3_vl_30b_a3b_later_frames": + video_text_transform = L(text_transforms_for_video.TextTransformForVideo)( + input_keys=[], + args={ + "captions_key": "metas", + "embeddings_key": embedding_type, + "caption_windows_key": "i2w_windows_later_frames", + "caption_type": "qwen3_vl_30b_a3b_caption", + "embedding_caption_type": "i2w_qwen3_vl_30b_a3b_later_frames", + "t5_tokens": {"num": 512}, + "is_mask_all_ones": True, + "caption_probs": { + "long": long_caption_ratio, + "medium": medium_caption_ratio, + "short": short_caption_ratio, + "user": user_caption_ratio, + }, + }, + ) + else: + raise ValueError(f"Unsupported caption type ({caption_type}) for video data") + + return video_text_transform + + +@augmentor_register("video_basic_augmentor_v1") +def get_video_augmentor_v1( + resolution: str, + caption_type: str = "vila_caption", + embedding_type: str = "t5_xxl", + min_fps: int = 10, + max_fps: int = 60, + long_caption_ratio: int = 7, + medium_caption_ratio: int = 2, + short_caption_ratio: int = 1, + user_caption_ratio: int = 90, +): + """Video augmentor V1. It relies on a separate video decoder to decode videos of required number of frames. + Augmentors here will resize the video, add reflection padding, and extract captions and embeddings. + + Supported caption_type include vila_caption. + Supported embedding_type include t5_xxl. + """ + assert caption_type == "vila_caption", f"Unsupported caption type ({caption_type}) for video data" + assert embedding_type == "t5_xxl", f"Unsupported embeddings type ({embedding_type}) for video data" + video_text_transform = get_video_text_transform( + caption_type=caption_type, + embedding_type=embedding_type, + long_caption_ratio=long_caption_ratio, + medium_caption_ratio=medium_caption_ratio, + short_caption_ratio=short_caption_ratio, + user_caption_ratio=user_caption_ratio, + ) + + return { + "merge_datadict": L(merge_datadict.DataDictMerger)( + input_keys=["video"], + output_keys=[ + "video", + "fps", + "num_frames", + "chunk_index", + "frame_start", + "frame_end", + "n_orig_video_frames", + ], + ), + "resize_largest_side_aspect_ratio_preserving": L(resize.ResizeLargestSideAspectPreserving)( + input_keys=["video"], + args={"size": VIDEO_RES_SIZE_INFO[resolution]}, + ), + "reflection_padding": L(padding.ReflectionPadding)( + input_keys=["video"], + args={"size": VIDEO_RES_SIZE_INFO[resolution]}, + ), + "text_transform": video_text_transform, + } + + +@augmentor_register("video_basic_augmentor_v2") +def get_video_augmentor_v2( + resolution: str, + caption_type: str = "t2w_qwen2p5_7b", + embedding_type: Optional[str] = "t5_xxl", + min_fps: int = 10, + max_fps: int = 60, + long_caption_ratio: int = 7, + medium_caption_ratio: int = 2, + short_caption_ratio: int = 1, + user_caption_ratio: int = 90, + num_video_frames: int = -1, + use_native_fps: bool = True, + use_original_fps: bool = False, + use_random_consecutive_frames: bool = False, + use_random_interleaved_frames: bool = False, +): + """ + num_video_frames: -1 means use all frames, otherwise use the number of frames specified. + + Video augmentor V2. It works with a naive video decoder ("video_naive_bytes") that does nothing. + Augmentors here include: + - a basic video decoder that fetches frames within a window and delegates further subsampling or duplication to the modeling code to produce videos with the required number of frames. + - resize the video + - add reflection padding + - extract captions and embeddings. + + When use_random_consecutive_frames is True, the augmentor will sample random consecutive frames, preserving the original fps. + When use_random_interleaved_frames is True, the augmentor will sample random interleaved frames, making fractional fps interpolation possible (e.g. 24->30fps). + + Supported caption_type include t2w_qwen2p5_7b and i2w_qwen2p5_7b_later_frames. + Supported embedding_type include t5_xxl and umt5_xxl. + """ + video_text_transform = get_video_text_transform( + caption_type=caption_type, + embedding_type=embedding_type, + long_caption_ratio=long_caption_ratio, + medium_caption_ratio=medium_caption_ratio, + short_caption_ratio=short_caption_ratio, + user_caption_ratio=user_caption_ratio, + ) + if caption_type.startswith("t2w_qwen"): + key_for_caption = "t2w_windows" + elif caption_type.startswith("i2w_qwen"): + key_for_caption = "i2w_windows_later_frames" + else: + f"Unsupported caption type ({caption_type}) for video data" + if embedding_type is not None: + assert embedding_type in ( + "t5_xxl", + "umt5_xxl", + ), f"Unsupported embeddings type ({embedding_type}) for video data" + + return { + "video_parsing": L(video_parsing.VideoParsing)( + input_keys=["metas", "video"], + args={ + "key_for_caption": key_for_caption, + "min_duration": 4.0, + "min_fps": min_fps, + "max_fps": max_fps, + "video_decode_num_threads": 4, + "num_video_frames": num_video_frames, + "use_native_fps": use_native_fps, + "use_original_fps": use_original_fps, + # use_random_consecutive_frames: + # If True, samples random consecutive frames within the window, preserving the original fps between frames. + # This enables consecutive clips from the source, without evenly-spaced dropping/duplication. + "use_random_consecutive_frames": use_random_consecutive_frames, + # use_random_interleaved_frames: + # If True, enables random interleaved frame subsampling (e.g., for fractional fps upsampling/downsampling such as 24->30 FPS). + # Produces non-consecutive, randomly-traced clip segments by mixing different strides, for more varied temporal sampling. + "use_random_interleaved_frames": use_random_interleaved_frames, + }, + ), + "merge_datadict": L(merge_datadict.DataDictMerger)( + input_keys=["video"], + output_keys=[ + "video", + "fps", + "num_frames", + "chunk_index", + "frame_start", + "frame_end", + "n_orig_video_frames", + ], + ), + "resize_largest_side_aspect_ratio_preserving": L(resize.ResizeLargestSideAspectPreserving)( + input_keys=["video"], + args={"size": VIDEO_RES_SIZE_INFO[resolution]}, + ), + "reflection_padding": L(padding.ReflectionPadding)( + input_keys=["video"], + args={"size": VIDEO_RES_SIZE_INFO[resolution]}, + ), + "text_transform": video_text_transform, + } + + +@augmentor_register("noframedrop_nocameramove_video_augmentor_v1") +def get_noframedrop_nocameramove_video_augmentor_v1( + resolution: str, + caption_type: str = "t2w_qwen2p5_7b", + embedding_type: Optional[str] = "t5_xxl", + min_fps: int = 10, + max_fps: int = 60, + long_caption_ratio: int = 7, + medium_caption_ratio: int = 2, + short_caption_ratio: int = 1, + user_caption_ratio: int = 90, + num_video_frames: int = -1, + use_native_fps: bool = True, + use_original_fps: bool = False, + use_random_consecutive_frames: bool = False, +): + """ + This augmentor is v2 + the following: + - no frame drop by ensure num_multipler is always 1 + - no camera move (indiciated by the camera related bad words in the caption) + """ + video_text_transform = get_video_text_transform( + caption_type=caption_type, + embedding_type=embedding_type, + long_caption_ratio=long_caption_ratio, + medium_caption_ratio=medium_caption_ratio, + short_caption_ratio=short_caption_ratio, + user_caption_ratio=user_caption_ratio, + ) + if caption_type.startswith("t2w_qwen"): + key_for_caption = "t2w_windows" + elif caption_type.startswith("i2w_qwen"): + key_for_caption = "i2w_windows_later_frames" + else: + f"Unsupported caption type ({caption_type}) for video data" + if embedding_type is not None: + assert embedding_type in ( + "t5_xxl", + "umt5_xxl", + ), f"Unsupported embeddings type ({embedding_type}) for video data" + + contain_keyword = False # ensure no camera move + augmentations = { + "video_parsing": L(video_parsing.VideoParsing)( + input_keys=["metas", "video"], + args={ + "key_for_caption": key_for_caption, + "min_duration": 4.0, + "min_fps": min_fps, + "max_fps": max_fps, + "video_decode_num_threads": 4, + "num_video_frames": num_video_frames, + "use_native_fps": use_native_fps, + "use_original_fps": use_original_fps, + "use_random_consecutive_frames": use_random_consecutive_frames, + # Both use_original_fps=True and "allowed_num_multiplers": [1] prevent frame dropping. + # Key differences: + # - use_original_fps=True: Hard-codes num_multiplier=1 and ignores allowed_num_multiplers setting. + # Won't skip entire videos, but may discard head/tail frames, potentially causing + # video-caption misalignment. + # - "allowed_num_multiplers": [1]: Uses the multiplier system but restricts it to 1x only. May skip videos, causing slower dataloader + "allowed_num_multiplers": [1], + }, + ), + "merge_datadict": L(merge_datadict.DataDictMerger)( + input_keys=["video"], + output_keys=[ + "video", + "fps", + "num_frames", + "chunk_index", + "frame_start", + "frame_end", + "n_orig_video_frames", + ], + ), + "resize_largest_side_aspect_ratio_preserving": L(resize.ResizeLargestSideAspectPreserving)( + input_keys=["video"], + args={"size": VIDEO_RES_SIZE_INFO[resolution]}, + ), + "reflection_padding": L(padding.ReflectionPadding)( + input_keys=["video"], + args={"size": VIDEO_RES_SIZE_INFO[resolution]}, + ), + "text_transform": video_text_transform, + "caption_filter": L(caption_filter.CaptionFilter)( + input_keys=["ai_caption"], # Works with ai_caption from TextTransformForVideo + args={ + "keywords": CAMERA_MOVEMENT_PHRASES, + "contain_keyword": contain_keyword, + "log_filtered": False, # Enable logging to see what gets filtered + "filter_stats": True, + # For 4k and physics AI datasets, even if this has camera movement, it is still good + "dont_apply_on_webdataset_names": [ + "4k_", + "a2d2_", + "agibot_", + "alpamayo_", + "bridgev2p1_", + "droid_", + "gr00t_", + "nexar", + "onex", + "openx", + "physical-ai-special", + "physics-cosmos-db", + "wisa", + "robomind", + "smartspace_", + ], + }, + ), + } + mode_str = "contain" if contain_keyword else "exclude" + log.info( + f"[video] noframedrop_nocameramove_video_augmentor_v1: Added caption filter in '{mode_str}' mode " + f"with {len(CAMERA_MOVEMENT_PHRASES)} camera movement phrases" + ) + return augmentations + + +@augmentor_register("nocameramove_video_augmentor_v1") +def get_nocameramove_video_augmentor_v1( + resolution: str, + caption_type: str = "t2w_qwen2p5_7b", + embedding_type: Optional[str] = "t5_xxl", + min_fps: int = 10, + max_fps: int = 60, + long_caption_ratio: int = 7, + medium_caption_ratio: int = 2, + short_caption_ratio: int = 1, + user_caption_ratio: int = 90, + num_video_frames: int = -1, + use_native_fps: bool = True, + use_original_fps: bool = False, + use_random_consecutive_frames: bool = False, +): + """ + This augmentor is based on noframedrop_nocameramove_video_augmentor_v1 but: + - allows limited frame drop by setting allowed_num_multiplers to [1,2] + - no camera move (indicated by the camera related bad words in the caption) + """ + # Get the base augmentations from the no-frame-drop version + augmentations = get_noframedrop_nocameramove_video_augmentor_v1( + resolution=resolution, + caption_type=caption_type, + embedding_type=embedding_type, + min_fps=min_fps, + max_fps=max_fps, + long_caption_ratio=long_caption_ratio, + medium_caption_ratio=medium_caption_ratio, + short_caption_ratio=short_caption_ratio, + user_caption_ratio=user_caption_ratio, + num_video_frames=num_video_frames, + use_native_fps=use_native_fps, + use_original_fps=use_original_fps, + use_random_consecutive_frames=use_random_consecutive_frames, + ) + + # Modify only the allowed_num_multiplers parameter + augmentations["video_parsing"].args["allowed_num_multiplers"] = [1, 2] + + log.info( + "[video] nocameramove_video_augmentor_v1: Modified allowed_num_multiplers to [1, 2] " + "for limited frame dropping capability" + ) + return augmentations + + +@augmentor_register("image_basic_augmentor") +def get_image_augmentor( + resolution: str, + caption_type: str = "ai_v3p1", + embedding_type: str = "t5_xxl", +): + augmentation = { + "resize_largest_side_aspect_ratio_preserving": L(resize.ResizeLargestSideAspectPreserving)( + input_keys=["images"], + args={"size": IMAGE_RES_SIZE_INFO[resolution]}, + ), + "reflection_padding": L(padding.ReflectionPadding)( + input_keys=["images"], + args={"size": IMAGE_RES_SIZE_INFO[resolution]}, + ), + "normalize": L(normalize.Normalize)( + input_keys=["images"], + args={"mean": 0.5, "std": 0.5}, + ), + "text_transform": L(text_transforms_for_image.TextTransformForImage)( + input_keys=[], + args={ + "caption_type": caption_type, + "embedding_type": embedding_type, + "weight_captions_gt": 0.05, + "caption_probs": {"ground_truth": 0.05, "vfc_fidelity": 0.95}, + "t5_tokens": {"num": 512, "dim": 1024}, + "is_mask_all_ones": True, + }, + ), + "append_fps_frames": L(append_fps_frames_for_image.AppendFPSFramesForImage)(), + } + + return augmentation + + +@augmentor_register("image_basic_augmentor_without_embeddings") +def get_image_augmentor_without_embeddings( + resolution: str, + caption_type: str = "ai_v3p1", + embedding_type: Optional[str] = None, +): + augmentation = { + "resize_largest_side_aspect_ratio_preserving": L(resize.ResizeLargestSideAspectPreserving)( + input_keys=["images"], + args={"size": IMAGE_RES_SIZE_INFO[resolution]}, + ), + "reflection_padding": L(padding.ReflectionPadding)( + input_keys=["images"], + args={"size": IMAGE_RES_SIZE_INFO[resolution]}, + ), + "normalize": L(normalize.Normalize)( + input_keys=["images"], + args={"mean": 0.5, "std": 0.5}, + ), + "text_transform": L(text_transforms_for_image.TextTransformForImageWithoutEmbeddings)( + input_keys=[], + args={ + "caption_type": caption_type, + }, + ), + "append_fps_frames": L(append_fps_frames_for_image.AppendFPSFramesForImage)(), + } + + return augmentation diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/append_fps_frames_for_image.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/append_fps_frames_for_image.py new file mode 100644 index 0000000000000000000000000000000000000000..b0b2daf12c7c11efe750a7461039204112868002 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/append_fps_frames_for_image.py @@ -0,0 +1,37 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Optional + +from cosmos_policy._src.imaginaire.datasets.webdataset.augmentors.augmentor import Augmentor + + +class AppendFPSFramesForImage(Augmentor): + def __init__( + self, input_keys: Optional[list] = None, output_keys: Optional[list] = None, args: Optional[dict] = None + ) -> None: + super().__init__(input_keys, output_keys, args) + + def __call__(self, data_dict: dict) -> dict: + r"""Remove the input keys from the data dict. + + Args: + data_dict (dict): Input data dict + Returns: + data_dict (dict): Output dict with keys removed. + """ + data_dict["fps"] = 30.0 # set image model fps = 30, which is the most common fps we used to train video. + data_dict["num_frames"] = 1 + return data_dict diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/caption_filter.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/caption_filter.py new file mode 100644 index 0000000000000000000000000000000000000000..4bae2feb0218aabaead269b59904a56322268cd3 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/caption_filter.py @@ -0,0 +1,173 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Optional + +from cosmos_policy._src.imaginaire.datasets.webdataset.augmentors.augmentor import Augmentor +from cosmos_policy._src.imaginaire.utils import log + + +class CaptionFilter(Augmentor): + """ + Caption filter augmentor for predict2 training. + + This augmentor filters video samples based on caption content with configurable behavior: + - contain_keyword=True: Only return videos that contain keywords in captions + - contain_keyword=False: Only return videos that do NOT contain keywords in captions + + When a sample doesn't match the filter criteria, it returns None, which causes + the webdataset pipeline to skip that sample and continue to the next one. + """ + + def __init__(self, input_keys: list, output_keys: Optional[list] = None, args: Optional[dict] = None) -> None: + """ + Initialize the caption filter. + + Args: + input_keys: List containing the caption key (e.g., ["ai_caption"] or text embeddings key) + output_keys: Not used for filtering, can be None + args: Dictionary with filtering parameters: + - "keywords": List of keywords to filter by (e.g., ["camera pan"]) + - "contain_keyword": Boolean flag for filtering behavior: + * True: Only return videos that contain keywords + * False: Only return videos that do NOT contain keywords + - "log_filtered": Whether to log filtered samples (default: False) + - "filter_stats": Whether to track filtering statistics (default: True) + - "dont_apply_on_webdataset_names": List of webdataset names to not apply the filter on, it will just pass through without checking contain or not contain keywords + """ + super().__init__(input_keys, output_keys, args) + + # Parse arguments + if args is None: + args = {} + + self.keywords = args.get("keywords", []) + self.contain_keyword = args.get("contain_keyword", False) # Default to exclude mode + self.log_filtered = args.get("log_filtered", False) + self.filter_stats = args.get("filter_stats", True) + self.dont_apply_on_webdataset_names = args.get("dont_apply_on_webdataset_names", []) + + # Validate input_keys + if not input_keys or len(input_keys) == 0: + raise ValueError("CaptionFilter requires at least one input key for the caption field") + + self.caption_key = input_keys[0] # Use the first input key as the caption key + + # Statistics tracking + if self.filter_stats: + self.total_samples = 0 + self.filtered_samples = 0 + + # Validate configuration + if not self.keywords: + log.warning("CaptionFilter: No keywords provided, filter will not filter any samples") + + mode_str = "contain" if self.contain_keyword else "exclude" + log.info( + f"CaptionFilter initialized in '{mode_str}' mode with {len(self.keywords)} keywords using caption key '{self.caption_key}': {self.keywords}" + ) + + def __call__(self, data_dict: dict) -> Optional[dict]: + """ + Filter data based on caption content. + + This checks the caption field specified by the input_keys parameter. + Depending on contain_keyword flag: + - True: Returns data_dict only if caption contains any keyword, None otherwise + - False: Returns data_dict only if caption contains NO keywords, None otherwise + + Args: + data_dict: Input data dictionary containing the caption field specified in input_keys + + Returns: + data_dict: Original data dict if caption passes filter + None: If caption should be filtered out (causes sample to be skipped) + """ + data_dict_root = data_dict["__url__"].root + if any(n in data_dict_root for n in self.dont_apply_on_webdataset_names): + return data_dict + + if self.filter_stats: + self.total_samples += 1 + + # Check if caption key exists + if self.caption_key not in data_dict: + if self.log_filtered: + log.warning(f"CaptionFilter: No '{self.caption_key}' found in data_dict, passing through") + return data_dict + + caption = data_dict[self.caption_key] + if not isinstance(caption, str) or not caption.strip(): + if self.log_filtered: + log.warning(f"CaptionFilter: '{self.caption_key}' is empty or not a string, got {type(caption)}") + return data_dict + + # Check if any keywords are found in the caption + search_caption = caption.lower() + keyword_found = False + matched_keyword = None + + for keyword in self.keywords: + if keyword.lower() in search_caption: + keyword_found = True + matched_keyword = keyword + break + + # Apply filtering logic based on contain_keyword flag + should_filter = False + if self.contain_keyword: + # Include mode: filter out if NO keywords found + should_filter = not keyword_found + else: + # Exclude mode: filter out if ANY keyword found + should_filter = keyword_found + + if should_filter: + if self.log_filtered: + if self.contain_keyword: + log.info(f"CaptionFilter: excluded sample (no keywords found) - caption: '{caption[:100]}...'") + else: + log.info( + f"CaptionFilter: excluded sample due to keyword '{matched_keyword}' - caption: '{caption[:100]}...'" + ) + + if self.filter_stats: + self.filtered_samples += 1 + return None + + # Sample passes filter + return data_dict + + def get_filter_stats(self) -> dict: + """ + Get filtering statistics. + + Returns: + Dictionary with filtering statistics + """ + if not self.filter_stats: + return {"stats_disabled": True} + + filter_rate = (self.filtered_samples / self.total_samples * 100) if self.total_samples > 0 else 0 + mode_str = "contain" if self.contain_keyword else "exclude" + + return { + "total_samples": self.total_samples, + "filtered_samples": self.filtered_samples, + "passed_samples": self.total_samples - self.filtered_samples, + "filter_rate_percent": filter_rate, + "mode": mode_str, + "keywords": self.keywords, + } diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/merge_datadict.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/merge_datadict.py new file mode 100644 index 0000000000000000000000000000000000000000..0bc69364920de855918c5a71ca09b8697446152e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/merge_datadict.py @@ -0,0 +1,50 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Optional + +from cosmos_policy._src.imaginaire.datasets.webdataset.augmentors.augmentor import Augmentor +from cosmos_policy._src.imaginaire.utils import log + + +class DataDictMerger(Augmentor): + def __init__(self, input_keys: list, output_keys: Optional[list] = None, args: Optional[dict] = None) -> None: + super().__init__(input_keys, output_keys, args) + + def __call__(self, data_dict: dict) -> dict: + r"""Merge the dictionary associated with the input keys into data_dict. Only keys in output_keys are merged. + + Args: + data_dict (dict): Input data dict + Returns: + data_dict (dict): Output dict with dictionary associated with the input keys merged. + """ + for key in self.input_keys: + if key not in data_dict: + log.warning( + f"DataDictMerger dataloader error: missing {key}, {data_dict['__url__']}, {data_dict['__key__']}", + rank0_only=False, + ) + return None + key_dict = data_dict.pop(key) + if key == "depth" and "depth" in self.output_keys: + data_dict["depth"] = key_dict + elif key == "segmentation" and "segmentation" in self.output_keys: + data_dict["segmentation"] = key_dict + for sub_key in key_dict: + if sub_key in self.output_keys: + data_dict[sub_key] = key_dict[sub_key] + del key_dict + return data_dict diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/text_transforms_for_image.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/text_transforms_for_image.py new file mode 100644 index 0000000000000000000000000000000000000000..0f82266888787357062884cab81935da92ee9a0a --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/text_transforms_for_image.py @@ -0,0 +1,155 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import random +from typing import Optional + +from cosmos_policy._src.imaginaire.datasets.augmentors.v3_text_transforms import pad_and_resize +from cosmos_policy._src.imaginaire.datasets.webdataset.augmentors.augmentor import Augmentor +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.predict2.datasets.data_sources.data_registration import _CAPTION_EMBEDDING_KEY_MAPPING_IMAGES + +# For the qwen captions, we have 3 variants: short, medium, long +# In addition, for synthetic data, we create prompt embeddings as well. +# There is quite a bit of entropy in the way prompt data is saved. +# Captions are saved as "prompts", while the corresponding embeddings are saved as "original_prompt" +# This part will be cleaned after synthetic data is cleaned to be in the same format as real data. +_AVAILABLE_QWEN_CAPTIONS = ["qwen2p5_7b_short", "qwen2p5_7b_medium", "qwen2p5_7b_long"] +_CAPTION_EMBEDDING_MAPPING = { + "qwen2p5_7b_short": "qwen2p5_7b_short", + "qwen2p5_7b_medium": "qwen2p5_7b_medium", + "qwen2p5_7b_long": "qwen2p5_7b_long", + "prompts": "original_prompt", +} + + +class TextTransformForImage(Augmentor): + def __init__(self, input_keys: list, output_keys: Optional[list] = None, args: Optional[dict] = None) -> None: + super().__init__(input_keys, output_keys, args) + + def __call__(self, data_dict: dict) -> dict: + r"""Performs camera transformation. + + Args: + data_dict (dict): Input data dict + Returns: + data_dict (dict): Output dict with camera attributes added + """ + + caption_type = self.args["caption_type"] + embedding_key_in_dict = _CAPTION_EMBEDDING_KEY_MAPPING_IMAGES[caption_type] + embedding_type = self.args["embedding_type"] + embedding_input_key_prefix = "" if embedding_type == "t5_xxl" else "umt5_" + + captions_key, embeddings_key = ( + f"captions_{caption_type}", + f"{embedding_input_key_prefix}embeddings_captions_{embedding_key_in_dict}", + ) + decoded_captions_ai = data_dict[captions_key] + decoded_embeddings_ai = data_dict[embeddings_key] + + try: + # Hotfix: Some captions are labeled as "captions" and some are labeled as "caption" + # This issue needs to be fixed in the synthetic data. This is a hack and will be removed + # once the data is cleaned. + caption_key = "captions" if "captions" in decoded_captions_ai else "caption" + embedding_key = "t5_xxl_fp8" if embedding_type == "t5_xxl" else "umt5_xxl" + if caption_type == "qwen2p5_7b_v4": + selected_caption_type = random.choice(_AVAILABLE_QWEN_CAPTIONS) + data_dict["ai_caption"] = decoded_captions_ai[caption_key][selected_caption_type] + t5_embedding = decoded_embeddings_ai[selected_caption_type]["embeddings"][embedding_key] + data_dict["selected_caption_type"] = selected_caption_type + elif caption_type == "prompts": + data_dict["ai_caption"] = decoded_captions_ai["caption"]["prompt"] + t5_embedding = decoded_embeddings_ai[_CAPTION_EMBEDDING_MAPPING[caption_type]]["embeddings"][ + embedding_key + ] + data_dict["selected_caption_type"] = caption_type + else: + assert caption_type == "ai_v3p1", f"Caption type {caption_type} not supported" + if decoded_captions_ai["had_parse_issue"]: + data_dict["ai_caption"] = decoded_captions_ai["captions"]["kosmos_2"] + t5_embedding = decoded_embeddings_ai["kosmos2"]["embeddings"][embedding_key] + else: + data_dict["ai_caption"] = decoded_captions_ai["captions"]["vfc"] + t5_embedding = decoded_embeddings_ai["vfc_fidelity"]["embeddings"][embedding_key] + + out_t5, out_t5_mask = pad_and_resize( + t5_embedding, + self.args["t5_tokens"]["num"], + is_mask_all_ones=self.args["is_mask_all_ones"], + ) + data_dict["t5_text_embeddings"] = out_t5 + data_dict["t5_text_mask"] = out_t5_mask + except Exception as e: + log.warning( + f"TextTransform dataloader error: {data_dict['__url__']}, {data_dict['__key__']}\n error {e}", + rank0_only=False, + ) + return None + + del data_dict[captions_key] + del data_dict[embeddings_key] + + return data_dict + + +class TextTransformForImageWithoutEmbeddings(Augmentor): + def __init__(self, input_keys: list, output_keys: Optional[list] = None, args: Optional[dict] = None) -> None: + super().__init__(input_keys, output_keys, args) + + def __call__(self, data_dict: dict) -> dict: + r"""Performs text transform without any embedding loading. + This is useful for online computation. + + Args: + data_dict (dict): Input data dict + Returns: + data_dict (dict): Output dict with camera attributes added + """ + + caption_type = self.args["caption_type"] + captions_key = f"captions_{caption_type}" + decoded_captions_ai = data_dict[captions_key] + + try: + # Hotfix: Some captions are labeled as "captions" and some are labeled as "caption" + # This issue needs to be fixed in the synthetic data. This is a hack and will be removed + # once the data is cleaned. + caption_key = "captions" if "captions" in decoded_captions_ai else "caption" + if caption_type == "qwen2p5_7b_v4": + selected_caption_type = random.choice(_AVAILABLE_QWEN_CAPTIONS) + data_dict["ai_caption"] = decoded_captions_ai[caption_key][selected_caption_type] + data_dict["selected_caption_type"] = selected_caption_type + elif caption_type == "prompts": + data_dict["ai_caption"] = decoded_captions_ai["caption"]["prompt"] + data_dict["selected_caption_type"] = caption_type + else: + assert caption_type == "ai_v3p1", f"Caption type {caption_type} not supported" + if decoded_captions_ai["had_parse_issue"]: + data_dict["ai_caption"] = decoded_captions_ai["captions"]["kosmos_2"] + else: + data_dict["ai_caption"] = decoded_captions_ai["captions"]["vfc"] + + except Exception as e: + log.warning( + f"TextTransform dataloader error: {data_dict['__url__']}, {data_dict['__key__']}\n error {e}", + rank0_only=False, + ) + return None + + del data_dict[captions_key] + + return data_dict diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/text_transforms_for_video.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/text_transforms_for_video.py new file mode 100644 index 0000000000000000000000000000000000000000..88a4425df42feb3c017c4cd17294a816b9cba5d6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/text_transforms_for_video.py @@ -0,0 +1,174 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import random +from typing import Optional + +from cosmos_policy._src.imaginaire.datasets.augmentors.v3_text_transforms import pad_and_resize +from cosmos_policy._src.imaginaire.datasets.webdataset.augmentors.augmentor import Augmentor +from cosmos_policy._src.imaginaire.utils import log + + +class TextTransformForVideo(Augmentor): + def __init__(self, input_keys: dict, output_keys: Optional[list] = None, args: Optional[dict] = None) -> None: + super().__init__(input_keys, output_keys, args) + + # our caption is saved in json with format: {"": "xxx", "": [{"start_frame": x, "end_frame": x, "": xxx}, ...], "": [{"start_frame":...]} + # our t5 embedding is saved in pickle with format: [{"": array1, "": array2}, ...] + self.captions_key: str = args[ + "captions_key" + ] # s3 folder that saves the captions; this get mapped to the key in data_dict to fetch the caption field + self.embeddings_key: Optional[str] = args[ + "embeddings_key" + ] # s3 folder that saves the embeddings; this get mapped to the key in data_dict to fetch the embedding field + self.caption_windows_key: str = args[ + "caption_windows_key" + ] # key to get the caption windows from the caption field + self.caption_type: str = args["caption_type"] # key of caption type to fetch the caption from caption windows + + self._load_embeddings = self.embeddings_key is not None + + if not self._load_embeddings: + # In this case, we don't load the embeddings + log.info("No embeddings key provided, we will not load embeddings") + self.embedding_caption_type = None + self.t5_tokens_num = None + self.is_mask_all_ones = None + self.embedding_style_mapping = None + else: + self.embedding_caption_type: str = args[ + "embedding_caption_type" + ] # key to get the embedding of a particular caption type from the embedding field + self.t5_tokens_num = args["t5_tokens"]["num"] # number of tokens we cap after padding + self.is_mask_all_ones = args["is_mask_all_ones"] # if true, set mask for t5 to all ones + + self.embedding_style_mapping = { + "long": self.embedding_caption_type, + "short": f"{self.embedding_caption_type}_short", + "medium": f"{self.embedding_caption_type}_medium", + "user": f"{self.embedding_caption_type}_user", + } + + self.caption_probs: dict[str, float] = args[ + "caption_probs" + ] # probabilities for user/short/medium/long captions + self.caption_style_mapping = { + "long": self.caption_type, + "short": f"{self.caption_type}_short", + "medium": f"{self.caption_type}_medium", + "user": f"{self.caption_type}_user", + } + assert self.caption_probs.keys() == self.caption_style_mapping.keys(), ( + "The keys for caption_probs, caption_style_mapping, and embedding_style_mapping should match" + ) + + if self._load_embeddings: + assert self.caption_style_mapping.keys() == self.embedding_style_mapping.keys(), ( + "The keys for caption_style_mapping and embedding_style_mapping should match" + ) + + def __call__(self, data_dict: dict) -> dict: + r"""Performs text transformation. + + Args: + data_dict (dict): Input data dict + Returns: + data_dict (dict): Output dict with captions and t5 embeddings added + """ + + try: + windows = data_dict[self.captions_key][self.caption_windows_key] + n_windows = len(windows) + chunk_index = data_dict["chunk_index"] + + if chunk_index == n_windows: + # This will only happen when the number of captions does not match number of chunks due to re-transcoding the videos. + log.warning( + f"TextTransform dataloader error: Found {data_dict['n_orig_video_frames']} in video but captioning is done with videos of {windows[-1]['end_frame']} frames. This mismatch is due to video re-transcoding.", + rank0_only=False, + ) + chunk_index -= 1 + + selected_caption_window = windows[chunk_index] + except Exception as e: + log.warning( + f"TextTransform dataloader error -- url: {data_dict['__url__']}, key: {data_dict['__key__']}, chunk_index: {data_dict['chunk_index']}\n error {e}", + rank0_only=False, + ) + return None + + sampled_caption_style = None + try: + available_caption_styles = [] + for k in selected_caption_window.keys(): + caption_style = k.replace(self.caption_type, "").replace("_", "") + if caption_style == "": # it is long caption by default + available_caption_styles.append("long") + elif caption_style in self.caption_style_mapping: + available_caption_styles.append(caption_style) + else: + assert caption_style in ["startframe", "endframe"], f"Unsupported caption_type {caption_style}" + + probabilities_for_available_caption_styles = { + k: v for k, v in self.caption_probs.items() if k in available_caption_styles + } + sampled_caption_style = random.choices( + list(probabilities_for_available_caption_styles), + weights=probabilities_for_available_caption_styles.values(), + )[0] + data_dict["ai_caption"] = selected_caption_window[self.caption_style_mapping[sampled_caption_style]] + except Exception as e: + log.warning( + f"TextTransform dataloader error -- url: {data_dict['__url__']}, key: {data_dict['__key__']}, selected_caption_window: {selected_caption_window}\n error {e}", + rank0_only=False, + ) + return None + if data_dict["ai_caption"] == "": + log.warning( + f"TextTransform dataloader error -- empty caption! url: {data_dict['__url__']}, key: {data_dict['__key__']}, selected_caption_window: {selected_caption_window}", + rank0_only=False, + ) + return None + + assert data_dict["ai_caption"] is not None and sampled_caption_style is not None + data_dict["sampled_caption_style"] = sampled_caption_style + + del data_dict[self.captions_key] # delete the field as we have extracted ai_caption from it + + if self._load_embeddings: + ai_caption_embedding_data = data_dict[self.embeddings_key] + try: + if self.embedding_caption_type == "vila_caption": + t5_embedding = ai_caption_embedding_data[chunk_index] + else: + t5_embedding = ai_caption_embedding_data[chunk_index][ + self.embedding_style_mapping[sampled_caption_style] + ] + except Exception as e: + log.warning( + f"TextTransform dataloader error -- url: {data_dict['__url__']}, key: {data_dict['__key__']}, chunk_index: {data_dict['chunk_index']}, n embeddings: {len(ai_caption_embedding_data)}, n captions: {n_windows} \n error {e}", + rank0_only=False, + ) + return None + out_t5, out_t5_mask = pad_and_resize( + t5_embedding, + self.t5_tokens_num, + is_mask_all_ones=self.is_mask_all_ones, + ) + data_dict["t5_text_embeddings"] = out_t5 + data_dict["t5_text_mask"] = out_t5_mask + del data_dict[self.embeddings_key] # delete the field as we have extracted t5 embedding from it + + return data_dict diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/video_parsing.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/video_parsing.py new file mode 100644 index 0000000000000000000000000000000000000000..bae3aca49392dd3f440de521a6b1f82d15e8e408 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/augmentors/video_parsing.py @@ -0,0 +1,200 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import io +import random +from typing import Optional + +import decord +import numpy as np +import torch +from einops import rearrange +from torchvision.transforms.v2 import UniformTemporalSubsample + +from cosmos_policy._src.imaginaire.datasets.webdataset.augmentors.augmentor import Augmentor +from cosmos_policy._src.imaginaire.utils import log + + +class VideoParsing(Augmentor): + """ + This augmentor is used to parse the video bytes and get the video frames. + the return dict is back-compatible with old datasets, which video decoding happens in the decoder stage. + """ + + def __init__(self, input_keys: list, output_keys: Optional[list] = None, args: Optional[dict] = None) -> None: + super().__init__(input_keys, output_keys, args) + assert len(input_keys) == 2, "VideoParsing augmentor only supports two input keys" + self.meta_key = input_keys[0] + self.video_key = input_keys[1] + + self.key_for_caption = args["key_for_caption"] + assert self.key_for_caption in [ + "t2w_windows", + "i2w_windows_later_frames", + ], "key_for_caption must be either t2w_windows or i2w_windows_later_frames" + self.min_duration = args["min_duration"] + self.min_fps = args["min_fps"] + self.max_fps = args["max_fps"] + self.num_frames = args["num_video_frames"] + self.use_native_fps = args["use_native_fps"] # orginal fps if (total_frames // self.num_frames == 1). + # two variables for frame interpolation to subsample the video frames to different fps. + self.use_random_consecutive_frames = args.get("use_random_consecutive_frames", False) + self.use_random_interleaved_frames = args.get("use_random_interleaved_frames", False) + # a list of allowed num_multiplers (how many frames are skipped) + # default is 1 - 100 which allows virtually any num_multipler possible + self.allowed_num_multiplers = args.get("allowed_num_multiplers", list(range(1, 100))) + log.info(f"allowed_num_multiplers in video_parsing with use_native_fps: {self.allowed_num_multiplers}") + self.use_original_fps = args["use_original_fps"] # use original fps without sampling + if self.use_native_fps or self.use_original_fps: + assert self.num_frames > 0, "num_frames must be greater than 0 when use_native_fps is True" + if self.num_frames > 0: + self.sampler = UniformTemporalSubsample(self.num_frames) + self.video_decode_num_threads = args["video_decode_num_threads"] + + def __call__(self, data_dict: dict) -> dict | None: + try: + meta_dict = data_dict[self.meta_key] + video = data_dict[self.video_key] + except Exception: + log.warning( + f"Cannot find video. url: {data_dict['__url__']}, key: {data_dict['__key__']}", rank0_only=False + ) + return None + + if not isinstance(video, bytes): + return data_dict + + video_info = { + "fps": meta_dict["framerate"], + "n_orig_video_frames": meta_dict["nb_frames"], + } + + if video_info["fps"] < self.min_fps: + log.warning(f"Video FPS {video_info['fps']} is less than min_fps {self.min_fps}", rank0_only=False) + return None + if video_info["fps"] > self.max_fps: + log.warning(f"Video FPS {video_info['fps']} is greater than max_fps {self.max_fps}", rank0_only=False) + return None + + options: list = list((i, item) for i, item in enumerate(meta_dict[self.key_for_caption])) + + # Skip the last window if possible. + # All windows except the last are 5 seconds long. The last window has a duration in the range [2.5s, 7.5), which is less preferred. + if len(options) > 1: + options = options[:-1] + + # shuffle options + random.shuffle(options) + video_frames = None + for chunk_index, option in options: + start_frame = option["start_frame"] + end_frame = option["end_frame"] + if (end_frame - start_frame) < self.min_duration * video_info["fps"]: + continue + + if self.use_native_fps or self.use_original_fps: + if (end_frame - start_frame) < self.num_frames: + continue + + video_buffer = io.BytesIO(video) + video_reader = decord.VideoReader(video_buffer, num_threads=self.video_decode_num_threads) + + if self.use_native_fps or self.use_original_fps: + if "alpamayo" in data_dict["__url__"].root: + start_frame += 5 + if (end_frame - start_frame) < self.num_frames: + continue + if self.use_random_consecutive_frames: + # Random consecutive sampling for frame interpolation + total_frames = end_frame - start_frame + max_start_idx = total_frames - self.num_frames + random_offset = random.randint(0, max_start_idx) + _start_frame = start_frame + random_offset + _end_frame = _start_frame + self.num_frames + frame_indices = np.arange(_start_frame, _end_frame).tolist() + + assert len(frame_indices) == self.num_frames, ( + f"frame_indices length {len(frame_indices)} should be == {self.num_frames}" + ) + elif self.use_random_interleaved_frames: + # Random interleaved sampling for fractional frame interpolation (e.g. 24->30fps) + total_frames = end_frame - start_frame + max_start_idx = total_frames - self.num_frames + random_offset = random.randint(0, max_start_idx) + _start_frame = start_frame + random_offset + _end_frame = _start_frame + self.num_frames + frame_indices = sorted( + np.arange(_start_frame, _end_frame, 4).tolist() + + np.arange(_start_frame, _end_frame, 5).tolist() + )[1:-1] # remove duplicate first and last frame + else: + # take mid self.num_frames frames from start frame to end frame. + total_frames = end_frame - start_frame + # always try lower fps if possible. + if self.use_native_fps: + num_multiplier = total_frames // self.num_frames + if num_multiplier not in self.allowed_num_multiplers: + continue + else: # self.use_original_fps + num_multiplier = 1 + expected_length = self.num_frames * num_multiplier + _start_frame = start_frame + (total_frames - expected_length) // 2 + _end_frame = _start_frame + expected_length + frame_indices = np.arange(_start_frame, _end_frame, num_multiplier).tolist() + assert len(frame_indices) == self.num_frames, "frame_indices length is not equal to num_frames" + video_frames = video_reader.get_batch(frame_indices).asnumpy() + video_frames = torch.from_numpy(video_frames).permute(3, 0, 1, 2) # (T, H, W, C) -> (C, T, H, W) + video_reader.seek(0) # set video reader point back to 0 to clean up cache + del video_reader # delete the reader to avoid memory leak + break + + else: + frame_indices = np.arange(start_frame, end_frame).tolist() + + # online hot-fix for alpamayo data. Skip the first 5 frames as there is chance that the first five frames contain black frames. + if "alpamayo" in data_dict["__url__"].root: + assert len(frame_indices) >= 5, ( + "Getting less than 5 frames for alpamayo videos. There is no way to skip the first five frames." + ) + frame_indices = frame_indices[5:] + start_frame += 5 + video_frames = video_reader.get_batch(frame_indices).asnumpy() + + video_frames = torch.from_numpy(video_frames).permute(3, 0, 1, 2) # (T, H, W, C) -> (C, T, H, W) + + # Clean up + video_reader.seek(0) # set video reader point back to 0 to clean up cache + del video_reader # delete the reader to avoid memory leak + break + + if video_frames is None: + log.warning( + f"No valid video frames found, return None. url: {data_dict['__url__']}, key: {data_dict['__key__']}", + rank0_only=False, + ) + return None + + video_info["chunk_index"] = chunk_index + video_info["frame_start"] = start_frame + video_info["frame_end"] = end_frame + video_info["num_frames"] = end_frame - start_frame # type: ignore + if self.num_frames > 0 and not (self.use_native_fps or self.use_original_fps): + video_frames = rearrange(self.sampler(rearrange(video_frames, "c t h w -> t c h w")), "t c h w -> c t h w") + video_info["video"] = video_frames + + # update data_dict, make it back-compatible with old datasets, which video decoding happens in the decoder stage. + data_dict[self.video_key] = video_info + + return data_dict diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/cached_replay_dataloader.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/cached_replay_dataloader.py new file mode 100644 index 0000000000000000000000000000000000000000..5ed670b8d53fd50e31dc169251a4a5e5953eb8a2 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/cached_replay_dataloader.py @@ -0,0 +1,510 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import copy +import threading +import traceback +from typing import Callable, Dict, Iterator, List, Optional + +import numpy as np +import torch +from torch.utils.data import DataLoader + +from cosmos_policy._src.predict2.datasets.watchdog import OperationWatchdog + + +def generate_multiple_image_batches(data_batch, nh, nw, output_height, output_width): + """ + Generate (nh * nw) data batches, each with images randomly cropped from the original. + + Args: + data_batch: Input batch containing images and other elements + nh: Number of replications in height direction + nw: Number of replications in width direction + output_height: Target image height + output_width: Target image width + + Returns: + List of data batches with randomly cropped images + """ + # Access the original image tensor + original_images = data_batch["images"] + + # Get the original image dimensions + B, C, H, W = original_images.shape + + # Check if output dimensions are valid + if output_height > H or output_width > W: + raise ValueError("Output dimensions cannot be larger than original dimensions") + + # Calculate step sizes for uniform sampling with minimal overlap + h_step = max(1, (H - output_height) // max(1, nh - 1)) if nh > 1 else 0 + w_step = max(1, (W - output_width) // max(1, nw - 1)) if nw > 1 else 0 + + # Initialize list to store all created batches + all_batches = [] + + # Generate crops + for h_idx in range(nh): + for w_idx in range(nw): + # Create a deep copy of the original data batch + new_batch = copy.deepcopy(data_batch) + + # Calculate starting positions for this crop + # Add some randomness within each grid cell to avoid exact same positions + h_start = min(H - output_height, h_idx * h_step + torch.randint(0, max(1, h_step), (1,)).item()) + w_start = min(W - output_width, w_idx * w_step + torch.randint(0, max(1, w_step), (1,)).item()) + + # For the edge case where there's only one crop position, center the crop + if nh == 1: + h_start = (H - output_height) // 2 + if nw == 1: + w_start = (W - output_width) // 2 + + # Crop the image + new_batch["images"] = original_images[ + :, + :, + h_start : h_start + output_height, + w_start : w_start + output_width, + ] + + # Add to our list of batches + all_batches.append(new_batch) + + return all_batches + + +def generate_multiple_video_batches(data_batch, nt, nh, nw, output_length, output_height, output_width): + """ + Generate (nt * nh * nw) data batches, each with videos randomly cropped from the original. + + Args: + data_batch: Input batch containing videos and other elements + nt: Number of replications in temporal (length) direction + nh: Number of replications in height direction + nw: Number of replications in width direction + output_length: Target video length (temporal dimension) + output_height: Target video height + output_width: Target video width + + Returns: + List of data batches with randomly cropped videos + """ + # Access the original video tensor + original_video = data_batch["video"] + + # Get the original video dimensions + B, C, T, H, W = original_video.shape + + # Check if output dimensions are valid + if output_length > T or output_height > H or output_width > W: + raise ValueError("Output dimensions cannot be larger than original dimensions") + + # Calculate step sizes for uniform sampling with minimal overlap + t_step = max(1, (T - output_length) // max(1, nt - 1)) if nt > 1 else 0 + h_step = max(1, (H - output_height) // max(1, nh - 1)) if nh > 1 else 0 + w_step = max(1, (W - output_width) // max(1, nw - 1)) if nw > 1 else 0 + + # Initialize list to store all created batches + all_batches = [] + + # Generate crops + for t_idx in range(nt): + for h_idx in range(nh): + for w_idx in range(nw): + # Create a deep copy of the original data batch + new_batch = copy.deepcopy(data_batch) + + # Calculate starting positions for this crop + # Add some randomness within each grid cell to avoid exact same positions + t_start = min(T - output_length, t_idx * t_step + torch.randint(0, max(1, t_step), (1,)).item()) + h_start = min(H - output_height, h_idx * h_step + torch.randint(0, max(1, h_step), (1,)).item()) + w_start = min(W - output_width, w_idx * w_step + torch.randint(0, max(1, w_step), (1,)).item()) + + # For the edge case where there's only one crop position, center the crop + if nt == 1: + t_start = (T - output_length) // 2 + if nh == 1: + h_start = (H - output_height) // 2 + if nw == 1: + w_start = (W - output_width) // 2 + + # Crop the video + new_batch["video"] = original_video[ + :, + :, + t_start : t_start + output_length, + h_start : h_start + output_height, + w_start : w_start + output_width, + ] + + # Add to our list of batches + all_batches.append(new_batch) + + return all_batches + + +def duplicate_batches(data_batch, n: int) -> List[Dict]: + """ + Duplicate a data batch n times. + """ + return [copy.deepcopy(data_batch) for _ in range(n)] + + +_RNG = np.random.default_rng(123) + + +def duplicate_batches_random(data_batch, n: float) -> List[Dict]: + floor = int(np.floor(n)) + ceil = int(np.ceil(n)) + # generate a random number uniformly from [floor, ceil) + random_number = _RNG.uniform(floor, ceil) + if random_number > n: + return [copy.deepcopy(data_batch) for _ in range(floor)] + else: + return [copy.deepcopy(data_batch) for _ in range(ceil)] + + +def concatenate_batches(n: int, data_batches: List[Dict]) -> List[Dict]: + """ + Smartly concatenate n input data batches into m output data batches. + Each data batch is a dictionary with values that can be torch tensor, string, or list. + + Args: + n (int): Number of input batches to process per output batch + data_batches (list): List of dictionary data batches + + Returns: + list: List of concatenated data batches + """ + if n <= 0: + raise ValueError("n must be a positive integer") + + # Calculate m based on the input + total_batches = len(data_batches) + if total_batches % n != 0: + raise ValueError(f"Length of data_batches ({total_batches}) must be divisible by n ({n})") + + m = total_batches // n + + # Initialize output batches + output_batches = [] + + # Process in groups of n + for i in range(m): + # Get the corresponding batch from each group + batches_to_concat = [] + for j in range(n): + batch_idx = j * m + i + batches_to_concat.append(data_batches[batch_idx]) + + # Create a new merged dictionary + merged_batch = {} + + # Get all unique keys from the dictionaries + all_keys = set() + for batch in batches_to_concat: + all_keys.update(batch.keys()) + + # Process each key + for key in all_keys: + # Collect values for this key from all batches + values = [] + for batch in batches_to_concat: + if key in batch: + values.append(batch[key]) + + if not values: + continue + + # Determine the type of the first non-None value + first_value = next((v for v in values if v is not None), None) + if first_value is None: + merged_batch[key] = None + continue + + # Handle different types + if isinstance(first_value, torch.Tensor): + # Assuming this is a tensor-like object with cat method (e.g., torch.Tensor) + merged_batch[key] = torch.cat(values, dim=0) + elif isinstance(first_value, str): + merged_batch[key] = values[0] + elif isinstance(first_value, list): + # Extend lists + merged_list = [] + for v in values: + merged_list.extend(v) + merged_batch[key] = merged_list + else: + # For other types, just use the list of values + merged_batch[key] = values + + output_batches.append(merged_batch) + + return output_batches + + +class CachedReplayDataLoader: + """A DataLoader wrapper that asynchronously caches and replays data batches to + mitigate slow loading issues. Assumes the underlying DataLoader is infinite. + + This class delegates all augmentation logic to an external augmentation function, + which takes a batch from the data loader and returns multiple augmented versions. + The class handles caching these augmented batches and optionally concatenating + them when yielded. + + Attributes: + data_loader (DataLoader): The underlying infinite DataLoader. + cache_size (int): Maximum number of augmented batches to store in the cache. + cache_augmentation_fn (Callable): Function to create multiple augmented versions of each batch. + concat_size (int): Number of batches to concatenate when yielding from the iterator. + rng (numpy.random.Generator): Controlled random number generator for deterministic behavior. + """ + + def __init__( + self, + data_loader: DataLoader, + cache_size: int, + cache_augmentation_fn: Callable[[Dict], List[Dict]], + concat_size: int = 1, + name: str = "cached_replay_dataloader", + ) -> None: + """Initialize the CachedReplayDataLoader. + + Args: + data_loader (DataLoader): The infinite DataLoader to fetch data batches from. + cache_size (int): Maximum number of augmented data batches to store in the cache. + cache_augmentation_fn (Callable[[Dict], List[Dict]]): Function that takes a batch and returns + a list of augmented batches. + concat_size (int, optional): Number of batches to concatenate when yielding. Defaults to 1. + """ + self.data_loader = data_loader + self.cache_size = cache_size + self.cache_augmentation_fn = cache_augmentation_fn + self.concat_size = concat_size + + # Create controlled random number generator for deterministic behavior + self.rng = np.random.default_rng(123) + + # Create an iterator over the infinite DataLoader. + self._data_iter: Iterator = iter(self.data_loader) + # Internal cache to store augmented batches. + self._cache: List[Dict] = [] + # Condition variable to manage cache access. + self._cache_cond = threading.Condition() + # Event to signal the background thread to stop. + self._stop_event = threading.Event() + # Store exceptions from the background thread + self._prefetch_exception = None + + self._watchdog = OperationWatchdog(warning_threshold=100, verbose_interval=600, name=name) + self._prefetch_thread = threading.Thread( + target=self._prefetch_loop, daemon=True, name=f"{name}_prefetch_thread" + ) + self._prefetch_thread.start() + + def _prefetch_loop(self) -> None: + """Continuously fetch batches from the DataLoader, augment them, and store in the cache. + + If the cache is full (reaches `cache_size`), this loop waits until space is available. + Catches exceptions and stores them for later propagation to the main thread. + """ + try: + while not self._stop_event.is_set(): + try: + with self._watchdog.watch("fetch raw batch", verbose_first_n=5): + batch = next(self._data_iter) + except Exception as e: + # Capture DataLoader errors + self._set_exception(e, "Error fetching batch from DataLoader") + break + + try: + # Apply augmentation function to generate multiple augmented batches + with self._watchdog.watch("augmentation", verbose_first_n=5): + augmented_batches = self.cache_augmentation_fn(batch) + except Exception as e: + # Capture augmentation function errors + self._set_exception(e, "Error in augmentation function") + break + + try: + # Use controlled random generator for shuffling + permutation = self.rng.permutation(len(augmented_batches)) + augmented_batches = [augmented_batches[i] for i in permutation] + + for aug_batch in augmented_batches: + with self._cache_cond: + while len(self._cache) >= self.cache_size and not self._stop_event.is_set(): + self._cache_cond.wait(timeout=1.0) + if self._stop_event.is_set(): + break + self._cache.append(aug_batch) + self._cache_cond.notify_all() + except Exception as e: + # Capture other errors during caching + self._set_exception(e, "Error adding batch to cache") + break + except Exception as e: + # Catch any other unforeseen errors + self._set_exception(e, "Unexpected error in prefetch thread") + + def _set_exception(self, exception: Exception, context: str = "") -> None: + """Store an exception from the background thread with context information. + + Args: + exception (Exception): The exception that was raised + context (str, optional): Additional context about where the error occurred + """ + error_info = f"{context}: {str(exception)}\n{traceback.format_exc()}" + with self._cache_cond: + self._prefetch_exception = RuntimeError(error_info) + self._cache_cond.notify_all() # Wake up any waiting threads + + def _check_for_errors(self) -> None: + """Check if the background thread has encountered an error and raise it if so.""" + if self._prefetch_exception is not None: + raise self._prefetch_exception + + def __iter__(self) -> Iterator[Dict]: + """Yield augmented data batches from the cache, optionally concatenated based on concat_size. + + This method starts the background prefetch thread if it hasn't been started yet. + If concat_size > 1, it collects multiple batches and concatenates them. + + Raises: + RuntimeError: If the background thread encountered an error + """ + while not self._stop_event.is_set(): + if self.concat_size <= 1: + # Simple case: yield single batches + with self._watchdog.watch("main thread fetch single batch", verbose_first_n=5): + with self._cache_cond: + while not self._cache and not self._stop_event.is_set() and self._prefetch_exception is None: + self._cache_cond.wait(timeout=1.0) # Add timeout to periodically check for errors + + # Check for errors before proceeding + self._check_for_errors() + + if self._stop_event.is_set(): + break + + if not self._cache: # If cache is still empty after timeout + continue + + # Use controlled random generator to select batch index + idx = self.rng.integers(0, len(self._cache)) + batch = self._cache.pop(idx) + self._cache_cond.notify_all() + yield batch + else: + # Collect concat_size batches and concatenate them + with self._watchdog.watch("main thread fetch smaples", verbose_first_n=5): + collected_batches = [] + for _ in range(self.concat_size): + with self._cache_cond: + while ( + not self._cache and not self._stop_event.is_set() and self._prefetch_exception is None + ): + self._cache_cond.wait(timeout=1.0) # Add timeout to periodically check for errors + + # Check for errors before proceeding + self._check_for_errors() + + if self._stop_event.is_set(): + break + + if not self._cache: # If cache is still empty after timeout + continue + + # Use controlled random generator to select batch index + idx = self.rng.integers(0, len(self._cache)) + batch = self._cache.pop(idx) + self._cache_cond.notify_all() + collected_batches.append(batch) + + if self._stop_event.is_set(): + break + + if not collected_batches: + continue + + if len(collected_batches) < self.concat_size: + # Not enough batches collected, just concatenate the ones we have + concat_batches = concatenate_batches(len(collected_batches), collected_batches) + yield concat_batches[0] + else: + # Concatenate the collected batches + try: + concat_batches = concatenate_batches(self.concat_size, collected_batches) + yield concat_batches[0] + except Exception as e: + # Handle errors in batch concatenation + raise RuntimeError(f"Error concatenating batches: {str(e)}") from e + + def __len__(self) -> int: + """Return the length of the underlying DataLoader.""" + return len(self.data_loader) + + def close(self) -> None: + """Stop the prefetch thread and clear the cache. + Also checks for any errors in the background thread and raises them. + """ + self._stop_event.set() + with self._cache_cond: + self._cache_cond.notify_all() + if self._prefetch_thread is not None: + self._prefetch_thread.join(timeout=5.0) # Add timeout to avoid hanging on thread join + with self._cache_cond: + self._cache.clear() + + # Check and propagate any errors from the background thread + self._check_for_errors() + + +def get_cached_replay_dataloader( + use_cache: bool = False, + cache_size: int = 32, + concat_size: int = 1, + cache_augment_fn: Optional[Callable] = None, + cache_replay_name: str = "cached_replay_dataloader", + webdataset: bool = True, + **kwargs, +): + if webdataset: + from cosmos_policy._src.imaginaire.datasets.webdataset.dataloader import DataLoader as _DataLoader + else: + from torch.utils.data import DataLoader as _DataLoader + + if not use_cache: + return _DataLoader(**kwargs) + + expected_batch_size = kwargs["batch_size"] + assert expected_batch_size % concat_size == 0, ( + f"Batch size {expected_batch_size} must be divisible by concat_size {concat_size}" + ) + kwargs["batch_size"] = expected_batch_size // concat_size + + dataloader = _DataLoader(**kwargs) + + # wrapper it with cached replay dataloader + return CachedReplayDataLoader( + data_loader=dataloader, + cache_size=cache_size, + concat_size=concat_size, + cache_augmentation_fn=cache_augment_fn, + name=cache_replay_name, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/cached_replay_dataloader_test.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/cached_replay_dataloader_test.py new file mode 100644 index 0000000000000000000000000000000000000000..4d309fc293c0d3a21692c4d51e2fd69f6239416c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/cached_replay_dataloader_test.py @@ -0,0 +1,166 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Dict, List + +import pytest +import torch +from torch.utils.data import DataLoader, Dataset + +from cosmos_policy._src.predict2.datasets.cached_replay_dataloader import CachedReplayDataLoader, concatenate_batches + + +class DummyDataset(Dataset): + """A dummy dataset that returns a dictionary with a 'videos' tensor of shape (1, 3, frames_per_batch). + + Each item is a tensor with sequential values and a batch-specific offset for uniqueness. + """ + + def __init__(self, num_batches: int, frames_per_batch: int) -> None: + self.num_batches = num_batches + self.frames_per_batch = frames_per_batch + + def __len__(self) -> int: + return 999 + + def __getitem__(self, index: int) -> Dict: + # Create a tensor with shape (1, 3, frames_per_batch) with sequential values. + video = torch.arange(self.frames_per_batch, dtype=torch.float32).unsqueeze(0).repeat(3, 1) + video = video + index * 1000 # Offset each batch for uniqueness. + return {"videos": video} + + +def temporal_slice_augmentation(batch: Dict) -> List[Dict]: + """Augmentation function that creates multiple temporal slice variants. + + This simulates the original CachedReplayDataLoader behavior but as an external function. + """ + videos = batch["videos"] + total_frames = videos.shape[2] + num_video_frames = 20 # Number of frames per slice + replay_num = 5 # Number of slices to create + + if total_frames < num_video_frames: + raise ValueError(f"Total frames ({total_frames}) is less than required frames ({num_video_frames}).") + + # Compute evenly spaced starting offsets along the T dimension + if replay_num == 1: + offsets = [0] + else: + max_start = total_frames - num_video_frames + offsets = [int(round(i * max_start / (replay_num - 1))) for i in range(replay_num)] + + # Create clones with different temporal slices + augmented_batches = [] + for offset in offsets: + clone = {k: v.clone() if isinstance(v, torch.Tensor) else v for k, v in batch.items()} + # Slice the video tensor along T dimension (index 2) + clone["videos"] = videos[:, :, offset : offset + num_video_frames, ...] + augmented_batches.append(clone) + + return augmented_batches + + +def brightness_augmentation(batch: Dict) -> List[Dict]: + """Augmentation function that creates variants with different brightness levels.""" + videos = batch["videos"] + scales = [0.8, 1.0, 1.2] # Brightness adjustment factors + + augmented_batches = [] + for scale in scales: + clone = {k: v.clone() if isinstance(v, torch.Tensor) else v for k, v in batch.items()} + clone["videos"] = videos * scale + augmented_batches.append(clone) + + return augmented_batches + + +@pytest.mark.L1 +def test_augmentation(): + """Test that the dataloader correctly uses an external augmentation function.""" + total_frames = 100 + dataset = DummyDataset(num_batches=1, frames_per_batch=total_frames) + data_loader = DataLoader(dataset, batch_size=1) + + cached_loader = CachedReplayDataLoader( + data_loader, + cache_size=10, + cache_augmentation_fn=temporal_slice_augmentation, + ) + + # Collect augmented batches + augmented_batches = [] + for batch in cached_loader: + augmented_batches.append(batch) + if len(augmented_batches) >= 5: # We expect 5 variants from our augmentation function + break + + # Verify each augmented batch has the correct shape + for batch in augmented_batches: + video = batch["videos"] + # Expected shape after temporal slicing (1, 3, 20) + assert video.shape[0] == 1 + assert video.shape[1] == 3 + assert video.shape[2] == 20 + + # Check that the variants cover different parts of the video + offsets = [int(video[0, 0, 0].item() % 1000) for batch in augmented_batches for video in [batch["videos"]]] + assert len(set(offsets)) > 1 # Should have different starting offsets + + cached_loader.close() + + +@pytest.mark.L1 +def test_batch_concatenation(): + """Test that batch concatenation works correctly.""" + total_frames = 80 + concat_size = 2 + + dataset = DummyDataset(num_batches=10, frames_per_batch=total_frames) + data_loader = DataLoader(dataset, batch_size=1) + + cached_loader = CachedReplayDataLoader( + data_loader, + cache_size=10, + cache_augmentation_fn=brightness_augmentation, + concat_size=concat_size, + ) + + # Collect a few batches and check their shapes + for i, batch in enumerate(cached_loader): + # Should have batch_size batches concatenated along dim 0 + assert batch["videos"].shape[0] == concat_size + + if i >= 3: + break + + cached_loader.close() + + +@pytest.mark.L1 +def test_external_concatenate_batches(): + """Test the concatenate_batches function separately.""" + # Create sample batches with tensors + batch1 = {"videos": torch.ones(1, 3, 10), "labels": torch.tensor([1])} + batch2 = {"videos": torch.zeros(1, 3, 10), "labels": torch.tensor([0])} + + # Test concatenation + result = concatenate_batches(1, [batch1, batch2]) + assert len(result) == 2 + + result = concatenate_batches(2, [batch1, batch2]) + assert len(result) == 1 + assert result[0]["videos"].shape[0] == 2 + assert result[0]["labels"].shape[0] == 2 diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/data_sources/data_registration.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/data_sources/data_registration.py new file mode 100644 index 0000000000000000000000000000000000000000..edd6f567ea7904d7f6285ef986518b49b012c003 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/data_sources/data_registration.py @@ -0,0 +1,40 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Dataset registration for cosmos datasets with support for different caption types. +""" + +from cosmos_policy._src.imaginaire.utils import log + +DATASET_OPTIONS = {} + + +# embeddings are packed together. Need to clean data to reduce entropy. +_CAPTION_EMBEDDING_KEY_MAPPING_IMAGES = { + "ai_v3p1": "ai_v3p1", + "qwen2p5_7b_v4": "qwen2p5_7b_v4", + "prompts": "qwen2p5_7b_v4", +} + + +def dataset_register(key): + log.info(f"registering dataset {key}") + + def decorator(func): + DATASET_OPTIONS[key] = func + return func + + return decorator diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/data_sources/item_datasets_for_validation.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/data_sources/item_datasets_for_validation.py new file mode 100644 index 0000000000000000000000000000000000000000..8bb521bf6a500dfe1daaf592991e094b15a97e5c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/data_sources/item_datasets_for_validation.py @@ -0,0 +1,53 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os + +from cosmos_policy._src.predict2.datasets.item_dataset import ItemDatasetConfig + + +def get_itemdataset_option(name: str, text_embedding_type: str = "t5_xxl") -> ItemDatasetConfig: + item_dataset_config = ITEMDATASET_OPTIONS[name] + + if text_embedding_type != "t5_xxl": + # For all datasets other than T5_XXL, we save the dataset in the following path + # {data_root}/ablation_text_embeddings/{text_embedding_type}/{dataset_name} + dataset_path = item_dataset_config.path + dataset_path_split = dataset_path.split("/") + is_file = os.path.splitext(dataset_path)[1] != "" + + if is_file: + # In case of a file, we have + # {data_root}/ablation_text_embeddings/{text_embedding_type}/{dataset_name}/{filename.ext} + new_dataset_path = ( + dataset_path_split[0:-2] + + ["ablation_text_embeddings", f"{text_embedding_type}"] + + dataset_path_split[-2:] + ) + else: + new_dataset_path = ( + dataset_path_split[0:-1] + + ["ablation_text_embeddings", f"{text_embedding_type}"] + + dataset_path_split[-1:] + ) + + new_dataset_path = "/".join(new_dataset_path) + + return ItemDatasetConfig(path=new_dataset_path, length=item_dataset_config.length) + return item_dataset_config + + +# length must % 8 =0 to avoid mysterious hang bug of fsdp+CP!It is tested with cp4. +ITEMDATASET_OPTIONS = {} diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/data_sources/mock_data.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/data_sources/mock_data.py new file mode 100644 index 0000000000000000000000000000000000000000..e33300b933e3948502542a73b9b6c77be5763c81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/data_sources/mock_data.py @@ -0,0 +1,134 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Mock data for fast testing and debugging. See `projects/cosmos/diffusion/v1/datasets/mock_data_test.py` for usage. +""" + +from functools import partial + +import torch + +from cosmos_policy._src.imaginaire.datasets.mock_dataset import CombinedDictDataset, LambdaDataset +from cosmos_policy._src.predict2.datasets.utils import IMAGE_RES_SIZE_INFO, VIDEO_RES_SIZE_INFO + + +def get_image_dataset( + resolution: str = "512", + len_t5: int = 512, + t5_dim: int = 1024, + **kwargs, +): + h, w = IMAGE_RES_SIZE_INFO[resolution]["9,16"] + del kwargs + return CombinedDictDataset( + **{ + "images": LambdaDataset(partial(torch.randn, size=(3, h, w))), + "t5_text_embeddings": LambdaDataset(partial(torch.randn, size=(len_t5, t5_dim))), + "t5_text_mask": LambdaDataset(partial(torch.randint, low=0, high=2, size=(len_t5,), dtype=torch.int64)), + "fps": LambdaDataset(lambda: 1.0), + "image_size": LambdaDataset(partial(torch.tensor, [h, w, h, w], dtype=torch.float32)), + "num_frames": LambdaDataset(lambda: 1), + "padding_mask": LambdaDataset(partial(torch.zeros, size=(1, h, w))), + "dataset_name": LambdaDataset(lambda: "image_data"), + "raw_captions": LambdaDataset(lambda: "placeholder"), + "ai_caption": LambdaDataset(lambda: "placeholder"), # The text encoder augmentation method uses this field + "__url__": LambdaDataset(lambda: "placeholder"), + "__key__": LambdaDataset(lambda: "placeholder"), + } + ) + + +def get_video_dataset( + num_video_frames: int, + resolution: str = "512", + len_t5: int = 512, + t5_dim: int = 1024, + **kwargs, +): + del kwargs + h, w = VIDEO_RES_SIZE_INFO[resolution]["9,16"] + + def video_fn(): + return torch.randint(0, 255, size=(3, num_video_frames, h, w)).to(dtype=torch.uint8) + + return CombinedDictDataset( + **{ + "video": LambdaDataset(video_fn), + "t5_text_embeddings": LambdaDataset(partial(torch.randn, size=(len_t5, t5_dim))), + "t5_text_mask": LambdaDataset(partial(torch.randint, low=0, high=2, size=(len_t5,), dtype=torch.int64)), + "fps": LambdaDataset(lambda: 24.0), + "image_size": LambdaDataset(partial(torch.tensor, [h, w, h, w], dtype=torch.float32)), + "num_frames": LambdaDataset(lambda: num_video_frames), + "padding_mask": LambdaDataset(partial(torch.zeros, size=(1, h, w))), + "ai_caption": LambdaDataset(lambda: "placeholder"), + "dataset_name": LambdaDataset(lambda: "video_data"), + "chunk_index": LambdaDataset(lambda: 0), + "frame_end": LambdaDataset(lambda: 0), + "frame_start": LambdaDataset(lambda: 0), + "n_orig_video_frames": LambdaDataset(lambda: 0), + "__url__": LambdaDataset(lambda: "placeholder"), + "__key__": LambdaDataset(lambda: "placeholder"), + } + ) + + +def get_video_ctrlnet_dataset( + h: int, + w: int, + num_video_frames: int, + len_t5: int = 512, + hint_key: str = "control_input_canny", + **kwargs, +): + del kwargs + + def video_fn(): + return torch.randint(0, 255, size=(3, num_video_frames, h, w)).to(dtype=torch.uint8) + + return CombinedDictDataset( + **{ + "video": LambdaDataset(video_fn), + hint_key: LambdaDataset(video_fn), + "t5_text_embeddings": LambdaDataset(partial(torch.randn, size=(len_t5, 1024), dtype=torch.bfloat16)), + "t5_text_mask": LambdaDataset(partial(torch.randint, low=0, high=2, size=(len_t5,), dtype=torch.int64)), + "fps": LambdaDataset(lambda: 24.0), + "image_size": LambdaDataset(partial(torch.tensor, [h, w, h, w], dtype=torch.float32)), + "num_frames": LambdaDataset(lambda: num_video_frames), + "padding_mask": LambdaDataset(partial(torch.zeros, size=(1, h, w), dtype=torch.bfloat16)), + } + ) + + +def get_image_ctrlnet_dataset( + h: int, + w: int, + len_t5: int = 512, + hint_key: str = "control_input_canny", + **kwargs, +): + del kwargs + return CombinedDictDataset( + **{ + "images": LambdaDataset(partial(torch.randn, size=(3, h, w))), + hint_key: LambdaDataset(partial(torch.randn, size=(3, h, w))), + "t5_text_embeddings": LambdaDataset(partial(torch.randn, size=(len_t5, 1024))), + "t5_text_mask": LambdaDataset(partial(torch.randint, low=0, high=2, size=(len_t5,), dtype=torch.int64)), + "fps": LambdaDataset(lambda: 1.0), + "image_size": LambdaDataset(partial(torch.tensor, [h, w, h, w], dtype=torch.float32)), + "num_frames": LambdaDataset(lambda: 1), + "padding_mask": LambdaDataset(partial(torch.zeros, size=(1, h, w))), + } + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/dataset_provider.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/dataset_provider.py new file mode 100644 index 0000000000000000000000000000000000000000..8c28bf2f7a0f43916211df9c834c60de6388c593 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/dataset_provider.py @@ -0,0 +1,243 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import omegaconf + +try: + from megatron.core import parallel_state + + USE_MEGATRON = True +except ImportError: + USE_MEGATRON = False +from typing import Callable, Optional + +from webdataset.handlers import warn_and_continue + +import cosmos_policy._src.imaginaire.datasets.webdataset.decoders.image as image_decoders +import cosmos_policy._src.imaginaire.datasets.webdataset.decoders.pickle as pickle_decoders +import cosmos_policy._src.imaginaire.datasets.webdataset.distributors as distributors +import cosmos_policy._src.predict2.datasets.decoders.video_decoder as video_decoder +import cosmos_policy._src.predict2.datasets.distributor.parallel_sync_multi_aspect_ratio as parallel_sync_multi_aspect_ratio +import cosmos_policy._src.predict2.datasets.webdataset as webdataset +from cosmos_policy._src.imaginaire.datasets.webdataset.config.schema import DatasetConfig +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.predict2.datasets.augmentor_provider import AUGMENTOR_OPTIONS +from cosmos_policy._src.predict2.datasets.data_sources.data_registration import DATASET_OPTIONS +from cosmos_policy._src.predict2.datasets.utils import IMAGE_RES_SIZE_INFO, VIDEO_RES_SIZE_INFO + + +def get_video_dataset( + dataset_name: str, + video_decoder_name: str, + resolution: str, + is_train: bool = True, + num_video_frames: int = 121, + chunk_size: int = 0, + min_fps_thres: int = 10, + max_fps_thres: int = 60, + dataset_resolution_type: str = "all", + augmentor_name: str = "video_basic_augmentor_v1", + object_store: Optional[str] = "s3", + caption_type: str = "t2w_qwen2p5_7b", + embedding_type: str = "t5_xxl", + detshuffle: bool = False, + long_caption_ratio: int = 7, + medium_caption_ratio: int = 2, + short_caption_ratio: int = 1, + user_caption_ratio: int = 90, + dataset_info_fn: Optional[Callable] = None, + use_native_fps: bool = True, + use_original_fps: bool = False, + use_random_consecutive_frames: bool = False, # If True, sample random consecutive frames within clip window, preserving original fps between frames (no frame skipping or duplication); good for generating unique contiguous video segments for augmentation. + use_random_interleaved_frames: bool = False, # If True, enable random interleaved (non-consecutive) frame sampling for fractional fps upsampling/downsampling (e.g., 24->30fps), producing temporally varied clips by mixing frame strides. +) -> omegaconf.dictconfig.DictConfig: + assert resolution in VIDEO_RES_SIZE_INFO.keys(), "The provided resolution cannot be found in VIDEO_RES_SIZE_INFO." + assert object_store in [ + "s3", + "swiftstack", + "gcp", + False, + ], "We support s3 and swiftstack only, or False for local loading." + basic_augmentor_names = [ + "video_basic_augmentor_v2", + "video_basic_augmentor_v2_with_control", + "noframedrop_nocameramove_video_augmentor_v1", + ] + if video_decoder_name == "video_naive_bytes": + assert augmentor_name in basic_augmentor_names, ( + "We can only use video_basic_augmentor_v2 with video_naive_bytes decoder." + ) + if augmentor_name in basic_augmentor_names: + assert video_decoder_name == "video_naive_bytes", ( + "We can only use video_naive_bytes decoder with video_basic_augmentor_v2." + ) + + assert dataset_resolution_type in [ + "all", + "gt720p", + "gt1080p", + ], f"The provided dataset resolution type {dataset_resolution_type} is not supported." + # dataset_resolution_type + # -- all - uses all dataset resolutions + # -- gt720p - Uses only resolutions >= 720p + # -- gt1080p - Uses only resolutions >= 1080p + if not object_store: + assert dataset_info_fn is not None, "dataset_info_fn is required for local loading." + dataset_info = dataset_info_fn() + else: + dataset_info_fn = DATASET_OPTIONS[dataset_name] + dataset_info = dataset_info_fn(object_store, caption_type, embedding_type, dataset_resolution_type) # type: ignore + augmentor = AUGMENTOR_OPTIONS[augmentor_name]( + resolution=resolution, + caption_type=caption_type, + embedding_type=embedding_type, + min_fps=min_fps_thres, + max_fps=max_fps_thres, + long_caption_ratio=long_caption_ratio, + medium_caption_ratio=medium_caption_ratio, + short_caption_ratio=short_caption_ratio, + user_caption_ratio=user_caption_ratio, + num_video_frames=num_video_frames, + use_native_fps=use_native_fps, + use_original_fps=use_original_fps, + use_random_consecutive_frames=use_random_consecutive_frames, + use_random_interleaved_frames=use_random_interleaved_frames, + ) + + if ( + USE_MEGATRON + and parallel_state.is_initialized() + and ( + parallel_state.get_context_parallel_world_size() > 1 + or parallel_state.get_tensor_model_parallel_world_size() > 1 + ) + ): + log.critical( + f"Using parallelism size CP :{parallel_state.get_context_parallel_world_size()}, TP :{parallel_state.get_tensor_model_parallel_world_size()} for video dataset, switch to ShardlistMultiAspectRatioParallelSync distributor" + ) + distributor = parallel_sync_multi_aspect_ratio.ShardlistMultiAspectRatioParallelSync( + shuffle=True, + split_by_node=True, + split_by_worker=True, + resume_flag=True, + verbose=True, + is_infinite_loader=is_train, + ) + detshuffle = True # overwrite detshuffle. + else: + distributor = distributors.ShardlistMultiAspectRatio( + shuffle=True, + split_by_node=True, + split_by_worker=True, + resume_flag=True, + verbose=False, + is_infinite_loader=is_train, + ) + + video_data_config = DatasetConfig( + keys=[], # use the per_dataset_keys in DatasetInfo instead + buffer_size=100, + streaming_download=True, + dataset_info=dataset_info, + distributor=distributor, + decoders=[ + video_decoder.construct_video_decoder( + video_decoder_name=video_decoder_name, + sequence_length=num_video_frames, + chunk_size=chunk_size, + min_fps_thres=min_fps_thres, + max_fps_thres=max_fps_thres, + ), + pickle_decoders.pkl_decoder, + ], + augmentation=augmentor, + remove_extension_from_keys=True, + sample_keys_full_list_path=None, + ) + + return webdataset.Dataset(config=video_data_config, decoder_handler=warn_and_continue, detshuffle=detshuffle) + + +def get_image_dataset( + dataset_name: str, + resolution: str, + dataset_resolution_type: str = "all", + is_train: bool = True, + augmentor_name: str = "image_basic_augmentor", + object_store: str = "s3", + detshuffle: bool = False, + caption_type: str = "ai_v3p1", + embedding_type: str = "t5_xxl", +) -> omegaconf.dictconfig.DictConfig: + assert resolution in IMAGE_RES_SIZE_INFO.keys(), "The provided resolution cannot be found in IMAGE_RES_SIZE_INFO." + assert object_store in ["s3", "swiftstack", "gcp"], "We support s3, gcp and swiftstack only." + assert dataset_resolution_type in [ + "all", + "gt720p", + "gt1080p", + ], f"The provided dataset resolution type {dataset_resolution_type} is not supported." + # dataset_resolution_type + # -- all - uses all dataset resolutions + # -- gt720p - Uses only resolutions >= 720p + # -- gt1080p - Uses only resolutions >= 1080p + dataset_info_fn = DATASET_OPTIONS[dataset_name] + dataset_info = dataset_info_fn(object_store, caption_type, embedding_type, dataset_resolution_type) + augmentation = AUGMENTOR_OPTIONS[augmentor_name]( + resolution=resolution, + caption_type=caption_type, + embedding_type=embedding_type, + ) + + if parallel_state.is_initialized() and ( + parallel_state.get_context_parallel_world_size() > 1 + or parallel_state.get_tensor_model_parallel_world_size() > 1 + ): + log.critical( + f"Using parallelism size CP :{parallel_state.get_context_parallel_world_size()}, TP :{parallel_state.get_tensor_model_parallel_world_size()} for image dataset, switch to ShardlistMultiAspectRatioParallelSync distributor" + ) + distributor = parallel_sync_multi_aspect_ratio.ShardlistMultiAspectRatioParallelSync( + shuffle=True, + split_by_node=True, + split_by_worker=True, + resume_flag=True, + verbose=True, + is_infinite_loader=is_train, + ) + detshuffle = True # overwrite detshuffle. + else: + distributor = distributors.ShardlistMultiAspectRatio( + shuffle=True, + split_by_node=True, + split_by_worker=True, + resume_flag=True, + verbose=False, + is_infinite_loader=is_train, + ) + + image_data_config = DatasetConfig( + keys=[], + # https://gitlab-master.nvidia.com/dir/imaginaire4/-/issues/119 + buffer_size=25, + streaming_download=True, + dataset_info=dataset_info, + distributor=distributor, + decoders=[ + image_decoders.pil_loader, + pickle_decoders.pkl_decoder, + ], + augmentation=augmentation, + ) + + return webdataset.Dataset(config=image_data_config, detshuffle=detshuffle) diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/decoders/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/decoders/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/decoders/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/decoders/video_decoder.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/decoders/video_decoder.py new file mode 100644 index 0000000000000000000000000000000000000000..8d28d1b96439d585f5df7ee702f853a70c75cc07 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/decoders/video_decoder.py @@ -0,0 +1,546 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import io +import re +from random import randint +from typing import List, Tuple + +import decord +import numpy as np +import torch +from PIL import Image + +Image.MAX_IMAGE_PIXELS = 933120000 +_VIDEO_EXTENSIONS = "mp4 avi webm mov".split() + +VIDEO_DECODER_OPTIONS = {} + + +def video_decoder_register(key): + def decorator(func): + VIDEO_DECODER_OPTIONS[key] = func + return func + + return decorator + + +def basic_check_on_inputs( + n_video_frames: int, n_target_frames: int, video_fps: float, min_fps_thres: int, max_fps_thres: int +) -> str: + if n_video_frames <= 0: + return "n_video_frames must be positive" + if min_fps_thres <= 0: + return "min_fps_thres must be positive" + if video_fps < 1: + return "Video fps lower than 1, skipping" + if max_fps_thres < min_fps_thres: + return "max_fps_thres must be greater than or equal to min_fps_thres" + if n_target_frames <= 1: + return "sequence_length must be greater than 1" + if n_target_frames > n_video_frames: + return f"Specified sequence_length {n_target_frames} exceeds num frames in video {n_video_frames}." + + return "success" + + +def sample_chunk_index_from_chunked_video( + n_video_frames: int, + n_target_frames: int, + chunk_size: int, +) -> Tuple[int, int, str]: + """ + Sample a chunk from the chunked videos. Our videos are stored as regular mp4 files but with chunked captions. There is one caption per [chunk_size] frames. + If the last chunk has frames >= chunk_size / 2, it will be treated as a separate chunk and has its own caption. Otherwise, it will be treated as part of the previous chunk. + e.g. if chunk_size = 256, possible number of frames in a chunk: [1, 383] + + Args: + n_video_frames: total number of frames in the video + n_target_frames: number of requested frames + chunk_size: number of frames in each chunk. The last chunk will be treated differently + + Returns: + sampled_chunk_indexs + n_frames_in_chunk + message + """ + n_chunks = max(n_video_frames // chunk_size, 1) + + # Check if the last chunk has separate window + # This happens only if remainder frames >= chunk_size / 2 [data annotation was done this way] + # Else this is used as a part of previous window. + n_frames_in_last_chunk = n_video_frames - n_chunks * chunk_size + if n_frames_in_last_chunk >= int(0.5 * chunk_size): + if n_frames_in_last_chunk > n_target_frames: + n_chunks += 1 + + sampled_chunk_index = randint(0, n_chunks - 1) + if sampled_chunk_index == n_chunks - 1: + # For the last chunk, use all of the remaining frames + n_frames_in_chunk = n_video_frames - sampled_chunk_index * chunk_size + else: + # Else use only the chunk size + n_frames_in_chunk = chunk_size + + if n_target_frames > n_frames_in_chunk: + error_message = f"Requested sequence_length {n_target_frames} exceeds curr_chunk_size {n_frames_in_chunk}, n_video_frames={n_video_frames}, chunk_size={chunk_size}, sampled_chunk_index={sampled_chunk_index}." + return -1, 0, error_message + + return sampled_chunk_index, n_frames_in_chunk, "success" + + +@video_decoder_register("video_naive_bytes") +def video_naive_bytes(*args, **kwargs): + """ + do nothing, just return the video bytes + """ + del args, kwargs + + def video_decoder( + key: str, + data: bytes, + ): + extension = re.sub(r".*[.]", "", key) + if extension.lower() not in _VIDEO_EXTENSIONS: + return None + + return data + + return video_decoder + + +@video_decoder_register("chunked_video_decoder") +def chunked_video_decoder( + chunk_size: int = 0, + sequence_length: int = 34, + min_fps_thres: int = 1, + max_fps_thres: int = 9999, + num_threads=4, +): + """ + Video decoder for videos with chunked captions. + It first sample a chunk from the video then sample the start frame within the chunk. + It has a basic check to make sure the video fps falls within the range [min_fps_thres, max_fps_thres]. Otherwise, it will skip the video sample. + + Args: + - chunk_size (int): How the video is divided into chunks. Only return frames within a chunk. chunk_size=0 means we use full video length. Defaults to 0. + - sequence_length (int) : Number of frames returned by the function + - min_fps_thres (int): Minimum fps threshold allowed. + - max_fps_thres (int): Maximum fps threshold allowed. + - num_thread (int): Number of threads for decord. + + Returns: + dict with video frames tensor and additional attributes including + - fps + - orig_fps + - num_frames + - chunk_index + - frame_start + - frame_end + - n_orig_video_frames + """ + + def video_decoder( + key: str, + data: bytes, + ): + extension = re.sub(r".*[.]", "", key) + if extension.lower() not in _VIDEO_EXTENSIONS: + return None + + video_buffer = io.BytesIO(data) + video_reader = decord.VideoReader(video_buffer, num_threads=num_threads) + + n_target_frames = sequence_length if sequence_length > 0 else len(video_reader) + n_video_frames = len(video_reader) + video_fps = int(np.round(video_reader.get_avg_fps())) + cur_chunk_size = n_video_frames if chunk_size == 0 else chunk_size + + # basic check + message = basic_check_on_inputs( + n_video_frames=n_video_frames, + n_target_frames=n_target_frames, + video_fps=video_fps, + min_fps_thres=min_fps_thres, + max_fps_thres=max_fps_thres, + ) + if message != "success": + raise ValueError(message) + + # check if video fps is within the specified range + if video_fps < min_fps_thres: + raise ValueError(f"Video fps {video_fps} lower than {min_fps_thres}, skipping") + if video_fps > max_fps_thres: + raise ValueError(f"Video fps {video_fps} larger than {max_fps_thres}, skipping") + + sampled_chunk_index, n_frames_in_chunk, message = sample_chunk_index_from_chunked_video( + n_video_frames=n_video_frames, + n_target_frames=n_target_frames, + chunk_size=cur_chunk_size, + ) + if sampled_chunk_index == -1: + raise ValueError(message) + else: + assert message == "success" + + # Select the frame start index and frame end index + chunk_frame_start = sampled_chunk_index * chunk_size + # Start index is randomly selected in the chunk + frame_start = chunk_frame_start + int(np.random.choice(n_frames_in_chunk - n_target_frames, 1)) + frame_end = frame_start + n_target_frames + + # Subsample the frames + video_frames = video_reader.get_batch(np.arange(frame_start, frame_end).tolist()).asnumpy() + video_frames = torch.from_numpy(video_frames).permute(3, 0, 1, 2) # (T, H, W, C) -> (C, T, H, W) + + # Clean up + video_reader.seek(0) # set video reader point back to 0 to clean up cache + del video_reader # delete the reader to avoid memory leak + + return { + "video": video_frames, + "fps": float(video_fps), + "orig_fps": float(video_fps), + "num_frames": video_frames.shape[1], + "chunk_index": sampled_chunk_index, + "frame_start": frame_start, + "frame_end": frame_end, + "n_orig_video_frames": n_video_frames, + } + + return video_decoder + + +def get_frame_indices_w_lowered_fps( + n_video_frames: int, + video_fps: int, + min_fps_thres: int, + max_fps_thres: int, + n_target_frames: int, +) -> Tuple[List[int], float]: + """Generates frame indices for video sampling with FPS control. + + This function determines valid stride lengths for sampling frames from a video, + preferring lower FPS (larger strides) when multiple options are available. + It returns both the selected frame indices and the resulting FPS. + + Args: + n_video_frames: Total number of frames in the original video. + video_fps: Original video frames per second. + min_fps_thres: Minimum allowed frames per second. + max_fps_thres: Maximum allowed frames per second. + n_target_frames: Number of frames to sample. + + Returns: + A tuple containing: + - list[int]: Frame indices to sample from the original video. + - float: The resulting frames per second after sampling. + + Raises: + ValueError: If no valid stride options are available given the constraints. + ValueError: If input parameters are invalid (e.g., negative values). + """ + # Calculate stride range + min_stride = 1 + max_stride = (n_video_frames - 1) // (n_target_frames - 1) + + valid_strides = [] + for stride in range(min_stride, max_stride + 1): + # Check if we can get n_target_frames frames with this stride + if (n_video_frames - stride * (n_target_frames - 1)) > 0: + new_fps = video_fps / stride + if min_fps_thres <= new_fps <= max_fps_thres: + valid_strides.append(stride) + + if not valid_strides: + raise ValueError( + f"No valid stride options available for the given constraints. " + f"stride range = [{min_stride}, {max_stride}]; " + f"original FPS = {video_fps}; " + f"n_target_frames = {n_target_frames}; " + f"min_fps_thres = {min_fps_thres}; " + f"max_fps_thres = {max_fps_thres}; " + f"original num_frames = {n_video_frames}" + ) + + # Select stride with weighted probability + if len(valid_strides) >= 2: + stride_choices = valid_strides[-2:] # Taking last two as they're the largest + weights = [0.01, 0.99] # [smaller_stride, larger_stride] + selected_stride = np.random.choice(stride_choices, p=weights) + else: + selected_stride = valid_strides[0] + + # Calculate the maximum valid start index and random start frame + max_start_idx = n_video_frames - (n_target_frames - 1) * selected_stride + frame_start = np.random.randint(0, max_start_idx) + + # Generate frame indices + frame_indices = [frame_start + i * selected_stride for i in range(n_target_frames)] + return frame_indices, video_fps / selected_stride + + +@video_decoder_register("chunked_video_decoder_w_lower_fps") +def chunked_video_decoder_w_lower_fps( + chunk_size: int = 0, + sequence_length: int = 34, + min_fps_thres: int = 4, + max_fps_thres: int = 30, + num_threads: int = 4, +) -> dict: + """ + Video decoder for videos with chunked captions. + It first sample a chunk from the video then sample the start frame within the chunk. + It has high probability (>99%) to lower the fps with frame sampling whenever allowed. + + Args: + - chunk_size (int): How the video is divided into chunks. Only return frames within a chunk. chunk_size=0 means we use full video length. Defaults to 0. + - sequence_length (int) : Number of frames returned by the function + - min_fps_thres: Minimum FPS threshold + - max_fps_thres: Maximum FPS threshold + - num_threads: Number of threads for decord + + Returns: + dict with video frames tensor and additional attributes including + - fps + - orig_fps + - num_frames + - chunk_index + - frame_start + - frame_end + - n_orig_video_frames + """ + + def video_decoder( + key: str, + data: bytes, + ) -> dict[str, torch.Tensor | int]: + # Check video extension + extension = re.sub(r".*[.]", "", key) + if extension.lower() not in _VIDEO_EXTENSIONS: + return None + + # Read video + video_buffer = io.BytesIO(data) + video_reader = decord.VideoReader(video_buffer, num_threads=num_threads) + + n_target_frames = sequence_length if sequence_length > 0 else len(video_reader) + n_video_frames = len(video_reader) + video_fps = int(np.round(video_reader.get_avg_fps())) + cur_chunk_size = n_video_frames if chunk_size == 0 else chunk_size + + # basic check + message = basic_check_on_inputs( + n_video_frames=n_video_frames, + n_target_frames=n_target_frames, + video_fps=video_fps, + min_fps_thres=min_fps_thres, + max_fps_thres=max_fps_thres, + ) + if message != "success": + raise ValueError(message) + + sampled_chunk_index, n_frames_in_chunk, message = sample_chunk_index_from_chunked_video( + n_video_frames=n_video_frames, + n_target_frames=n_target_frames, + chunk_size=cur_chunk_size, + ) + if sampled_chunk_index == -1: + raise ValueError(message) + else: + assert message == "success" + + chunk_frame_start = sampled_chunk_index * cur_chunk_size + + frame_indices, adjusted_fps = get_frame_indices_w_lowered_fps( + n_video_frames=n_frames_in_chunk, + video_fps=video_fps, + min_fps_thres=min_fps_thres, + max_fps_thres=max_fps_thres, + n_target_frames=n_target_frames, + ) + frame_indices = [chunk_frame_start + idx for idx in frame_indices] + + # Sample frames + video_frames = video_reader.get_batch(frame_indices).asnumpy() + video_frames = torch.from_numpy(video_frames).permute(3, 0, 1, 2) # (T, H, W, C) -> (C, T, H, W) + + # Clean up + video_reader.seek(0) + del video_reader + + output = { + "video": video_frames, + "fps": float(adjusted_fps), + "orig_fps": video_fps, + "num_frames": video_frames.shape[1], + "chunk_index": sampled_chunk_index, + "frame_start": frame_indices[0], + "frame_end": frame_indices[-1], + "n_orig_video_frames": n_video_frames, + } + return output + + return video_decoder + + +@video_decoder_register("chunked_video_decoder_with_fixed_fps") +def chunked_video_decoder_with_fixed_fps( + chunk_size: int = 0, + sequence_length: int = 34, + min_fps_thres: int = 4, + max_fps_thres: int = 30, + num_threads: int = 4, +) -> dict: + """ + Video decoder optimized for processing videos with chunked captions. + + unlike other video decoders which return video frames of requested sequence_length. + The function returns a randomly sampled chunk with duration between 4 seconds and 8 seconds whenever possible. + The chunk will be provided to modeling code and the frame subsampling happens on the modeling side. + !!! IMPORTANT: it can only work with batch size 1 otherwise, different length video can not be concatenated. + + This decoder first samples a chunk from the video, then selects frames within that chunk. + It dynamically adjusts the frame rate with a high probability (>99%) to lower the FPS + through frame sampling when conditions allow, ensuring efficient processing while + maintaining video quality. + + The decoder handles variable chunk durations with special processing for chunks that are + either too short or too long, ensuring consistent output regardless of input video properties. + + Args: + chunk_size (int): Size of video chunks in frames. If set to 0, the entire video length + is used as a single chunk. Defaults to 0. + sequence_length (int): Number of frames to extract from the video. Defaults to 34. + min_fps_thres (int): Minimum acceptable frames per second. Videos with lower FPS will + raise errors. Defaults to 4. + max_fps_thres (int): Maximum acceptable frames per second. Higher FPS videos will be + downsampled. Defaults to 30. + num_threads (int): Number of threads to use for video decoding with decord. Defaults to 4. + + Returns: + dict: A dictionary containing video frames tensor and metadata including: + - video: Tensor of shape (C, T, H, W) containing the sampled video frames + - fps: Actual frames per second (float) + - orig_fps: Original video frame rate (int) + - num_frames: Number of frames extracted + - chunk_index: Index of the sampled chunk + - frame_start: Starting frame index in original video + - frame_end: Ending frame index in original video + - n_orig_video_frames: Total number of frames in original video + + Raises: + ValueError: If video duration is too short, if FPS is outside acceptable range, + or if selected chunk has insufficient frames. + + Note: + - Chunks with duration < 4.0 seconds are skipped with an error. + - Chunks with duration > 8.0 seconds are capped to 8.0 seconds worth of frames. + - For best results, ensure videos have FPS between min_fps_thres and max_fps_thres. + """ + + def video_decoder( + key: str, + data: bytes, + ) -> dict[str, torch.Tensor | int]: + # Check video extension + extension = re.sub(r".*[.]", "", key) + if extension.lower() not in _VIDEO_EXTENSIONS: + return None + + # Read video + video_buffer = io.BytesIO(data) + video_reader = decord.VideoReader(video_buffer, num_threads=num_threads) + + n_target_frames = sequence_length if sequence_length > 0 else len(video_reader) + n_video_frames = len(video_reader) + video_fps_float = video_reader.get_avg_fps() + video_fps = int(np.round(video_fps_float)) + cur_chunk_size = n_video_frames if chunk_size == 0 else chunk_size + + # basic check + message = basic_check_on_inputs( + n_video_frames=n_video_frames, + n_target_frames=n_target_frames, + video_fps=video_fps, + min_fps_thres=min_fps_thres, + max_fps_thres=max_fps_thres, + ) + if message != "success": + raise ValueError(message) + + sampled_chunk_index, n_frames_in_chunk, message = sample_chunk_index_from_chunked_video( + n_video_frames=n_video_frames, + n_target_frames=n_target_frames, + chunk_size=cur_chunk_size, + ) + if sampled_chunk_index == -1: + raise ValueError(message) + else: + assert message == "success" + + chunk_frame_start = sampled_chunk_index * cur_chunk_size + + chunk_duration = n_frames_in_chunk / video_fps_float + if chunk_duration < 4.0: + raise ValueError(f"Chunk duration {chunk_duration} is less than 4.0 seconds, skipping") + + if chunk_duration > 8.0: + n_frames_needed = int(np.ceil(8.0 * video_fps)) + else: + n_frames_needed = n_frames_in_chunk + + chunk_frame_end = chunk_frame_start + n_frames_needed + + frame_indices = np.arange(chunk_frame_start, chunk_frame_end).tolist() + + # Sample frames + video_frames = video_reader.get_batch(frame_indices).asnumpy() + video_frames = torch.from_numpy(video_frames).permute(3, 0, 1, 2) # (T, H, W, C) -> (C, T, H, W) + + # Clean up + video_reader.seek(0) + del video_reader + + output = { + "video": video_frames, + "fps": video_fps_float, + "orig_fps": video_fps, + "num_frames": video_frames.shape[1], + "chunk_index": sampled_chunk_index, + "frame_start": frame_indices[0], + "frame_end": frame_indices[-1], + "n_orig_video_frames": n_video_frames, + } + return output + + return video_decoder + + +def construct_video_decoder( + video_decoder_name: str = "chunked_video_decoder", + chunk_size: int = 0, + sequence_length: int = 34, + min_fps_thres: int = 1, + max_fps_thres: int = 9999, + num_threads=4, +): + return VIDEO_DECODER_OPTIONS[video_decoder_name]( + chunk_size=chunk_size, + sequence_length=sequence_length, + min_fps_thres=min_fps_thres, + max_fps_thres=max_fps_thres, + num_threads=num_threads, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/decoders/video_decoder_test.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/decoders/video_decoder_test.py new file mode 100644 index 0000000000000000000000000000000000000000..feed6d41dbf3dde8f93bf7d70658c11e806e4bd0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/decoders/video_decoder_test.py @@ -0,0 +1,157 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import numpy as np +import pytest + +from cosmos_policy._src.predict2.datasets.decoders.video_decoder import ( + basic_check_on_inputs, + get_frame_indices_w_lowered_fps, + sample_chunk_index_from_chunked_video, +) + + +@pytest.fixture +def set_random_seed(): + """Fixture to ensure reproducible random numbers.""" + np.random.seed(42) + yield + np.random.seed(None) + + +@pytest.mark.L0 +def test_basic_functionality(set_random_seed): + """Test basic functionality with valid inputs.""" + indices, fps = get_frame_indices_w_lowered_fps( + n_video_frames=100, video_fps=30, min_fps_thres=4, max_fps_thres=30, n_target_frames=5 + ) + + assert len(indices) == 5 + assert all(0 <= idx < 100 for idx in indices) + assert 4 <= fps <= 30 + assert indices == sorted(indices) # Ensure indices are monotonically increasing + + +@pytest.mark.L0 +def test_sequence_spacing(): + """Test that frame indices are evenly spaced.""" + indices, _ = get_frame_indices_w_lowered_fps( + n_video_frames=100, video_fps=30, min_fps_thres=4, max_fps_thres=30, n_target_frames=5 + ) + + differences = [indices[i + 1] - indices[i] for i in range(len(indices) - 1)] + assert len(set(differences)) == 1 # All differences should be equal + + +@pytest.mark.L0 +def test_fps_bounds(): + """Test that resulting FPS is within specified bounds.""" + _, fps = get_frame_indices_w_lowered_fps( + n_video_frames=100, video_fps=30, min_fps_thres=4, max_fps_thres=30, n_target_frames=5 + ) + + assert 4 <= fps <= 30 + + +@pytest.mark.parametrize( + "invalid_input", + [ + {"n_video_frames": 0}, + {"video_fps": 0}, + {"min_fps_thres": 0}, + {"max_fps_thres": 3, "min_fps_thres": 4}, + {"n_target_frames": 1}, + {"n_target_frames": 101, "n_video_frames": 100}, + ], +) +@pytest.mark.L0 +def test_invalid_inputs(invalid_input): + """Test that invalid inputs raise appropriate errors.""" + default_args = { + "n_video_frames": 100, + "video_fps": 30, + "min_fps_thres": 4, + "max_fps_thres": 30, + "n_target_frames": 5, + } + + args = {**default_args, **invalid_input} + message = basic_check_on_inputs(**args) + assert message != "success" # assert raise error message + + +@pytest.mark.L0 +def test_stride_selection_bias(set_random_seed): + """Test that larger strides (lower FPS) are selected more frequently.""" + results = [] + for _ in range(100): + _, fps = get_frame_indices_w_lowered_fps( + n_video_frames=100, video_fps=30, min_fps_thres=4, max_fps_thres=30, n_target_frames=5 + ) + results.append(fps) + + # Check that lower FPS (larger strides) are selected more often + lower_fps_count = sum(1 for fps in results if fps < (4 + 30) / 2) + assert lower_fps_count > len(results) * 0.6 # Should be selected roughly 75% of the time + + +@pytest.mark.L0 +def test_extreme_case(): + """Test with minimal valid input values.""" + indices, fps = get_frame_indices_w_lowered_fps( + n_video_frames=5, video_fps=8, min_fps_thres=4, max_fps_thres=8, n_target_frames=2 + ) + + assert len(indices) == 2 + assert all(0 <= idx < 5 for idx in indices) + assert 4 <= fps <= 8 + + +@pytest.mark.L0 +def test_no_valid_strides(): + """Test that appropriate error is raised when no valid strides exist.""" + with pytest.raises(ValueError) as exc_info: + get_frame_indices_w_lowered_fps( + n_video_frames=10, video_fps=30, min_fps_thres=25, max_fps_thres=29, n_target_frames=9 + ) + assert "No valid stride options available" in str(exc_info.value) + + +@pytest.mark.L0 +def test_sample_chunk_index_from_chunked_video(): + sampled_chunk_index, n_frames_in_chunk, message = sample_chunk_index_from_chunked_video( + n_video_frames=383, + n_target_frames=4, + chunk_size=256, + ) + assert n_frames_in_chunk == 383 + + n_frames_in_chunk_list = set() + for _ in range(10): + sampled_chunk_index, n_frames_in_chunk, message = sample_chunk_index_from_chunked_video( + n_video_frames=641, + n_target_frames=4, + chunk_size=256, + ) + n_frames_in_chunk_list.add(n_frames_in_chunk) + assert n_frames_in_chunk_list == {256, 129} + + message = sample_chunk_index_from_chunked_video( + n_video_frames=4, + n_target_frames=121, + chunk_size=256, + ) + + assert message != "success" diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/distributor/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/distributor/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/distributor/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/distributor/parallel_sync_multi_aspect_ratio.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/distributor/parallel_sync_multi_aspect_ratio.py new file mode 100644 index 0000000000000000000000000000000000000000..eaa128b36ab3a8c95baddaf05de362637e89286d --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/distributor/parallel_sync_multi_aspect_ratio.py @@ -0,0 +1,151 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# This script contains the code for multi-aspect ratio shard iterator + +import os +import random +import time +from copy import deepcopy + +import torch + +try: + from megatron.core import parallel_state + + USE_MEGATRON = True +except ImportError: + USE_MEGATRON = False +from webdataset.utils import pytorch_worker_info + +from cosmos_policy._src.imaginaire.datasets.webdataset.config.schema import TarSample +from cosmos_policy._src.imaginaire.datasets.webdataset.distributors.multi_aspect_ratio import ShardlistMultiAspectRatio +from cosmos_policy._src.imaginaire.utils import log + + +class ShardlistMultiAspectRatioParallelSync(ShardlistMultiAspectRatio): + r""" + An iterable dataset that parses and yields tar files. + This distributor is based on ShardlistMultiAspectRatio. + Additionally, it allows users to synchronize inputs for context/tensor parallelism. This is achieved by specifying the context/tensor parallel group size during initialization. + """ + + def __init__(self, **kwargs): + r"""Create a multi-aspect ratio ShardList.""" + super().__init__(**kwargs) + self.enable_parallel() + + def _obtain_node_worker_url_mapping( + self, + url_aspect_split: dict[str, list[TarSample]], + num_urls_per_worker: int, + group_id: int, + group_num: int, + worker_id: int, + num_workers: int, + ): + r"""This function obtains the worker-URL mapping. It assigns the tar list seen by + each workers. + + Args: + url_aspect_split (dict[list[TarSample]]: TarSample split by aspect ratio + num_urls_per_worker (int): Number of tar files seen by each worker + group_id (int): Rank of the current GPU + group_num (int): Total number of groups + worker_id (int): ID for the current worker in the dataloader + num_workers (int): Total number of workers in the dataloader + + Returns: + URL list for the current worker + """ + assert self.split_by_node is True and self.split_by_worker is True + + # First chunk the tars + chunk_mappings = [] + for aspect_ratio in url_aspect_split: + samples_asp = url_aspect_split[aspect_ratio] + nchunks_asp = int(len(samples_asp) / num_urls_per_worker) + for chunk_id in range(nchunks_asp): + chunk_mappings.append((aspect_ratio, samples_asp[chunk_id::nchunks_asp])) + # Split by rank and workers + chunk_mappings = chunk_mappings[group_id::group_num] + chunk_mappings = chunk_mappings[worker_id::num_workers] + + assert len(chunk_mappings) == 1, f"Length of chunk_mappings {len(chunk_mappings)} != 1" + return chunk_mappings[0][1] + + def enable_parallel(self): + # Ranks of the same pp/tp/cp group will have the same dp rank and thus share the same group id. + self.group_id = parallel_state.get_data_parallel_rank() + # The size of the group is how many GPUs we use to process one batch of data. + self.group_size = torch.distributed.get_world_size() // parallel_state.get_data_parallel_world_size() + + def obtain_url_list(self): + r"""Return an iterator over the shards.""" + + rank, world_size, worker_id, num_workers = pytorch_worker_info() + + num_groups = world_size // self.group_size + # Setting epoch and start index + if self.resume_flag: + self.epoch = int(os.environ.get("WDS_EPOCH_NUM", 0)) + + # This tells us number of chunks that have been seen by one GPU + self.start_index = int(os.environ.get("WDS_START_INDEX", 0)) // self.chunk_size + + url_aspect_split = deepcopy(self.url_aspect_split) + + # nworkers_all is no longer world_size * num_workers, since self.group_size workers duplicate + nworkers_all = num_groups * num_workers + + if self.verbose: + log.info(f"Total {nworkers_all} workers are in effect") + + # Perform DDP equalization + url_aspect_split, num_urls_per_worker = self._ddp_equalize(url_aspect_split, nworkers_all) + + # Form a mapping of url_aspect_split to node and workers + urls = self._obtain_node_worker_url_mapping( + url_aspect_split, num_urls_per_worker, self.group_id, num_groups, worker_id, num_workers + ) + + if self.shuffle: + random.Random(self.group_id).shuffle(urls) + + # This tells us the number of chunks seen by one worker. + # Do not iterate over the seen chunks. + start_index_per_worker = self.start_index // num_workers + if not self.is_infinite_loader: + urls = urls[start_index_per_worker:] + + if self.verbose: + log.info( + f"Rank {rank}, group {self.group_id}, worker {worker_id} of {num_workers}, group_size {self.group_size} got {len(urls)} urls, first five are {urls[:5]}" + ) + + return urls + + def __iter__(self): + url_list = self.obtain_url_list() + + if self.is_infinite_loader: + while True: + cur_time = time.time_ns() + random.Random(cur_time).shuffle(url_list) + for url in url_list: + yield dict(url=url) + else: + for url in url_list: + yield dict(url=url) diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/item_dataset.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/item_dataset.py new file mode 100644 index 0000000000000000000000000000000000000000..acc7e1608bf3264decffe1ac23f4ac76080a3f84 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/item_dataset.py @@ -0,0 +1,439 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import dataclasses +import os +from typing import Tuple + +import cv2 +import numpy as np +import torch +import torch.nn.functional as F + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + + +@dataclasses.dataclass +class ItemDatasetConfig: + path: str + length: int + + +class PromptOnlyItemDataset(torch.utils.data.Dataset): + """ + A simple dataset class for handling sequences of pickle data read from a specified path. + It supports reading from local paths or S3. It currently handles prompts and T5 embeddings. + The class is mainly for debug and testing purposes. + + Args: + path (str): The path to the dataset source. Can be a local file system path or an S3 bucket path. + start_index (int, optional): The starting index of the dataset to consider (inclusive). Defaults to 0. + end_index (int, optional): The ending index of the dataset to consider (exclusive). Defaults to 32. + max_t5_length (int, optional): The maximum length for T5 embedding. Defaults to 512. The sequence will be padded to this length. + + Example: + >>> dataset = PromptOnlyItemDataset(path='local/path/to/dataset', start_index=100, end_index=1100) + >>> len(dataset) + 1000 + >>> item = dataset[0] # Retrieve the first item in the dataset + """ + + def __init__( + self, + path: str = "s3://bucket/edify_video/v4/validation/item_dataset/sora_veo_v1", + start_index: int = 0, # inclusive + end_index=32, # exclusive + max_t5_length: int = 512, + height=704, + width=1280, + num_video_frames=136, + ): + self.start_index = start_index + self.end_index = end_index + self.path = path + self.height = height + self.width = width + self.num_video_frames = num_video_frames + log.warning( + f"using path: {path} and default s3 credentials in easy_io. It is user's responsibility to set up the correct credentials." + ) + max_length = easy_io.load(os.path.join(self.path, "meta_info.json"))["length"] + assert max_length >= end_index, f"dataset {path} max_length: {max_length}, end_index: {end_index}" + self.max_t5_length = max_t5_length + + def __len__(self): + return self.end_index - self.start_index + + def __getitem__(self, idx): + while True: + try: + return self._getitem(idx) + except Exception as e: + log.error(f"Error in __getitem__ {e}") + continue + + def _getitem(self, idx): + """ + Retrieves a specific pickle file based on its index, performing preprocessing + on text and image data and handling padding as needed. + + Args: + idx (int): Index of the dataset item to retrieve. + + Returns: + dict: A dictionary containing preprocessed dataset items, including embeddings, masks, + and potentially transformed images. + """ + item_fp = os.path.join(self.path, f"{self.start_index + idx:06d}.pkl") + item = easy_io.load(item_fp) + + if item is None: + raise ValueError(f"item is None: {item_fp}") + # t5 + mask = torch.LongTensor(self.max_t5_length).zero_() + length = len(item["t5_text_embeddings"]) + mask[0:length] = 1 + item["t5_text_mask"] = mask + if length < self.max_t5_length: + item["t5_text_embeddings"] = F.pad( + item["t5_text_embeddings"], (0, 0, 0, self.max_t5_length - length), value=0 + ).float() + item["t5_text_embeddings"] = item["t5_text_embeddings"][: self.max_t5_length] + + item["prompt"] = item["prompt"] + item["__idx__"] = idx + self.start_index + # hard coded conds + item["fps"] = 30.0 + item["image_size"] = torch.Tensor([self.height, self.width, self.height, self.width]).float() + item["padding_mask"] = torch.zeros((1, self.height, self.width)).float() + item["num_frames"] = torch.zeros((1)) + self.num_video_frames + + return item + + +class PromptImageItemDataset(PromptOnlyItemDataset): + def __init__( + self, + path: str = "s3://bucket/edify_video/v4/validation/item_dataset/sora_veo_v1", + num_videos: int = 32, + start_index: int = 0, # inclusive + end_index=32, # exclusive + max_t5_length: int = 512, + height=704, + width=1280, + num_max_frames=136, + augs=[], + aug_labels=[], + sigma_max_list=[], + control_weight_list=[], + pad_all_videos_to_same_length=True, + ): + self.start_index = start_index + self.end_index = end_index + self.path = path + self.num_videos = num_videos + self.height = height + self.width = width + self.num_max_frames = num_max_frames + log.warning( + f"using path: {path} and default s3 credentials in easy_io. It is user's responsibility to set up the correct credentials." + ) + self.max_t5_length = max_t5_length + + valid_idx = [i for i, aug in enumerate(augs) if len(aug.comb) > 0] + self.augs = [augs[i] for i in valid_idx] + self.aug_labels = [aug_labels[i] for i in valid_idx] + self.sigma_max_list = sigma_max_list + self.control_weight_list = control_weight_list + self.pad_all_videos_to_same_length = pad_all_videos_to_same_length + + def _getitem(self, idx): + cur_idx = self.start_index + idx + file_idx = cur_idx % self.num_videos + aug_idx = (cur_idx // self.num_videos) % len(self.augs) + sigma_idx = (cur_idx // (self.num_videos * len(self.augs))) % len(self.sigma_max_list) + control_weight_idx = (cur_idx // (self.num_videos * len(self.augs) * len(self.sigma_max_list))) % len( + self.control_weight_list + ) + file_paths = easy_io.list_dir_or_file(self.path, recursive=True, list_dir=False, suffix=".mp4") + file_paths = sorted([path[:-4] for path in file_paths]) + item_fp = os.path.join(self.path, file_paths[file_idx] + ".pkl") + item = easy_io.load(item_fp) + if item is None: + raise ValueError(f"item is None: {item_fp}") + # t5 + mask = torch.LongTensor(self.max_t5_length).zero_() + if isinstance(item["t5_text_embeddings"], dict): + t5_text_embeddings = list(item["t5_text_embeddings"].values()) + t5_text_embeddings = torch.cat([t5[:, :, None] for t5 in t5_text_embeddings], dim=2) + else: + t5_text_embeddings = item["t5_text_embeddings"] + length = len(t5_text_embeddings) + mask[0:length] = 1 + item["t5_text_mask"] = mask + if length < self.max_t5_length: + t5_text_embeddings = F.pad(t5_text_embeddings, (0, 0, 0, self.max_t5_length - length), value=0).float() + item["t5_text_embeddings"] = t5_text_embeddings[: self.max_t5_length] + + item["prompt"] = item["prompt"] + item["__idx__"] = file_idx + item["filename"] = file_paths[file_idx] + # hard coded conds + item["image_size"] = torch.Tensor([self.height, self.width, self.height, self.width]).float() + item["padding_mask"] = torch.zeros((1, self.height, self.width)).float() + + # Augmentation + aug = self.augs[aug_idx] + aug_label = self.aug_labels[aug_idx] + + # video mp4 file + item_fp = os.path.join(self.path, file_paths[file_idx] + ".mp4") + log.info(f"reading from {item_fp}") + video_np, video_meta_data = easy_io.load(item_fp) # (TxHxWx3) + log.info(f"finished reading from {item_fp}") + num_max_frames = self.num_max_frames + if not self.pad_all_videos_to_same_length: + num_max_frames = min(num_max_frames, video_np.shape[0]) + resized_video = np.zeros((num_max_frames, self.height, self.width, 3), dtype=np.uint8) # THWC + log.info(f"loading video: {resized_video.shape} from {item_fp}") + for i in range(min(num_max_frames, video_np.shape[0])): + resized_video[i] = cv2.resize(video_np[i], (self.width, self.height), interpolation=cv2.INTER_AREA) + log.info(f"finished loading video: {resized_video.shape} from {item_fp}") + item["video"] = resized_video.transpose((3, 0, 1, 2)) + + # convert np array to tensor + item["raw_video"] = torch.from_numpy(resized_video.transpose((3, 0, 1, 2))) # CTHW + item["num_frames"] = torch.zeros((1)) + video_np.shape[0] + + item = aug(item) + item["aug_label"] = aug_label + item["sigma_max"] = self.sigma_max_list[sigma_idx] + item["hint_key"], item["control_weight"] = self.control_weight_list[control_weight_idx] + item["fps"] = int(video_meta_data.get("fps")) + return item + + +class PromptVideoItemDataset(PromptOnlyItemDataset): + """ + Dataset for evaluation with prompt and video pairs. + Expects pickle files with 'prompt' field and corresponding MP4 video files. + + Note: + We intentionally do NOT load saved T5 embeddings from disk. + Embeddings will be computed online in the model/callback (see text2world_model and ValLossComputation). + + Args: + path (str): Path to dataset containing .pkl and .mp4 files + start_index (int): Starting index (inclusive) + end_index (int): Ending index (exclusive) + max_t5_length (int): Maximum T5 sequence length + height (int): Video height after resizing + width (int): Video width after resizing + num_video_frames (int): Number of video frames to load + """ + + def __init__( + self, + path: str = "s3://bucket/projects/edify_video/v4/validation/item_dataset/ptbench_video_val", + start_index: int = 0, + end_index: int = 32, + max_t5_length: int = 512, + height: int = 704, + width: int = 1280, + num_video_frames: int = 136, + ): + self.path = path + self.height = height + self.width = width + self.num_video_frames = num_video_frames + self.max_t5_length = max_t5_length + + log.warning( + f"using path: {path} and default s3 credentials in easy_io. " + f"It is user's responsibility to set up the correct credentials." + ) + + # Discover available MP4 files and create file list + file_paths = easy_io.list_dir_or_file(self.path, recursive=True, list_dir=False, suffix=".mp4") + self.file_paths = sorted([path[:-4] for path in file_paths]) # Remove .mp4 extension + + # Apply start_index and end_index to the discovered files + self.file_paths = self.file_paths[start_index:end_index] + + log.info( + f"Build PromptVideoItemDataset with path: {path}, " + f"discovered {len(self.file_paths)} files (after slicing {start_index}:{end_index}), " + f"video shape: ({num_video_frames}, {height}, {width})" + ) + + def __len__(self): + return len(self.file_paths) + + def _getitem(self, idx): + """ + Load a single item with prompt and video. T5 embeddings are NOT loaded. + """ + # Use discovered file paths instead of sequential numbering + file_idx = idx + if file_idx >= len(self.file_paths): + raise IndexError(f"Index {file_idx} out of range for {len(self.file_paths)} files") + + pkl_path = os.path.join(self.path, self.file_paths[file_idx] + ".pkl") + video_path = os.path.join(self.path, self.file_paths[file_idx] + ".mp4") + + # Load pickle data (expects at least 'prompt') + item = easy_io.load(pkl_path) + if item is None: + raise ValueError(f"item is None: {pkl_path}") + # Do not load or pad t5_text_embeddings here; we switch to online computation + + # Load and process video + log.info(f"Loading video from {video_path}") + video_np, video_meta_data = easy_io.load(video_path) # (T, H, W, 3) + log.info(f"Loaded video with shape {video_np.shape} from {video_path}") + + # Resize video frames + num_frames_to_load = min(self.num_video_frames, video_np.shape[0]) + resized_video = np.zeros((self.num_video_frames, self.height, self.width, 3), dtype=np.uint8) + + for i in range(min(num_frames_to_load, video_np.shape[0])): + resized_video[i] = cv2.resize(video_np[i], (self.width, self.height), interpolation=cv2.INTER_AREA) + + # Convert to tensor format (C, T, H, W) + video_cthw = resized_video.transpose((3, 0, 1, 2)) # CTHW + item["video"] = torch.from_numpy(video_cthw) # uint8 tensor, normalized later on GPU + item["raw_video"] = item["video"].clone() + + # Set metadata + # Keep prompt as-is. Online text embedding will use this field + item["prompt"] = item["prompt"] + item["__idx__"] = idx # Use idx instead of item_idx + item["__file__"] = self.file_paths[file_idx] # Store actual file name + item["fps"] = float(video_meta_data.get("fps", 30.0)) + item["image_size"] = torch.Tensor([self.height, self.width, self.height, self.width]).float() + item["padding_mask"] = torch.zeros((1, self.height, self.width)).float() + item["num_frames"] = torch.zeros((1)) + video_np.shape[0] + + log.info( + f"Processed item {self.file_paths[file_idx]}: video shape {item['raw_video'].shape}, prompt length {len(item['prompt'])}" + ) + + return item + + +class PromptLVGItemDataset(PromptOnlyItemDataset): + def __init__( + self, + path: str = "s3://bucket/projects/edify_video/v4/validation/item_dataset/lvg_video_extend_v0_val", + start_index: int = 0, # inclusive + end_index=32, # exclusive + max_t5_length: int = 512, + height=704, + width=1280, + video_length=121, # length of each video clip + num_overlap_frames=4, # number of frames to encode + ): + self.start_index = start_index + self.end_index = end_index + self.path = path + self.height = height + self.width = width + self.video_length = video_length + log.warning( + f"using path: {path} and default s3 credentials in easy_io. It is user's responsibility to set up the correct credentials." + ) + log.info( + f"Build item dataset with path: {path}, and start_index: {start_index}, end_index: {end_index}. video shape THW = ({video_length}, {height}, {width})" + ) + self.max_t5_length = max_t5_length + self.num_overlap_frames = num_overlap_frames + + def _getitem(self, idx): + cur_idx = self.start_index + idx + file_idx = cur_idx + + item_fp = os.path.join(self.path, f"{file_idx:06d}.pkl") + item = easy_io.load(item_fp) + input_image_or_video_ext = item["ext"] + input_image_or_video_path = item_fp.replace(".pkl", f".{input_image_or_video_ext}") + if item is None: + raise ValueError(f"item is None: {item_fp}") + # t5 + mask = torch.LongTensor(self.max_t5_length).zero_() + length = len(item["t5_text_embeddings"]) + mask[0:length] = 1 + item["t5_text_mask"] = mask + if length < self.max_t5_length: + item["t5_text_embeddings"] = F.pad( + item["t5_text_embeddings"], (0, 0, 0, self.max_t5_length - length), value=0 + ).float() + item["t5_text_embeddings"] = item["t5_text_embeddings"][: self.max_t5_length] + + item["prompt"] = item["prompt"] + item["__idx__"] = file_idx + # hard coded conds + item["fps"] = 30.0 + item["image_size"] = torch.Tensor([self.height, self.width, self.height, self.width]).float() + item["padding_mask"] = torch.zeros((1, self.height, self.width)).float() + item["num_frames"] = torch.zeros((1)) + self.video_length + + # video mp4 file + if input_image_or_video_path.endswith(".mp4"): + video_np, video_meta_data = easy_io.load(input_image_or_video_path) # (TxHxWx3) + assert len(video_np) > self.num_overlap_frames, ( + f"to support num_overlap_frames={self.num_overlap_frames}, need at least {self.num_overlap_frames} frames, but current video only have {len(video_np)} frames" + ) + video_np = video_np[-self.num_overlap_frames :] # Select the last num_overlap_frames frames + else: + video_np = np.array(easy_io.load(input_image_or_video_path))[None] # (1xHxWx3) + assert self.num_overlap_frames == 1, ( + f"image data is not supported when num_overlap_frames({self.num_overlap_frames}) > 1, need to set num_overlap_frames=1 or use video input" + ) + + resized_video = np.zeros((self.video_length, self.height, self.width, 3), dtype=np.uint8) + for i in range(min(video_np.shape[0], resized_video.shape[0])): + resized_video[i] = cv2.resize(video_np[i], (self.width, self.height), interpolation=cv2.INTER_AREA) + item["video"] = resized_video.transpose((3, 0, 1, 2)) + return item + + +def calculate_indices(dataset_length: int, world_size: int, rank: int) -> Tuple[int, int, bool]: + """ + Calculate the start and end indices for a given rank in a distributed setting. + + Args: + dataset_length (int): The total length of the dataset. + world_size (int): The number of distributed processes. + rank (int): The rank of the current process. + + Returns: + Tuple[int, int]: A tuple containing the start index (inclusive) and end index (exclusive). + """ + # Calculate the number of samples per rank + samples_per_rank = dataset_length // world_size + remainder = dataset_length % world_size + is_overflow = False + + # Calculate the start and end indices for this rank + start_index = rank * samples_per_rank + min(rank, remainder) + end_index = start_index + samples_per_rank + (1 if rank < remainder else 0) + # take care of corner case where dataset_length is smaller than world size. + if start_index >= dataset_length: # when number of samples are not enough + start_index = 0 + end_index = 1 + is_overflow = True + + return start_index, end_index, is_overflow diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/joint_dataloader.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/joint_dataloader.py new file mode 100644 index 0000000000000000000000000000000000000000..ee28b238ca49182a34cb476e03e479b48af0dee5 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/joint_dataloader.py @@ -0,0 +1,147 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Dict, Union + +import numpy as np +import torch +import webdataset + +from cosmos_policy._src.imaginaire.lazy_config import instantiate + + +class IterativeJointDataLoader(webdataset.WebLoader): + r""" + A joint dataloader that supports loading both images and videos. + """ + + def __init__( + self, dataloaders: Dict[str, Dict[str, Union[torch.utils.data.DataLoader, webdataset.WebLoader, int]]], **kwargs + ): + """ + Initialize the JointDataLoader with multiple datasets. + + Args: + dataloaders: key - dataset_name; value - {"dataloader": dataloader, "ratio": data_ratio} + + Example: + joint_loader = IterativeJointDataLoader( + dataloaders{ + "image_data": { + "dataloader": webdataset.WebLoader(...), + "ratio": 4, + }, + "video_data": { + "dataloader": torch.utils.data.DataLoader(...), + "ratio": 1, + }, + } + ) + """ + self.dataloader_list, self.dataset_name_list, self.data_ratios = [], [], [] + + for dataset_name, dataloader_data in dataloaders.items(): + assert set(dataloader_data.keys()) == {"dataloader", "ratio"}, f"Invalid config: {dataloader_data}" + self.dataset_name_list.append(dataset_name) + self.dataloader_list.append(instantiate(dataloader_data["dataloader"])) + self.data_ratios.append(dataloader_data["ratio"]) + + self.global_id = 0 + self.ratio_sum = sum(self.data_ratios) + + self.data_len = 0 + self.dataloaders = [iter(dataloader) for dataloader in self.dataloader_list] + for data in self.dataloader_list: + self.data_len += len(data) + + def __len__(self) -> int: + return self.data_len + + def __iter__(self): + while True: + data_id = self.global_id % self.ratio_sum + index_id = self._get_dataloader_index(data_id) + curr_dataloader = self.dataloaders[index_id] + output = next(curr_dataloader) + output["dataset_name"] = self.dataset_name_list[index_id] + self.global_id += 1 + del curr_dataloader + yield output + + def _get_dataloader_index(self, data_id): + """Maps global id to the corresponding dataloader index based on ratio.""" + for i, r in enumerate(self.data_ratios): + if data_id < r: + return i + data_id -= r + raise ValueError("Invalid data_id") + + +class RandomJointDataLoader(webdataset.WebLoader): + r""" + A joint dataloader that supports randomly samples batches from multiple datasets. + """ + + # def __init__(self, **kwargs): + def __init__( + self, dataloaders: Dict[str, Dict[str, Union[torch.utils.data.DataLoader, webdataset.WebLoader, int]]] + ): + """ + Initialize the JointDataLoader with multiple datasets. + + Args: + **kwargs: Arbitrary keyword arguments where each key is a string + representing the dataset name, and each value is either + a `webdataset.WebLoader` or `torch.utils.data.DataLoader` + instance. + + Raises: + AssertionError: If any value in kwargs is not an instance of + `webdataset.WebLoader` or `torch.utils.data.DataLoader`. + AssertionError: If any key in kwargs is not a string. + + Example: + joint_loader = JointDataLoader( + images=webdataset.WebLoader(...), + videos=torch.utils.data.DataLoader(...) + ) + """ + self.dataloader_list, self.dataset_name_list, self.data_ratios = [], [], [] + + for dataset_name, dataloader_data in dataloaders.items(): + assert set(dataloader_data.keys()) == {"dataloader", "ratio"}, f"Invalid config: {dataloader_data}" + self.dataset_name_list.append(dataset_name) + self.dataloader_list.append(instantiate(dataloader_data["dataloader"])) + self.data_ratios.append(dataloader_data["ratio"]) + + assert np.isclose(sum(self.data_ratios), 1.0), "Sum of sample probabilities should be equal to 1." + + self.data_len = 0 + self.dataloaders = [iter(dataloader) for dataloader in self.dataloader_list] + for data in self.dataloader_list: + self.data_len += len(data) + + def __len__(self) -> int: + return self.data_len + + def __iter__(self): + while True: + # Sample a random dataset + data_id = int(np.random.choice(len(self.dataloader_list), 1, p=self.data_ratios)[0]) + curr_dataloader = self.dataloaders[data_id] + output = next(curr_dataloader) + output["dataset_name"] = self.dataset_name_list[data_id] + del curr_dataloader + yield output diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/local_datasets/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/local_datasets/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/local_datasets/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/local_datasets/dataset_utils.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/local_datasets/dataset_utils.py new file mode 100644 index 0000000000000000000000000000000000000000..7fa7e7d1e284ff944ff17de2b73e933673595a64 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/local_datasets/dataset_utils.py @@ -0,0 +1,312 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Any, Optional, Union + +import torch +import torchvision.transforms.functional as F +from PIL import Image + + +def obtain_image_size(data_dict: dict, input_keys: list) -> tuple[int, int]: + r"""Function for obtaining the image size from the data dict. + + Args: + data_dict (dict): Input data dict + input_keys (list): List of input keys + Returns: + width (int): Width of the input image + height (int): Height of the input image + """ + + data1 = data_dict[input_keys[0]] + if isinstance(data1, Image.Image): + width, height = data1.size + elif isinstance(data1, torch.Tensor): + height, width = data1.size()[-2:] + else: + raise ValueError("data to random crop should be PIL Image or tensor") + + return width, height + + +def obtain_augmentation_size(data_dict: dict, augmentor_cfg: dict) -> Union[int, tuple]: + r"""Function for obtaining size of the augmentation. + When dealing with multi-aspect ratio dataloaders, we need to + find the augmentation size from the aspect ratio of the data. + + Args: + data_dict (dict): Input data dict + augmentor_cfg (dict): Augmentor config + Returns: + aug_size (int): Size of augmentation + """ + aspect_ratio = data_dict["aspect_ratio"] + aug_size = augmentor_cfg["size"][aspect_ratio] + return aug_size + + +class Augmentor: + def __init__(self, input_keys: list, output_keys: Optional[list] = None, args: Optional[dict] = None) -> None: + r"""Base augmentor class + + Args: + input_keys (list): List of input keys + output_keys (list): List of output keys + args (dict): Arguments associated with the augmentation + """ + self.input_keys = input_keys + self.output_keys = output_keys + self.args = args + + def __call__(self, *args: Any, **kwds: Any) -> Any: + raise ValueError("Augmentor not implemented") + + +class ResizeSmallestSideAspectPreserving(Augmentor): + def __init__(self, input_keys: list, output_keys: Optional[list] = None, args: Optional[dict] = None) -> None: + super().__init__(input_keys, output_keys, args) + + def __call__(self, data_dict: dict) -> dict: + r"""Performs aspect-ratio preserving resizing. + Image is resized to the dimension which has the smaller ratio of (size / target_size). + First we compute (w_img / w_target) and (h_img / h_target) and resize the image + to the dimension that has the smaller of these ratios. + + Args: + data_dict (dict): Input data dict + Returns: + data_dict (dict): Output dict where images are resized + """ + + if self.output_keys is None: + self.output_keys = self.input_keys + assert self.args is not None, "Please specify args in augmentations" + + img_w, img_h = self.args["img_w"], self.args["img_h"] + + orig_w, orig_h = obtain_image_size(data_dict, self.input_keys) + scaling_ratio = max((img_w / orig_w), (img_h / orig_h)) + target_size = (int(scaling_ratio * orig_h + 0.5), int(scaling_ratio * orig_w + 0.5)) + + assert target_size[0] >= img_h and target_size[1] >= img_w, ( + f"Resize error. orig {(orig_w, orig_h)} desire {(img_w, img_h)} compute {target_size}" + ) + + for inp_key, out_key in zip(self.input_keys, self.output_keys): + data_dict[out_key] = transforms_F.resize( # noqa: F821 + data_dict[inp_key], + size=target_size, # type: ignore + interpolation=getattr(self.args, "interpolation", transforms_F.InterpolationMode.BICUBIC), # noqa: F821 + antialias=True, + ) + + if out_key != inp_key: + del data_dict[inp_key] + return data_dict + + +class CenterCrop(Augmentor): + def __init__(self, input_keys: list, output_keys: Optional[list] = None, args: Optional[dict] = None) -> None: + super().__init__(input_keys, output_keys, args) + + def __call__(self, data_dict: dict) -> dict: + r"""Performs center crop. + + Args: + data_dict (dict): Input data dict + Returns: + data_dict (dict): Output dict where images are center cropped. + We also save the cropping parameters in the aug_params dict + so that it will be used by other transforms. + """ + assert (self.args is not None) and ("img_w" in self.args) and ("img_h" in self.args), ( + "Please specify size in args" + ) + + img_w, img_h = self.args["img_w"], self.args["img_h"] + + orig_w, orig_h = obtain_image_size(data_dict, self.input_keys) + for key in self.input_keys: + data_dict[key] = transforms_F.center_crop(data_dict[key], [img_h, img_w]) # noqa: F821 + + # We also add the aug params we use. This will be useful for other transforms + crop_x0 = (orig_w - img_w) // 2 + crop_y0 = (orig_h - img_h) // 2 + cropping_params = { + "resize_w": orig_w, + "resize_h": orig_h, + "crop_x0": crop_x0, + "crop_y0": crop_y0, + "crop_w": img_w, + "crop_h": img_h, + } + + if "aug_params" not in data_dict: + data_dict["aug_params"] = dict() + + data_dict["aug_params"]["cropping"] = cropping_params + data_dict["padding_mask"] = torch.zeros((1, cropping_params["crop_h"], cropping_params["crop_w"])) + return data_dict + + +class Normalize(Augmentor): + def __init__(self, input_keys: list, output_keys: Optional[list] = None, args: Optional[dict] = None) -> None: + super().__init__(input_keys, output_keys, args) + + def __call__(self, data_dict: dict) -> dict: + r"""Performs data normalization. + + Args: + data_dict (dict): Input data dict + Returns: + data_dict (dict): Output dict where images are center cropped. + """ + assert self.args is not None, "Please specify args" + + mean = self.args["mean"] + std = self.args["std"] + + for key in self.input_keys: + if isinstance(data_dict[key], torch.Tensor): + data_dict[key] = data_dict[key].to(dtype=torch.get_default_dtype()).div(255) + else: + data_dict[key] = transforms_F.to_tensor( # noqa: F821 + data_dict[key] + ) # division by 255 is applied in to_tensor() # noqa: F821 + + data_dict[key] = transforms_F.normalize(tensor=data_dict[key], mean=mean, std=std) # noqa: F821 + return data_dict + + +class ResizePreprocess: + def __init__(self, size: tuple[int, int]): + """ + Initialize the preprocessing class with the target size. + Args: + size (tuple): The target height and width as a tuple (height, width). + """ + self.size = size + + def __call__(self, video_frames): + """ + Apply the transformation to each frame in the video. + Args: + video_frames (torch.Tensor): A tensor representing a batch of video frames. + Returns: + torch.Tensor: The transformed video frames. + """ + if video_frames.ndim == 4: + # Resize each frame in the video + resized_frames = torch.stack([F.resize(frame, self.size, antialias=True) for frame in video_frames]) + else: + # Resize a single image + resized_frames = F.resize(video_frames, self.size, antialias=True) + + return resized_frames + + +class ToTensorImage: + """ + Convert tensor data type from uint8 to float, divide value by 255.0 and + permute the dimensions of clip tensor + """ + + def __init__(self): + pass + + def __call__(self, image): + """ + Args: + image (torch.tensor, dtype=torch.uint8): Size is (C, H, W) + Return: + image (torch.tensor, dtype=torch.float): Size is (C, H, W) + """ + return to_tensor_image(image) + + def __repr__(self) -> str: + return self.__class__.__name__ + + +class ToTensorVideo: + """ + Convert tensor data type from uint8 to float, divide value by 255.0 and + permute the dimensions of clip tensor + """ + + def __init__(self): + pass + + def __call__(self, clip): + """ + Args: + clip (torch.tensor, dtype=torch.uint8): Size is (T, C, H, W) + Return: + clip (torch.tensor, dtype=torch.float): Size is (T, C, H, W) + """ + return to_tensor(clip) + + def __repr__(self) -> str: + return self.__class__.__name__ + + +def to_tensor_image(image): + """ + Convert tensor data type from uint8 to float, divide value by 255.0 and + permute the dimensions of image tensor + Args: + image (torch.tensor, dtype=torch.uint8): Size is (T, C, H, W) + Return: + image (torch.tensor, dtype=torch.float): Size is (T, C, H, W) + """ + _is_tensor_image(image) + if not image.dtype == torch.uint8: + raise TypeError("image tensor should have data type uint8. Got %s" % str(image.dtype)) + return image.float() / 255.0 + + +def to_tensor(clip): + """ + Convert tensor data type from uint8 to float, divide value by 255.0 and + permute the dimensions of clip tensor + Args: + clip (torch.tensor, dtype=torch.uint8): Size is (T, C, H, W) + Return: + clip (torch.tensor, dtype=torch.float): Size is (T, C, H, W) + """ + _is_tensor_video_clip(clip) + if not clip.dtype == torch.uint8: + raise TypeError("clip tensor should have data type uint8. Got %s" % str(clip.dtype)) # noqa: UP031 + return clip.float() / 255.0 + + +def _is_tensor_image(image: torch.Tensor) -> bool: + if not torch.is_tensor(image): + raise TypeError("image should be Tensor. Got %s" % type(image)) + + if not image.ndimension() == 3: + raise ValueError("image should be 3D. Got %dD" % image.dim()) + + return True + + +def _is_tensor_video_clip(clip) -> bool: + if not torch.is_tensor(clip): + raise TypeError("clip should be Tensor. Got %s" % type(clip)) # noqa: UP031 + + if not clip.ndimension() == 4: + raise ValueError("clip should be 4D. Got %dD" % clip.dim()) # noqa: UP031 + + return True diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/local_datasets/dataset_video.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/local_datasets/dataset_video.py new file mode 100644 index 0000000000000000000000000000000000000000..ff798a89be89080a5edea8cee421781311f7dc6f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/local_datasets/dataset_video.py @@ -0,0 +1,322 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""Generic video dataset loader for Cosmos Predict2.""" + +import json +import os +import random +import traceback +from pathlib import Path +from typing import Any, Callable, Optional + +import numpy as np +import torch +from decord import VideoReader, cpu +from megatron.core import parallel_state +from torch.utils.data import DataLoader, Dataset, DistributedSampler +from torchvision import transforms as T + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.predict2.datasets.local_datasets.dataset_utils import ResizePreprocess, ToTensorVideo + + +class VideoDataset(Dataset): + def __init__( + self, + dataset_dir: str, + num_frames: int, + video_size: tuple[int, int], + prompt_type: str | None = None, # "long", "short", "medium", or None for auto + caption_format: str = "auto", # "text", "json", or "auto" + video_paths: Optional[list[str]] = None, + ) -> None: + """Dataset class for loading image-text-to-video generation data. + + Args: + dataset_dir (str): Base path to the dataset directory + num_frames (int): Number of frames to load per sequence + video_size (tuple[int, int]): Target size (H,W) for video frames + prompt_type (str | None): Which prompt to use from JSON ("long", "short", "medium"). + If None, uses the first available prompt type. + Only applicable when using JSON format. + caption_format (str): Caption format - "text", "json", or "auto" to detect automatically + + Returns dict with: + - video: RGB frames tensor [T,C,H,W] + - video_name: Dict with episode/frame metadata + """ + + super().__init__() + self.dataset_dir = dataset_dir + self.sequence_length = num_frames + self.prompt_type = prompt_type + self.caption_format = caption_format + + # Determine caption format and directory + self._setup_caption_format() + + video_dir = os.path.join(self.dataset_dir, "videos") + + if video_paths is None: + self.video_paths = [os.path.join(video_dir, f) for f in os.listdir(video_dir) if f.endswith(".mp4")] + self.video_paths = sorted(self.video_paths) + else: + self.video_paths = video_paths + log.info(f"{len(self.video_paths)} videos in total") + + self.num_failed_loads = 0 + self.preprocess = T.Compose([ToTensorVideo(), ResizePreprocess((video_size[0], video_size[1]))]) + + def __str__(self) -> str: + return f"{len(self.video_paths)} samples from {self.dataset_dir}" + + def __len__(self) -> int: + return len(self.video_paths) + + def _load_video(self, video_path: str) -> tuple[np.ndarray, float]: + vr = VideoReader(video_path, ctx=cpu(0), num_threads=2) + total_frames = len(vr) + if total_frames < self.sequence_length: + raise ValueError( + f"Video {video_path} has only {total_frames} frames, " + f"at least {self.sequence_length} frames are required." + ) + + # randomly sample a sequence of frames + max_start_idx = total_frames - self.sequence_length + start_frame = np.random.randint(0, max_start_idx) + end_frame = start_frame + self.sequence_length + frame_ids = np.arange(start_frame, end_frame).tolist() + + frame_data = vr.get_batch(frame_ids).asnumpy() + vr.seek(0) # set video reader point back to 0 to clean up cache + + try: + fps = vr.get_avg_fps() + except Exception: # failed to read FPS, assume it is 16 + fps = 16 + del vr # delete the reader to avoid memory leak + return frame_data, fps + + def _setup_caption_format(self) -> None: + """Determine the caption format and set up the caption directory.""" + metas_dir = os.path.join(self.dataset_dir, "metas") + captions_dir = os.path.join(self.dataset_dir, "captions") + + if self.caption_format == "auto": + # Auto-detect based on directory existence + if os.path.exists(captions_dir) and any(f.endswith(".json") for f in os.listdir(captions_dir)): + self.caption_format = "json" + self.caption_dir = captions_dir + elif os.path.exists(metas_dir) and any(f.endswith(".txt") for f in os.listdir(metas_dir)): + self.caption_format = "text" + self.caption_dir = metas_dir + else: + raise ValueError( + f"Could not auto-detect caption format. Neither 'metas/*.txt' nor 'captions/*.json' found in {self.dataset_dir}" + ) + elif self.caption_format == "json": + if not os.path.exists(captions_dir): + raise ValueError(f"JSON format specified but 'captions' directory not found in {self.dataset_dir}") + self.caption_dir = captions_dir + elif self.caption_format == "text": + if not os.path.exists(metas_dir): + raise ValueError(f"Text format specified but 'metas' directory not found in {self.dataset_dir}") + self.caption_dir = metas_dir + else: + raise ValueError(f"Invalid caption_format: {self.caption_format}. Must be 'text', 'json', or 'auto'") + + def _load_text(self, text_source: Path) -> str: + """Load text caption from file.""" + try: + return text_source.read_text().strip() + except Exception as e: + log.warning(f"Failed to read caption file {text_source}: {e}") + return "" + + def _load_json_caption(self, json_path: Path) -> str: + """Load caption from JSON file with prompt type selection.""" + try: + with open(json_path, "r") as f: + content = f.read() + # Handle JSON that might not have top-level object + if not content.strip().startswith("{"): + # Wrap in object if needed + data = json.loads("{" + content + "}") + else: + data = json.loads(content) + + # Get the first model's captions (e.g., "qwen3_vl_30b_a3b") + model_key = next(iter(data.keys())) + captions = data[model_key] + + if self.prompt_type: + # Use specified prompt type + if self.prompt_type in captions: + return captions[self.prompt_type] + else: + log.warning( + f"Prompt type '{self.prompt_type}' not found in {json_path}. " + f"Available: {list(captions.keys())}. Using first available." + ) + + # Use first available prompt type + first_prompt = next(iter(captions.values())) + return first_prompt + + except Exception as e: + log.warning(f"Failed to read JSON caption file {json_path}: {e}") + return "" + + def _get_frames(self, video_path: str) -> tuple[torch.Tensor, float]: + frames, fps = self._load_video(video_path) + frames = frames.astype(np.uint8) + frames = torch.from_numpy(frames).permute(0, 3, 1, 2) # [T, C, H, W] + frames = self.preprocess(frames) + frames = torch.clamp(frames * 255.0, 0, 255).to(torch.uint8) + return frames, fps + + def __getitem__(self, index: int) -> dict | Any: + try: + data = dict() + video, fps = self._get_frames(self.video_paths[index]) + video = video.permute(1, 0, 2, 3) # Rearrange from [T, C, H, W] to [C, T, H, W] + + # Load caption based on format + video_path = self.video_paths[index] + video_basename = os.path.basename(video_path).replace(".mp4", "") + + if self.caption_format == "json": + caption_path = os.path.join(self.caption_dir, f"{video_basename}.json") + caption = self._load_json_caption(Path(caption_path)) + else: # text format + caption_path = os.path.join(self.caption_dir, f"{video_basename}.txt") + caption = self._load_text(Path(caption_path)) + + data["video"] = video + data["ai_caption"] = caption + + _, _, h, w = video.shape + + data["fps"] = fps + data["image_size"] = torch.tensor([h, w, h, w]) + data["num_frames"] = self.sequence_length + data["padding_mask"] = torch.zeros(1, h, w) + + return data + except Exception as e: + self.num_failed_loads += 1 + log.warning( + f"Failed to load video {self.video_paths[index]} (total failures: {self.num_failed_loads}): {e}\n" + f"{traceback.format_exc()}", + rank0_only=False, + ) + # Randomly sample another video + return self[np.random.randint(len(self.video_paths))] + + +def get_generic_dataloader( + dataset: Dataset, + batch_size: int = 1, + sampler: Optional[Any] = None, + num_workers: int = 0, + pin_memory: bool = False, + drop_last: bool = False, + prefetch_factor: Optional[int] = None, + persistent_workers: bool = False, + collate_fn: Optional[Callable] = None, + **kwargs, # Ignore extra arguments +) -> DataLoader: + """Create DataLoader with commonly used parameters. + + Args: + dataset: Dataset instance + batch_size: Batch size + sampler: Optional sampler for data loading + num_workers: Number of worker processes + pin_memory: Pin memory for CUDA transfer + drop_last: Drop incomplete last batch + prefetch_factor: Number of batches to prefetch per worker + persistent_workers: Keep workers alive between epochs + collate_fn: Custom collate function + **kwargs: Extra arguments (ignored) + + Returns: + Configured DataLoader + """ + return DataLoader( + dataset=dataset, + batch_size=batch_size, + shuffle=False, # False when using sampler + sampler=sampler, + num_workers=num_workers, + pin_memory=pin_memory, + drop_last=drop_last, + prefetch_factor=prefetch_factor, + persistent_workers=persistent_workers, + collate_fn=collate_fn, + ) + + +def get_sampler(dataset) -> DistributedSampler: + """Create a distributed sampler for the dataset.""" + return DistributedSampler( + dataset, + num_replicas=parallel_state.get_data_parallel_world_size(), + rank=parallel_state.get_data_parallel_rank(), + shuffle=True, + seed=0, + ) + + +def get_train_val_dataloaders( + dataset_path: str, val_percentage: float, seed: int, video_size: tuple[int, int] = (704, 1280) +): + video_dir = os.path.join(dataset_path, "videos") + if not os.path.exists(video_dir): + log.debug(f"Dataset path {dataset_path} does not exist, returning empty dataloaders") + return dict(), dict() + video_paths = [os.path.join(video_dir, f) for f in os.listdir(video_dir) if f.endswith(".mp4")] + random.seed(seed) + random.shuffle(video_paths) + + cutoff = int(len(video_paths) * val_percentage) + val_video_paths = video_paths[:cutoff] + train_video_paths = video_paths[cutoff:] + + def get_dataset(video_paths): + return L(VideoDataset)( + video_paths=video_paths, + num_frames=93, + video_size=video_size, + dataset_dir=dataset_path, + ) + + ipn_hand_train_dataset = get_dataset(train_video_paths) + ipn_hand_val_dataset = get_dataset(val_video_paths) + + def get_dataloader(dataset): + return L(get_generic_dataloader)( + dataset=dataset, + sampler=L(get_sampler)(dataset=dataset), + batch_size=1, + drop_last=True, + num_workers=4, + pin_memory=True, + ) + + return get_dataloader(ipn_hand_train_dataset), get_dataloader(ipn_hand_val_dataset) diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/utils.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/utils.py new file mode 100644 index 0000000000000000000000000000000000000000..515346171d5cb17e8fa572dc1b26c48b6630c42f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/utils.py @@ -0,0 +1,83 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import re +from typing import List, Tuple + +IMAGE_RES_SIZE_INFO: dict[str, dict[str, tuple[int, int]]] = { + "1080": { + "1,1": (1024, 1024), + "4,3": (1440, 1056), + "3,4": (1056, 1440), + "16,9": (1920, 1056), + "9,16": (1056, 1920), + }, + # "1024": {"1,1": (1024, 1024), "4,3": (1280, 1024), "3,4": (1024, 1280), "16,9": (1280, 768), "9,16": (768, 1280)}, + "1024": {"1,1": (1024, 1024), "4,3": (1168, 880), "3,4": (880, 1168), "16,9": (1360, 768), "9,16": (768, 1360)}, + "720": {"1,1": (960, 960), "4,3": (960, 704), "3,4": (704, 960), "16,9": (1280, 704), "9,16": (704, 1280)}, + "512": {"1,1": (512, 512), "4,3": (640, 512), "3,4": (512, 640), "16,9": (640, 384), "9,16": (384, 640)}, + "480": {"1,1": (480, 480), "4,3": (640, 480), "3,4": (480, 640), "16,9": (768, 432), "9,16": (432, 768)}, + "480p": {"1,1": (640, 640), "4,3": (640, 480), "3,4": (480, 640), "16,9": (832, 480), "9,16": (480, 832)}, + "720robocasa": {"1,1": (720, 720), "4,3": (960, 720), "3,4": (720, 960), "16,9": (1280, 720), "9,16": (720, 1280)}, + "256": { + "1,1": (256, 256), + "4,3": (320, 256), + "3,4": (256, 320), + "16,9": (320, 192), + "9,16": (192, 320), + }, +} + + +VIDEO_RES_SIZE_INFO: dict[str, tuple[int, int]] = { + "1080": { + "1,1": (1024, 1024), + "4,3": (1440, 1056), + "3,4": (1056, 1440), + "16,9": (1920, 1056), + "9,16": (1056, 1920), + }, + "1024": {"1,1": (1024, 1024), "4,3": (1280, 1024), "3,4": (1024, 1280), "16,9": (1280, 768), "9,16": (768, 1280)}, + "720": {"1,1": (960, 960), "4,3": (960, 704), "3,4": (704, 960), "16,9": (1280, 704), "9,16": (704, 1280)}, + "512": {"1,1": (512, 512), "4,3": (640, 512), "3,4": (512, 640), "16,9": (640, 384), "9,16": (384, 640)}, + "480": {"1,1": (480, 480), "4,3": (640, 480), "3,4": (480, 640), "16,9": (768, 432), "9,16": (432, 768)}, + # 720, 1280 is Wan2.1 specs + "480p": {"1,1": (640, 640), "4,3": (640, 480), "3,4": (480, 640), "16,9": (832, 480), "9,16": (480, 832)}, + "720p": {"1,1": (960, 960), "4,3": (960, 720), "3,4": (720, 960), "16,9": (1280, 720), "9,16": (720, 1280)}, + "720robocasa": {"1,1": (720, 720), "4,3": (960, 720), "3,4": (720, 960), "16,9": (1280, 720), "9,16": (720, 1280)}, + "256": { + "1,1": (256, 256), + "4,3": (320, 256), + "3,4": (256, 320), + "16,9": (320, 192), + "9,16": (192, 320), + }, +} + + +def get_aspect_ratios_from_wdinfos(wdinfos: list[str]) -> list[str]: + aspect_ratios = [] + for wdinfo in wdinfos: + aspect_ratio_match = re.search(r"aspect_ratio_(\d+_\d+)", wdinfo) + aspect_ratios.append(aspect_ratio_match.group(1)) + + return aspect_ratios + + +def get_wdinfos_w_aspect_ratio(wdinfos: list[str]) -> List[Tuple[str, str]]: + aspect_ratios = get_aspect_ratios_from_wdinfos(wdinfos) + + # return a list of (wdinfo_path, aspect_ratio) pairs + return [(wdinfo, aspect_ratio.replace("_", ",")) for wdinfo, aspect_ratio in zip(wdinfos, aspect_ratios)] diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/watchdog.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/watchdog.py new file mode 100644 index 0000000000000000000000000000000000000000..a4333c1070e47d8f9268bd6b4b128c648152bf82 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/watchdog.py @@ -0,0 +1,282 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import contextlib +import threading +import time +from collections.abc import Iterator +from typing import Any, Optional, Union + +from cosmos_policy._src.imaginaire.utils import log + + +class OperationWatchdog: + """A watchdog that monitors operations for hangs and collects performance statistics. + + This class provides a mechanism to detect when operations take longer than expected, + which can help identify potential deadlocks or performance issues. It also collects + statistics about operations to help identify bottlenecks. + + Attributes: + warning_threshold: Time in seconds before warning about potential hangs. + check_interval: Time in seconds between checks for hung operations. + verbose_interval: Time in seconds between verbose logging. + """ + + def __init__( + self, + warning_threshold: int = 600, + check_interval: int = 30, + verbose_interval: int = -1, + name: str = "OperationWatchdog", + ) -> None: + """Initialize the watchdog. + + Args: + warning_threshold: Time in seconds before warning about potential hangs. + Defaults to 600 (10 minutes). + check_interval: Time in seconds between checks. Defaults to 30. + verbose_interval: Time in seconds between verbose logging. Defaults to -1 (disabled). + """ + self._warning_threshold = warning_threshold + self._check_interval = check_interval + self._verbose_interval = verbose_interval + self._name = name + self._ops: dict[str, dict[str, Any]] = {} # Active operations + self._stats: dict[str, dict[str, Union[int, float]]] = {} # Operation statistics + self._lock = threading.Lock() + self._stop_event = threading.Event() + self._thread: Optional[threading.Thread] = None + + # Auto-start the monitoring thread + self.start() + + def start(self) -> None: + """Start the watchdog monitoring thread. + + If the thread is already running, this method does nothing. + """ + if self._thread is None or not self._thread.is_alive(): + self._stop_event.clear() + self._thread = threading.Thread(target=self._monitor_loop, daemon=True, name=f"{self._name}_monitor_thread") + self._thread.start() + log.debug(f"[{self._name}] Watchdog monitoring thread started") + + def stop(self) -> None: + """Stop the watchdog monitoring thread. + + This method is typically called when shutting down the application. + """ + self._stop_event.set() + if self._thread and self._thread.is_alive(): + self._thread.join(timeout=1.0) + if self._thread.is_alive(): + log.warning(f"[{self._name}] Watchdog thread did not terminate within timeout", rank0_only=False) + else: + log.debug(f"[{self._name}] Watchdog monitoring thread stopped", rank0_only=False) + + @contextlib.contextmanager + def watch(self, operation_name: str, description: str = "", verbose_first_n: int = -1) -> Iterator[None]: + """Context manager for monitoring an operation. + + This is the primary interface for using the watchdog. It automatically + tracks the start and end time of the operation and updates statistics. + + Args: + operation_name: Name/type of the operation to monitor. + description: Optional description providing more context. + verbose_first_n: If positive, print verbose logs for first N operations for operation_name. + + Yields: + None + + Example: + with watchdog.watch("data_fetch", "Fetching user data"): + data = fetch_user_data(user_id) + """ + # Create unique ID for this specific operation instance + op_id = f"{operation_name}_{int(time.time() * 1000)}" + start_time = time.time() + + # Register operation + with self._lock: + self._ops[op_id] = { + "name": operation_name, + "desc": description or operation_name, + "start": start_time, + "update": start_time, + "warned": False, + } + + try: + # Yield control back to the with-block + yield + finally: + # Calculate duration and remove from active operations + duration = time.time() - start_time + if duration > self._warning_threshold: + log.warning( + f"[{self._name}] Operation id: {op_id}, name: '{operation_name}' took {duration:.2f}s", + rank0_only=False, + ) + + with self._lock: + # Remove from active operations + if op_id in self._ops: + del self._ops[op_id] + + # Update statistics + if operation_name not in self._stats: + self._stats[operation_name] = {"count": 0, "total_time": 0.0, "max_time": 0.0, "last_time": 0.0} + + stats = self._stats[operation_name] + stats["count"] += 1 + stats["total_time"] += duration + stats["max_time"] = max(stats["max_time"], duration) + stats["last_time"] = duration + + if verbose_first_n > 0 and stats["count"] <= verbose_first_n: + avg_time = stats["total_time"] / stats["count"] + log.info( + f"[{self._name}] name: '{operation_name}', count {stats['count']} / {verbose_first_n} took {duration:.2f}s. avg {avg_time:.2f}s, max {stats['max_time']:.2f}s, last {stats['last_time']:.2f}s", + rank0_only=False, + ) + + def heartbeat(self, operation_name: str) -> None: + """Send a heartbeat for all operations of a given type. + + Use this inside long operations to prevent false warnings. + + Args: + operation_name: The operation type to update. + """ + current_time = time.time() + with self._lock: + for _op_id, op in self._ops.items(): + if op["name"] == operation_name: + op["update"] = current_time + op["warned"] = False + + def get_stats(self, operation_name: Optional[str] = None) -> dict[str, Any]: + """Get statistics for operations. + + Args: + operation_name: Get stats for specific operation, or None for all. + + Returns: + Dictionary of operation statistics. For each operation, includes: + - count: Number of completed operations + - total_time: Total time spent in this operation type + - max_time: Maximum time spent in a single operation + - last_time: Time spent in the most recent operation + - avg_time: Average time per operation (if count > 0) + """ + with self._lock: + if operation_name: + if operation_name in self._stats: + stats = self._stats[operation_name].copy() + if stats["count"] > 0: + stats["avg_time"] = stats["total_time"] / stats["count"] + return stats + return {} + + # Return all stats + result = {} + for name, stats in self._stats.items(): + result[name] = stats.copy() + if stats["count"] > 0: + result[name]["avg_time"] = stats["total_time"] / stats["count"] + return result + + def print_stats(self) -> None: + """Print statistics for all operations. + + This is a convenience method that logs statistics at INFO level. + """ + stats = self.get_stats() + if not stats: + log.info(f"[{self._name}] No operation statistics available", rank0_only=False) + return + + log.info(f"[{self._name}] Operation Statistics:", rank0_only=False) + + for name, s in stats.items(): + if s["count"] > 0: + avg = s["total_time"] / s["count"] + log.info( + f"[{self._name}] {name}: count={s['count']}, " + f"avg={avg:.2f}s, max={s['max_time']:.2f}s, " + f"last={s['last_time']:.2f}s", + rank0_only=False, + ) + + def list_active_operations(self) -> dict[str, Any]: + """List all currently active operations. + + Returns: + Dictionary mapping operation IDs to information about active operations. + """ + with self._lock: + current_time = time.time() + result = {} + + for op_id, op in self._ops.items(): + result[op_id] = { + "name": op["name"], + "description": op["desc"], + "running_time": current_time - op["start"], + "time_since_update": current_time - op["update"], + } + + return result + + def reset_stats(self) -> None: + """Reset all operation statistics. + + This clears all accumulated statistics but does not affect active operations. + """ + with self._lock: + self._stats.clear() + + def _monitor_loop(self) -> None: + """Monitor registered operations for hangs. + + This is an internal method that runs in a separate thread. + """ + last_verbose_time = 0 + while not self._stop_event.is_set(): + now = time.time() + + with self._lock: + for _, op in list(self._ops.items()): + # Check if operation is hung + elapsed = now - op["update"] + if elapsed > self._warning_threshold and not op["warned"]: + total_time = now - op["start"] + log.warning( + f"[{self._name}] POTENTIAL HANG: '{op['name']}' ({op['desc']}) " + f"has been running for {total_time:.1f}s total, " + f"with {elapsed:.1f}s since last update", + rank0_only=False, + ) + op["warned"] = True + + # Sleep between checks + if self._verbose_interval > 0 and now - last_verbose_time > self._verbose_interval: + self.print_stats() + last_verbose_time = now + self._stop_event.wait(self._check_interval) diff --git a/REGEN-main/cosmos_policy/_src/predict2/datasets/webdataset.py b/REGEN-main/cosmos_policy/_src/predict2/datasets/webdataset.py new file mode 100644 index 0000000000000000000000000000000000000000..121a1fc3e949373a704992d49da2322b90be2fb0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/datasets/webdataset.py @@ -0,0 +1,118 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Callable, Optional + +import omegaconf +import webdataset as wds +from webdataset import filters +from webdataset.handlers import reraise_exception + +from cosmos_policy._src.imaginaire.datasets.webdataset.config.schema import DatasetConfig +from cosmos_policy._src.imaginaire.datasets.webdataset.utils.iterators import WebDataset +from cosmos_policy._src.imaginaire.datasets.webdataset.utils.misc import ( + remove_extensions_from_keys, + skip_keys, + update_url, +) +from cosmos_policy._src.imaginaire.datasets.webdataset.webdataset import Dataset as BaseDataset +from cosmos_policy._src.imaginaire.utils import log + + +class Dataset(BaseDataset): + def __init__( + self, + config: DatasetConfig, + handler: Callable = reraise_exception, + decoder_handler: Optional[Callable] = None, + detshuffle: bool = False, + ): + r"""Webdataloader class + + Args: + config: Dataset config + handler (Callable): Error handler for webdataset class + decoder_handler (Callable): Error handler during decoding + """ + super().__init__(config=config, handler=handler) + self.decoder_handler = decoder_handler + self.detshuffle = detshuffle + + def build_dataset(self, **kwargs) -> WebDataset: + r""" + Build the dataset object. + The function only diffs from BaseDataset.build_dataset by only adding the decoder_handler to the WebDataset object. + """ + tar_list = self.wdinfo.tar_files + num_tars = len(tar_list) + assert num_tars > 0, "Did not find any data." + + shuffle_buffer_size = getattr(self.config, "buffer_size", self.wdinfo.chunk_size) + + # update distributor urls and chunk size + distributor_fn = self.config.distributor + + distributor_fn.set_urls(tar_list) + distributor_fn.set_chunk_size(self.wdinfo.chunk_size) + + dataset = WebDataset( + distributor_fn, + load_from_object_store=self.use_object_store, + easy_io_backend=self.easy_io_backend, + s3_bucket_name=self.bucket, + streaming_download=self.streaming_download, + handler=self.handler, + ) + + # Creating a shuffle buffer + if self.detshuffle: + dataset.append(filters.detshuffle(shuffle_buffer_size)) + else: + dataset.append(wds.shuffle(shuffle_buffer_size)) + + # Adding decoders + # Decoders are functions that decode the input IO stream + decoder_list = getattr(self.config, "decoders", []) + decoder_functions = [] + for decoder in decoder_list: + # If the specified decoder is a string, use the webdataset decoder + # If its a callable function, use the defined function to decode data + assert isinstance(decoder, str) or callable(decoder), "Decoder should either be callable or a str" + decoder_functions.append(decoder) + dataset.append(wds.decode(*decoder_functions, handler=self.decoder_handler)) + + # After the decoders are added, remove extension from the keys + # Extensions in the data keys are needed for auto-detection of decoders in webdataset. + if self.config.remove_extension_from_keys: + dataset.append(remove_extensions_from_keys) + + # Function to skip keys + dataset.append(skip_keys) + # Building augmentors + augmentor_cfg = getattr(self.config, "augmentation", None) + assert isinstance(augmentor_cfg, (dict, omegaconf.dictconfig.DictConfig)), ( + f"getting type: {type(augmentor_cfg)}" + ) + augmentation_fn = self.build_data_augmentor(augmentor_cfg) + dataset.append(augmentation_fn) + + # Updates URL names so that the collate function can handle + dataset.append(update_url) + + dataset.total_images = self.wdinfo.total_key_count # type: ignore + log.info("Total number of training shards: %d" % num_tars) + log.info("Total training key count: %d" % dataset.total_images) # type: ignore + + return dataset diff --git a/REGEN-main/cosmos_policy/_src/predict2/functional/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/functional/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/functional/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/inference/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/inference/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/inference/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/inference/get_t5_emb.py b/REGEN-main/cosmos_policy/_src/predict2/inference/get_t5_emb.py new file mode 100644 index 0000000000000000000000000000000000000000..7d86da4e5a60a314f7f5e82e10374b4ce1d8c25f --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/inference/get_t5_emb.py @@ -0,0 +1,174 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import ClassVar, List, Optional, Tuple, Union + +import attrs +import torch +import transformers +from transformers import T5EncoderModel, T5TokenizerFast + +transformers.logging.set_verbosity_error() + +T5_MODEL_DIR = "checkpoints/google-t5/t5-11b" + + +class CosmosT5TextEncoder(torch.nn.Module): + """Handles T5 text encoding operations.""" + + def __init__( + self, model_name: str = "google-t5/t5-11b", device: str = "cuda", cache_dir=None, local_files_only=False + ): + """Initializes the T5 tokenizer and encoder. + + Args: + model_name: The name of the T5 model to use. + device: The device to use for computations. + """ + super().__init__() + self.tokenizer = T5TokenizerFast.from_pretrained( + model_name, cache_dir=cache_dir, local_files_only=local_files_only + ) + self.text_encoder = T5EncoderModel.from_pretrained( + model_name, cache_dir=cache_dir, local_files_only=local_files_only + ).to(device) + self.text_encoder.eval() + self.device = device + + @torch.inference_mode() + def encode_prompts( + self, prompts: Union[str, List[str]], max_length: int = 512, return_mask: bool = False + ) -> Union[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]: + """Encodes text prompts into hidden state representations using a T5 encoder. + + This function tokenizes the input prompts, processes them through a T5 text encoder, + and returns the last hidden states. The encoded outputs beyond the actual sequence + length are zero-padded. All prompts in a batch are padded to max_length. + + Args: + prompts: Input text to encode. Can be a single string or a list of strings. + max_length: Maximum sequence length for tokenization and padding. Longer + sequences will be truncated. Defaults to 512. + return_mask: If True, returns the attention mask along with encoded text. + Defaults to False. + + Returns: + If return_mask is False: + torch.Tensor: Encoded text embeddings of shape (batch_size, max_length, hidden_size). + If return_mask is True: + tuple[torch.Tensor, torch.Tensor]: A tuple containing: + - Encoded text embeddings of shape (batch_size, max_length, hidden_size) + - Attention mask of shape (batch_size, max_length) as boolean tensor + + Raises: + ValueError: If the input prompts list is empty. + + Example: + >>> encoder = CosmosT5TextEncoder() + >>> prompts = ["Hello world", "Another example"] + >>> embeddings = encoder.encode_prompts(prompts, max_length=128) + """ + if isinstance(prompts, str): + prompts = [prompts] + + if not prompts: + raise ValueError("The input prompt list is empty.") + + batch_encoding = self.tokenizer.batch_encode_plus( + prompts, + return_tensors="pt", + truncation=True, + padding="max_length", + max_length=max_length, + return_length=True, + return_offsets_mapping=False, + ) + + input_ids = batch_encoding.input_ids.to(self.device) + attn_mask = batch_encoding.attention_mask.to(self.device) + + outputs = self.text_encoder(input_ids=input_ids, attention_mask=attn_mask) + + encoded_text = outputs.last_hidden_state + lengths = attn_mask.sum(dim=1).cpu() + + for batch_id in range(encoded_text.shape[0]): + encoded_text[batch_id][lengths[batch_id] :] = 0 + + if return_mask: + return encoded_text, attn_mask.bool() + return encoded_text + + +@attrs.define(slots=False) +class CosmosT5TextEncoderConfig: + """ + Config for the T5 text encoder model + """ + + CKPT_PATH: ClassVar[str] = T5_MODEL_DIR + NUM_TOKENS: ClassVar[int] = 512 + EMBED_DIM: ClassVar[int] = 1024 + + ckpt_path: str = CKPT_PATH + num_tokens: int = NUM_TOKENS + embed_dim: int = EMBED_DIM + + +cosmos_encoder: Optional[CosmosT5TextEncoder] = None + + +def get_text_embedding( + prompts: Union[str, List[str]], + encoder: Optional[CosmosT5TextEncoder] = None, + device: str = "cuda", + max_length: int = 512, + return_mask: bool = False, + cache_dir: str = None, + local_files_only: str = False, + text_encoder_class: str = "T5", +) -> torch.Tensor: + """Encodes text prompts into T5 embeddings. + + Args: + prompts: A single text prompt or a list of text prompts. + encoder: An optional CosmosT5TextEncoder instance. If None, a global + instance will be created or reused. + device: The device to use for computations. + max_length: The maximum length for the padded embedding. + text_encoder_class: The class of the text encoder to use. + + Returns: + A tensor of T5 embeddings. + """ + assert text_encoder_class == "T5", f"text_encoder_class {text_encoder_class} is not supported" + + global cosmos_encoder + + if encoder is None: + if cosmos_encoder is None: + cosmos_encoder = CosmosT5TextEncoder(device=device, cache_dir=cache_dir, local_files_only=local_files_only) + encoder = cosmos_encoder + + encoder.text_encoder.to(device) + + if isinstance(prompts, str): + prompts = [prompts] + + return encoder.encode_prompts( + prompts, + max_length=max_length, + return_mask=return_mask, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/inference/get_t5_emb_test.py b/REGEN-main/cosmos_policy/_src/predict2/inference/get_t5_emb_test.py new file mode 100644 index 0000000000000000000000000000000000000000..2fd7c1d9e8513ae4eac9b1521ef493d5c97826e7 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/inference/get_t5_emb_test.py @@ -0,0 +1,78 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import pytest +import torch + +from cosmos_policy._src.predict2.inference.get_t5_emb import CosmosT5TextEncoder, get_text_embedding + + +@pytest.fixture(scope="module") +def device(): + return "cuda" if torch.cuda.is_available() else "cpu" + + +@pytest.fixture(scope="module") +def encoder(device): + return CosmosT5TextEncoder(device=device) + + +@pytest.mark.L2 +def test_single_prompt(encoder, device): + prompt = "This is a test prompt." + embedding = get_text_embedding(prompt, encoder=encoder, device=device) + assert embedding.shape == (1, 512, 1024) + + +@pytest.mark.L2 +def test_multiple_prompts(encoder, device): + prompts = ["First prompt.", "Second prompt.", "Third prompt."] + embeddings = get_text_embedding(prompts, encoder=encoder, device=device) + assert embeddings.shape == (3, 512, 1024) + + +@pytest.mark.L2 +def test_global_encoder(device): + prompt = "Testing global encoder." + embedding1 = get_text_embedding(prompt, device=device) + embedding2 = get_text_embedding(prompt, device=device) + assert torch.allclose(embedding1, embedding2) + + +@pytest.mark.L2 +def test_custom_max_length(encoder, device): + prompt = "Short prompt." + max_length = 20 + embedding = get_text_embedding(prompt, encoder=encoder, device=device, max_length=max_length) + assert embedding.shape == (1, max_length, 1024) + + +@pytest.mark.L2 +def test_encoder_device(encoder): + assert encoder.device in ["cuda", "cpu"] + assert next(encoder.text_encoder.parameters()).device.type == encoder.device + + +@pytest.mark.L2 +def test_empty_prompt_list(encoder, device): + with pytest.raises(ValueError, match="The input prompt list is empty."): + get_text_embedding([], encoder=encoder, device=device) + + +@pytest.mark.L2 +def test_long_prompt(encoder, device): + long_prompt = "This is a very long prompt. " * 100 + embedding = get_text_embedding(long_prompt, encoder=encoder, device=device) + assert embedding.shape == (1, 512, 1024) # Should be truncated to max_length diff --git a/REGEN-main/cosmos_policy/_src/predict2/inference/get_umt5_emb.py b/REGEN-main/cosmos_policy/_src/predict2/inference/get_umt5_emb.py new file mode 100644 index 0000000000000000000000000000000000000000..b2e9420a044f7b0ae916362570bdbbf4f7e2f344 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/inference/get_umt5_emb.py @@ -0,0 +1,615 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import html +import math +import string +from typing import List, Optional, Union + +import ftfy +import regex as re +import torch +import torch.distributed.checkpoint as dcp +import torch.nn as nn +import torch.nn.functional as F +from torch.distributed.checkpoint import FileSystemReader +from torch.distributed.checkpoint.default_planner import DefaultLoadPlanner +from torch.distributed.checkpoint.state_dict import StateDictOptions, get_model_state_dict, set_model_state_dict +from transformers import AutoTokenizer + +from cosmos_policy._src.imaginaire.checkpointer.s3_filesystem import S3StorageReader +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + +""" +Usage: + pytest -s projects/cosmos/diffusion/v2/networks/umt5.py --all +TODO: + + - [ ] using flex attention. + - [ ] FSDP shard +""" + + +def basic_clean(text): + text = ftfy.fix_text(text) + text = html.unescape(html.unescape(text)) + return text.strip() + + +def whitespace_clean(text): + text = re.sub(r"\s+", " ", text) + text = text.strip() + return text + + +def canonicalize(text, keep_punctuation_exact_string=None): + text = text.replace("_", " ") + if keep_punctuation_exact_string: + text = keep_punctuation_exact_string.join( + part.translate(str.maketrans("", "", string.punctuation)) + for part in text.split(keep_punctuation_exact_string) + ) + else: + text = text.translate(str.maketrans("", "", string.punctuation)) + text = text.lower() + text = re.sub(r"\s+", " ", text) + return text.strip() + + +class HuggingfaceTokenizer: + def __init__(self, name, seq_len=None, clean=None, **kwargs): + assert clean in (None, "whitespace", "lower", "canonicalize") + self.name = name + self.seq_len = seq_len + self.clean = clean + + # init tokenizer + self.tokenizer = AutoTokenizer.from_pretrained(name, **kwargs) + self.vocab_size = self.tokenizer.vocab_size + + def __call__(self, sequence, **kwargs): + return_mask = kwargs.pop("return_mask", False) + + # arguments + _kwargs = {"return_tensors": "pt"} + if self.seq_len is not None: + _kwargs.update({"padding": "max_length", "truncation": True, "max_length": self.seq_len}) + _kwargs.update(**kwargs) + + # tokenization + if isinstance(sequence, str): + sequence = [sequence] + if self.clean: + sequence = [self._clean(u) for u in sequence] + ids = self.tokenizer(sequence, **_kwargs) + + # output + if return_mask: + return ids.input_ids, ids.attention_mask + else: + return ids.input_ids + + def _clean(self, text): + if self.clean == "whitespace": + text = whitespace_clean(basic_clean(text)) + elif self.clean == "lower": + text = whitespace_clean(basic_clean(text)).lower() + elif self.clean == "canonicalize": + text = canonicalize(basic_clean(text)) + return text + + +def fp16_clamp(x): + if x.dtype == torch.float16 and torch.isinf(x).any(): + clamp = torch.finfo(x.dtype).max - 1000 + x = torch.clamp(x, min=-clamp, max=clamp) + return x + + +def init_weights(m): + if isinstance(m, T5LayerNorm): + nn.init.ones_(m.weight) + elif isinstance(m, T5Model): + nn.init.normal_(m.token_embedding.weight, std=1.0) + elif isinstance(m, T5FeedForward): + nn.init.normal_(m.gate[0].weight, std=m.dim**-0.5) + nn.init.normal_(m.fc1.weight, std=m.dim**-0.5) + nn.init.normal_(m.fc2.weight, std=m.dim_ffn**-0.5) + elif isinstance(m, T5Attention): + nn.init.normal_(m.q.weight, std=(m.dim * m.dim_attn) ** -0.5) + nn.init.normal_(m.k.weight, std=m.dim**-0.5) + nn.init.normal_(m.v.weight, std=m.dim**-0.5) + nn.init.normal_(m.o.weight, std=(m.num_heads * m.dim_attn) ** -0.5) + elif isinstance(m, T5RelativeEmbedding): + nn.init.normal_(m.embedding.weight, std=(2 * m.num_buckets * m.num_heads) ** -0.5) + + +class GELU(nn.Module): + def forward(self, x): + return 0.5 * x * (1.0 + torch.tanh(math.sqrt(2.0 / math.pi) * (x + 0.044715 * torch.pow(x, 3.0)))) + + +class T5LayerNorm(nn.Module): + def __init__(self, dim, eps=1e-6): + super(T5LayerNorm, self).__init__() + self.dim = dim + self.eps = eps + self.weight = nn.Parameter(torch.ones(dim)) + + def forward(self, x): + x = x * torch.rsqrt(x.float().pow(2).mean(dim=-1, keepdim=True) + self.eps) + if self.weight.dtype in [torch.float16, torch.bfloat16]: + x = x.type_as(self.weight) + return self.weight * x + + +class T5Attention(nn.Module): + def __init__(self, dim, dim_attn, num_heads, dropout=0.1): + assert dim_attn % num_heads == 0 + super(T5Attention, self).__init__() + self.dim = dim + self.dim_attn = dim_attn + self.num_heads = num_heads + self.head_dim = dim_attn // num_heads + + # layers + self.q = nn.Linear(dim, dim_attn, bias=False) + self.k = nn.Linear(dim, dim_attn, bias=False) + self.v = nn.Linear(dim, dim_attn, bias=False) + self.o = nn.Linear(dim_attn, dim, bias=False) + self.dropout = nn.Dropout(dropout) + + def forward(self, x, context=None, mask=None, pos_bias=None): + """ + x: [B, L1, C]. + context: [B, L2, C] or None. + mask: [B, L2] or [B, L1, L2] or None. + """ + # check inputs + context = x if context is None else context + b, n, c = x.size(0), self.num_heads, self.head_dim + + # compute query, key, value + q = self.q(x).view(b, -1, n, c) + k = self.k(context).view(b, -1, n, c) + v = self.v(context).view(b, -1, n, c) + + # attention bias + attn_bias = x.new_zeros(b, n, q.size(1), k.size(1)) + if pos_bias is not None: + attn_bias += pos_bias + if mask is not None: + assert mask.ndim in [2, 3] + mask = mask.view(b, 1, 1, -1) if mask.ndim == 2 else mask.unsqueeze(1) + attn_bias.masked_fill_(mask == 0, torch.finfo(x.dtype).min) + + # compute attention (T5 does not use scaling) + attn = torch.einsum("binc,bjnc->bnij", q, k) + attn_bias + attn = F.softmax(attn.float(), dim=-1).type_as(attn) + x = torch.einsum("bnij,bjnc->binc", attn, v) + + # output + x = x.reshape(b, -1, n * c) + x = self.o(x) + x = self.dropout(x) + return x + + +class T5FeedForward(nn.Module): + def __init__(self, dim, dim_ffn, dropout=0.1): + super(T5FeedForward, self).__init__() + self.dim = dim + self.dim_ffn = dim_ffn + + # layers + self.gate = nn.Sequential(nn.Linear(dim, dim_ffn, bias=False), GELU()) + self.fc1 = nn.Linear(dim, dim_ffn, bias=False) + self.fc2 = nn.Linear(dim_ffn, dim, bias=False) + self.dropout = nn.Dropout(dropout) + + def forward(self, x): + x = self.fc1(x) * self.gate(x) + x = self.dropout(x) + x = self.fc2(x) + x = self.dropout(x) + return x + + +class T5SelfAttention(nn.Module): + def __init__(self, dim, dim_attn, dim_ffn, num_heads, num_buckets, shared_pos=True, dropout=0.1): + super(T5SelfAttention, self).__init__() + self.dim = dim + self.dim_attn = dim_attn + self.dim_ffn = dim_ffn + self.num_heads = num_heads + self.num_buckets = num_buckets + self.shared_pos = shared_pos + + # layers + self.norm1 = T5LayerNorm(dim) + self.attn = T5Attention(dim, dim_attn, num_heads, dropout) + self.norm2 = T5LayerNorm(dim) + self.ffn = T5FeedForward(dim, dim_ffn, dropout) + self.pos_embedding = None if shared_pos else T5RelativeEmbedding(num_buckets, num_heads, bidirectional=True) + + def forward(self, x, mask=None, pos_bias=None): + e = pos_bias if self.shared_pos else self.pos_embedding(x.size(1), x.size(1)) + x = fp16_clamp(x + self.attn(self.norm1(x), mask=mask, pos_bias=e)) + x = fp16_clamp(x + self.ffn(self.norm2(x))) + return x + + +class T5CrossAttention(nn.Module): + def __init__(self, dim, dim_attn, dim_ffn, num_heads, num_buckets, shared_pos=True, dropout=0.1): + super(T5CrossAttention, self).__init__() + self.dim = dim + self.dim_attn = dim_attn + self.dim_ffn = dim_ffn + self.num_heads = num_heads + self.num_buckets = num_buckets + self.shared_pos = shared_pos + + # layers + self.norm1 = T5LayerNorm(dim) + self.self_attn = T5Attention(dim, dim_attn, num_heads, dropout) + self.norm2 = T5LayerNorm(dim) + self.cross_attn = T5Attention(dim, dim_attn, num_heads, dropout) + self.norm3 = T5LayerNorm(dim) + self.ffn = T5FeedForward(dim, dim_ffn, dropout) + self.pos_embedding = None if shared_pos else T5RelativeEmbedding(num_buckets, num_heads, bidirectional=False) + + def forward(self, x, mask=None, encoder_states=None, encoder_mask=None, pos_bias=None): + e = pos_bias if self.shared_pos else self.pos_embedding(x.size(1), x.size(1)) + x = fp16_clamp(x + self.self_attn(self.norm1(x), mask=mask, pos_bias=e)) + x = fp16_clamp(x + self.cross_attn(self.norm2(x), context=encoder_states, mask=encoder_mask)) + x = fp16_clamp(x + self.ffn(self.norm3(x))) + return x + + +class T5RelativeEmbedding(nn.Module): + def __init__(self, num_buckets, num_heads, bidirectional, max_dist=128): + super(T5RelativeEmbedding, self).__init__() + self.num_buckets = num_buckets + self.num_heads = num_heads + self.bidirectional = bidirectional + self.max_dist = max_dist + + # layers + self.embedding = nn.Embedding(num_buckets, num_heads) + + def forward(self, lq, lk): + device = self.embedding.weight.device + # rel_pos = torch.arange(lk).unsqueeze(0).to(device) - \ + # torch.arange(lq).unsqueeze(1).to(device) + rel_pos = torch.arange(lk, device=device).unsqueeze(0) - torch.arange(lq, device=device).unsqueeze(1) + rel_pos = self._relative_position_bucket(rel_pos) + rel_pos_embeds = self.embedding(rel_pos) + rel_pos_embeds = rel_pos_embeds.permute(2, 0, 1).unsqueeze(0) # [1, N, Lq, Lk] + return rel_pos_embeds.contiguous() + + def _relative_position_bucket(self, rel_pos): + # preprocess + if self.bidirectional: + num_buckets = self.num_buckets // 2 + rel_buckets = (rel_pos > 0).long() * num_buckets + rel_pos = torch.abs(rel_pos) + else: + num_buckets = self.num_buckets + rel_buckets = 0 + rel_pos = -torch.min(rel_pos, torch.zeros_like(rel_pos)) + + # embeddings for small and large positions + max_exact = num_buckets // 2 + rel_pos_large = ( + max_exact + + ( + torch.log(rel_pos.float() / max_exact) / math.log(self.max_dist / max_exact) * (num_buckets - max_exact) + ).long() + ) + rel_pos_large = torch.min(rel_pos_large, torch.full_like(rel_pos_large, num_buckets - 1)) + rel_buckets += torch.where(rel_pos < max_exact, rel_pos, rel_pos_large) + return rel_buckets + + +class T5Encoder(nn.Module): + def __init__(self, vocab, dim, dim_attn, dim_ffn, num_heads, num_layers, num_buckets, shared_pos=True, dropout=0.1): + super(T5Encoder, self).__init__() + self.dim = dim + self.dim_attn = dim_attn + self.dim_ffn = dim_ffn + self.num_heads = num_heads + self.num_layers = num_layers + self.num_buckets = num_buckets + self.shared_pos = shared_pos + + # layers + self.token_embedding = vocab if isinstance(vocab, nn.Embedding) else nn.Embedding(vocab, dim) + self.pos_embedding = T5RelativeEmbedding(num_buckets, num_heads, bidirectional=True) if shared_pos else None + self.dropout = nn.Dropout(dropout) + self.blocks = nn.ModuleList( + [ + T5SelfAttention(dim, dim_attn, dim_ffn, num_heads, num_buckets, shared_pos, dropout) + for _ in range(num_layers) + ] + ) + self.norm = T5LayerNorm(dim) + + # initialize weights + self.apply(init_weights) + + def forward(self, ids, mask=None): + x = self.token_embedding(ids) + x = self.dropout(x) + e = self.pos_embedding(x.size(1), x.size(1)) if self.shared_pos else None + for block in self.blocks: + x = block(x, mask, pos_bias=e) + x = self.norm(x) + x = self.dropout(x) + return x + + +class T5Decoder(nn.Module): + def __init__(self, vocab, dim, dim_attn, dim_ffn, num_heads, num_layers, num_buckets, shared_pos=True, dropout=0.1): + super(T5Decoder, self).__init__() + self.dim = dim + self.dim_attn = dim_attn + self.dim_ffn = dim_ffn + self.num_heads = num_heads + self.num_layers = num_layers + self.num_buckets = num_buckets + self.shared_pos = shared_pos + + # layers + self.token_embedding = vocab if isinstance(vocab, nn.Embedding) else nn.Embedding(vocab, dim) + self.pos_embedding = T5RelativeEmbedding(num_buckets, num_heads, bidirectional=False) if shared_pos else None + self.dropout = nn.Dropout(dropout) + self.blocks = nn.ModuleList( + [ + T5CrossAttention(dim, dim_attn, dim_ffn, num_heads, num_buckets, shared_pos, dropout) + for _ in range(num_layers) + ] + ) + self.norm = T5LayerNorm(dim) + + # initialize weights + self.apply(init_weights) + + def forward(self, ids, mask=None, encoder_states=None, encoder_mask=None): + b, s = ids.size() + + # causal mask + if mask is None: + mask = torch.tril(torch.ones(1, s, s).to(ids.device)) + elif mask.ndim == 2: + mask = torch.tril(mask.unsqueeze(1).expand(-1, s, -1)) + + # layers + x = self.token_embedding(ids) + x = self.dropout(x) + e = self.pos_embedding(x.size(1), x.size(1)) if self.shared_pos else None + for block in self.blocks: + x = block(x, mask, encoder_states, encoder_mask, pos_bias=e) + x = self.norm(x) + x = self.dropout(x) + return x + + +class T5Model(nn.Module): + def __init__( + self, + vocab_size, + dim, + dim_attn, + dim_ffn, + num_heads, + encoder_layers, + decoder_layers, + num_buckets, + shared_pos=True, + dropout=0.1, + ): + super(T5Model, self).__init__() + self.vocab_size = vocab_size + self.dim = dim + self.dim_attn = dim_attn + self.dim_ffn = dim_ffn + self.num_heads = num_heads + self.encoder_layers = encoder_layers + self.decoder_layers = decoder_layers + self.num_buckets = num_buckets + + # layers + self.token_embedding = nn.Embedding(vocab_size, dim) + self.encoder = T5Encoder( + self.token_embedding, dim, dim_attn, dim_ffn, num_heads, encoder_layers, num_buckets, shared_pos, dropout + ) + self.decoder = T5Decoder( + self.token_embedding, dim, dim_attn, dim_ffn, num_heads, decoder_layers, num_buckets, shared_pos, dropout + ) + self.head = nn.Linear(dim, vocab_size, bias=False) + + # initialize weights + self.apply(init_weights) + + def forward(self, encoder_ids, encoder_mask, decoder_ids, decoder_mask): + x = self.encoder(encoder_ids, encoder_mask) + x = self.decoder(decoder_ids, decoder_mask, x, encoder_mask) + x = self.head(x) + return x + + +def _t5(name, encoder_only=False, decoder_only=False, dtype=torch.float32, device="cpu", **kwargs): + # sanity check + assert not (encoder_only and decoder_only) + + # params + if encoder_only: + model_cls = T5Encoder + kwargs["vocab"] = kwargs.pop("vocab_size") + kwargs["num_layers"] = kwargs.pop("encoder_layers") + _ = kwargs.pop("decoder_layers") + elif decoder_only: + model_cls = T5Decoder + kwargs["vocab"] = kwargs.pop("vocab_size") + kwargs["num_layers"] = kwargs.pop("decoder_layers") + _ = kwargs.pop("encoder_layers") + else: + model_cls = T5Model + + # init model + with torch.device(device): + model = model_cls(**kwargs) + + # set device + model = model.to(dtype=dtype, device=device) + + return model + + +def umt5_xxl(**kwargs): + cfg = dict( + vocab_size=256384, + dim=4096, + dim_attn=4096, + dim_ffn=10240, + num_heads=64, + encoder_layers=24, + decoder_layers=24, + num_buckets=32, + shared_pos=False, + dropout=0.1, + ) + cfg.update(**kwargs) + return _t5("umt5-xxl", **cfg) + + +def get_storage_reader(checkpoint_path: str, credential_path: Optional[str] = None): + is_s3 = "s3://" in checkpoint_path + if is_s3: + return S3StorageReader( + credential_path=credential_path, + path=checkpoint_path, + ) + return FileSystemReader(checkpoint_path) + + +def load_model_dcp(model, ckpt_path, credential_path: Optional[str] = None): + storage_reader = get_storage_reader(ckpt_path, credential_path) + _state_dict = get_model_state_dict(model) + dcp.load(_state_dict, storage_reader=storage_reader, planner=DefaultLoadPlanner(allow_partial_load=True)) + log.info(set_model_state_dict(model, _state_dict, options=StateDictOptions(strict=False))) + return model + + +def load_model_torch(model, ckpt_path, credential_path: Optional[str] = None): + if distributed.is_rank0(): + if ckpt_path.startswith("s3://"): + backend_key = "_umt5_encoder" + easy_io.set_s3_backend( + key=backend_key, + backend_args={ + "backend": "s3", + "s3_credential_path": credential_path, + }, + ) + else: + backend_key = None + + ckpt = easy_io.load( + ckpt_path, + backend_key=backend_key, + map_location="cuda", + ) + model.load_state_dict(ckpt) + + distributed.sync_model_states(model, src=0) + return model + + +class UMT5EncoderModel: + def __init__( + self, + text_len=512, + dtype=torch.bfloat16, + device=torch.cuda.current_device(), + checkpoint_path="s3://bucket/cosmos_diffusion_v2/pretrain_weights/models_t5_umt5-xxl-enc-bf16.pth", + tokenizer_path="google/umt5-xxl", + credential_path: Optional[str] = "credentials/s3_training.secret", + enable_fsdp_shard: bool = False, + ): + assert not enable_fsdp_shard, "FSDP is not supported for UMT5" + self.text_len = text_len + self.dtype = dtype + self.device = device + + # init model + model = umt5_xxl(encoder_only=True, dtype=dtype, device=device).eval().requires_grad_(False) + log.info(f"loading {checkpoint_path}") + if checkpoint_path.endswith(".dcp"): + model = load_model_dcp(model, checkpoint_path, credential_path=credential_path) + else: + assert checkpoint_path.endswith(".pth"), "only .pth or .dcp are supported" + model = load_model_torch(model, checkpoint_path, credential_path=credential_path) + self.model = model + self.model.to(self.device) + # init tokenizer + self.tokenizer = HuggingfaceTokenizer(name=tokenizer_path, seq_len=text_len, clean="whitespace") + + def __call__(self, texts, device: Optional[torch.device] = None): + if device is None: + device = self.device + ids, mask = self.tokenizer(texts, return_mask=True, add_special_tokens=True) + ids = ids.to(device) + mask = mask.to(device) + seq_lens = mask.gt(0).sum(dim=1).long() + context = self.model(ids, mask) + # return [u[:v] for u, v in zip(context, seq_lens)] + stack_emb = [] + for u, length in zip(context, seq_lens): + if length > self.text_len: + stack_emb.append(u[: self.text_len]) + else: + # pad with zeros to max_length + zeros = torch.zeros(self.text_len - length, u.shape[1]).to(u) + stack_emb.append(torch.cat([u[:length], zeros], dim=0)) + return torch.stack(stack_emb) + + +t5_encoder: Optional[UMT5EncoderModel] = None + + +def get_umt5_embedding( + prompts: Union[str, List[str]], + device: str = "cuda", + max_length: int = 512, +) -> torch.Tensor: + global t5_encoder + if t5_encoder is None: + t5_encoder = UMT5EncoderModel(device=device) + return t5_encoder(prompts, device=device) + + +def get_negative_emb(): + easy_io.set_s3_backend( + backend_args={ + "backend": "s3", + "s3_credential_path": "credentials/s3_training.secret", + }, + ) + neg_prompt = "色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走" + emb = get_umt5_embedding(neg_prompt).to(dtype=torch.bfloat16).cpu() + print(emb.shape) + easy_io.dump(emb[0], "s3://bucket/cosmos_diffusion_v2/pretrain_weights/umT5_wan_negative_emb.pt") diff --git a/REGEN-main/cosmos_policy/_src/predict2/inference/get_umt5_emb_test.py b/REGEN-main/cosmos_policy/_src/predict2/inference/get_umt5_emb_test.py new file mode 100644 index 0000000000000000000000000000000000000000..1230318e3016911ef23882d31dcd06799f7da3ab --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/inference/get_umt5_emb_test.py @@ -0,0 +1,32 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import pytest + +from cosmos_policy._src.imaginaire.utils import misc +from cosmos_policy._src.predict2.inference.get_umt5_emb import UMT5EncoderModel + + +@pytest.mark.L2 +def test_encoder(): + with misc.timer("load model"): + model = UMT5EncoderModel( + checkpoint_path="s3://bucket/cosmos_diffusion_v2/pretrain_weights/models_t5_umt5-xxl-enc-bf16.pth" + ) + emb = model(texts=["hello world", "hello", "world"]) + assert len(emb) == 3 + assert emb[0].shape == (512, 4096) + assert emb[1].shape == (512, 4096) + assert emb[2].shape == (512, 4096) diff --git a/REGEN-main/cosmos_policy/_src/predict2/inference/interpolator_cli.py b/REGEN-main/cosmos_policy/_src/predict2/inference/interpolator_cli.py new file mode 100644 index 0000000000000000000000000000000000000000..9a2d64cdf8cbc99b932381ff8084e0630ceae4db --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/inference/interpolator_cli.py @@ -0,0 +1,458 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""Command-line interface for video frame interpolation using diffusion models. + +This script processes video files and generates interpolated frames between existing frames, +effectively increasing the frame rate of videos using trained diffusion models. + +Example usage: +# 720p 2X FRUC +run_docker -g 3 -i nvcr.io/nvidian/imaginaire4:v10.1.0 \ + "python3 -m cosmos_policy._src.predict2.inference.interpolator_cli \ + --experiment=Interpolation-2B-720p-16fps-to-32fps-HQ_V6_from_22 \ + --ckpt_path s3://bucket/predict2/frame_interpolation/Interpolation-2B-720p-16fps-to-32fps-HQ_V6_from_22/checkpoints/iter_000142000 \ + --ckpt_cred credentials/pbss_dir_share.secret \ + --video_pattern 'tmp/panda70m_test_0000071_00000.mp4' \ + --output_dir tmp/panda70m_test_0000071_00000 \ + --upsample_factor 2 \ + --num_frame_pairs 2 \ + --output_frames" + + # For Multi-GPU with context parallelism + append `--context_parallel_size ` to the command above. + +# 1080p 4X FRUC +run_docker -g 3 -i nvcr.io/nvidian/imaginaire4:v10.1.0 \ + "python3 -m cosmos_policy._src.predict2.inference.interpolator_cli \ + --experiment=Interpolation-2B-1080p-8fps-to-32fps-HQ_V6_from_22 \ + --ckpt_path s3://bucket/predict2/frame_interpolation/Interpolation-2B-1080p-16fps-to-48fps-HQ_V6_from_22/checkpoints/iter_000116000 \ + --ckpt_cred credentials/pbss_dir_share.secret \ + --video_pattern 'tmp/panda70m_test_0000071_00000.mp4' \ + --output_dir tmp/panda70m_test_0000071_00000 \ + --upsample_factor 4 \ + --num_frame_pairs 2 \ + --output_frames" + +# 1080p 24-to-30fps +CUDA_VISIBLE_DEVICES=0 torchrun --nproc_per_node=1 -m cosmos_policy._src.predict2.inference.interpolator_cli \ + --experiment=Interpolation-2B-1080p-24fps-to-30fps-HQ_V6_from_22 \ + --ckpt_path s3://bucket/cosmos_diffusion_v2/frame_interpolation/Interpolation-2B-1080p-24fps-to-30fps-HQ_V6_from_22/checkpoints/iter_000010000 \ + --ckpt_cred credentials/s3_training.secret \ + --video_pattern 'assets/upscaler/000005.mp4' \ + --output_dir results/interpolator/interleave/iter10k \ + --num_interleaved_frames 4 \ + --num_frame_pairs 2 \ + --output_frames + +CUDA_VISIBLE_DEVICES=0 torchrun --nproc_per_node=1 -m cosmos_policy._src.predict2.inference.interpolator_cli \ + --experiment=Interpolation-2B-1080p-24fps-to-30fps-HQ_V6_from_22_rectified_flow \ + --ckpt_path s3://bucket/cosmos_diffusion_v2/frame_interpolation/Interpolation-2B-1080p-24fps-to-30fps-HQ_V6_from_22_rectified_flow/checkpoints/iter_000010000 \ + --ckpt_cred credentials/s3_training.secret \ + --video_pattern 'assets/upscaler/000005.mp4' \ + --output_dir results/interpolator/interleave_rectified_flow/iter10k \ + --num_interleaved_frames 4 \ + --num_frame_pairs 2 \ + --output_frames + + +Expected input structure: + input_root/ + ├── video1.mp4 + ├── video1.txt (optional text prompt) + ├── video2.mp4 + ├── video2.txt (optional text prompt) + └── ... + +Generated output structure +(if `--output_dir` is provided, the filename subdirectory is omitted): + output_dir/ + ├── video1/ + ├── interpolated.mp4 + ├── interpolated_frames/ + ├── frame_000000.jpg + ├── frame_000001.jpg + └── ... + ├── video2/ + ├── interpolated.mp4 + ├── interpolated_frames/ + └── ... + └── ... + +# Method 1: Direct python for 1 GPU +run_docker -g 1 -i nvcr.io/nvidian/imaginaire4:v10.1.0 \ + "python3 -m cosmos_policy._src.predict2.inference.interpolator_cli \ + --experiment=Interpolation-2B-720p-16fps-to-32fps-HQ_V6_from_22 \ + --ckpt_path s3://bucket/cosmos_diffusion_v2/frame_interpolation/Interpolation-2B-720p-16fps-to-32fps-HQ_V6_from_22/checkpoints/iter_000370000 \ + --ckpt_cred credentials/s3_checkpoint.secret \ + --video_pattern 's3://cosmos2_results/qinshengz_Stage-c_pt_4-Index-22-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_20_iter-26000_task1_dataset-transition_change_issue_upsampled_prompts_v1/*/0.mp4' \ + --input_cred credentials/pdx_cosmos_benchmark.secret \ + --upsample_factor 2 \ + --num_frame_pairs -1 \ + --output_frames" + +# Method 2: torchrun with 4 GPUs (should work if mean_std_cli works) +run_docker -g 0,1,2,3 -i nvcr.io/nvidian/imaginaire4:v10.1.0 \ + "torchrun --nproc_per_node=4 -m cosmos_policy._src.predict2.inference.interpolator_cli \ + --experiment=Interpolation-2B-720p-16fps-to-32fps-HQ_V6_from_22 \ + --ckpt_path s3://bucket/cosmos_diffusion_v2/frame_interpolation/Interpolation-2B-720p-16fps-to-32fps-HQ_V6_from_22/checkpoints/iter_000370000 \ + --ckpt_cred credentials/s3_checkpoint.secret \ + --video_pattern 's3://cosmos2_results/qinshengz_Stage-c_pt_4-Index-22-Size-2B-Res-720-Fps-16-Note-HQ_V3_from_20_iter-26000_task1_dataset-transition_change_issue_upsampled_prompts_v1/*/0.mp4' \ + --input_cred credentials/pdx_cosmos_benchmark.secret \ + --upsample_factor 2 \ + --num_frame_pairs -1 \ + --output_frames" +""" + +import argparse +import os + +import numpy as np +import torch +import torch.distributed as dist +from loguru import logger + +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.context_managers import distributed_init +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.predict2.inference.interpolator_lib import Interpolator +from cosmos_policy._src.predict2.inference.utils import ( + get_filepaths, + numpy2tensor, + read_video, + set_s3_backend, + tensor2numpy, + write_image, + write_video, +) + +_DEFAULT_FPS = 24.0 + + +def parse_arguments() -> argparse.Namespace: + """Parse command-line arguments for the interpolator inference script.""" + parser = argparse.ArgumentParser(description="Video frame interpolation inference script") + + # Model and experiment configuration + parser.add_argument("--experiment", type=str, required=True, help="Experiment configuration name") + parser.add_argument( + "--ckpt_path", + type=str, + default=None, + help="Path to the model checkpoint (local or S3). If not provided, uses config default", + ) + parser.add_argument( + "--ckpt_cred", + type=str, + default="credentials/s3_checkpoint.secret", + help="Path to S3 credentials for checkpoint access", + ) + + # Input/output configuration + parser.add_argument( + "--video_pattern", type=str, default="path/to/videos/*.mp4", help="Glob pattern for input videos (local or S3)" + ) + parser.add_argument( + "--input_cred", + type=str, + default="credentials/pbss_dir_share.secret", + help="Path to S3 credentials for input access", + ) + parser.add_argument( + "--output_dir", + type=str, + default=None, + help="Output directory (local or S3). Defaults to same directory as input", + ) + parser.add_argument( + "--output_frames", + action="store_true", + help="Save individual interpolated frames as JPEG files", + ) + + # Interpolation parameters + parser.add_argument( + "--upsample_factor", + type=int, + default=2, + help="Temporal framerate upsampling factor (e.g., 2 for 2X FRUC, 4 for 4X FRUC)", + ) + parser.add_argument( + "--num_frame_pairs", + type=int, + default=-1, + help="Number of consecutive frame pairs to process from each input video. If -1, process all frame pairs", + ) + parser.add_argument( + "--resolution", + type=str, + default=None, + help="Target resolution as 'H,W'. Uses model's default resolution if not specified", + ) + parser.add_argument( + "--num_interleaved_frames", + type=int, + default=0, + choices=[0, 4], + help="Number of interleaved frames for interpolation. 0 means no interleaved frames.", + ) + + # Model inference parameters + parser.add_argument("--guidance", type=int, default=-1, help="Classifier-free guidance scale") + parser.add_argument("--seed", type=int, default=1, help="Random seed for reproducibility") + parser.add_argument( + "--negative_prompt", + type=str, + default=None, + help="Custom negative prompt for classifier-free guidance. Uses default S3 embeddings if not specified", + ) + + # Distributed processing + parser.add_argument( + "--context_parallel_size", + type=int, + default=1, + help="Number of GPUs for context parallelism. Use 2+ if encountering OOM errors", + ) + + return parser.parse_args() + + +def _read_prompt(prompt_path: str) -> str | None: + """Read text prompt from file if it exists. + + Args: + prompt_path: Path to the prompt text file. + + Returns: + Text prompt content if file exists, None otherwise. + """ + if easy_io.exists(prompt_path): + logger.info(f"Loading prompt from {prompt_path}") + prompt = easy_io.load(prompt_path, file_format="txt") + return prompt.strip() + return None + + +def _get_output_video_dir(input_video_filepath: str, output_dir: str = None, output_frames: bool = False) -> str: + """Generate output directory path for processed video. + + Args: + input_video_filepath: Path to input video file. + output_dir: Base output directory (optional). + output_frames: Whether frame output directory should be created. + + Returns: + Path to the output directory for this video. + """ + video_filename = os.path.basename(input_video_filepath).split(".")[0] + video_dirname = os.path.dirname(input_video_filepath) + output_video_dir = output_dir or os.path.join(video_dirname, video_filename) + + # Create directories for local output + if not output_video_dir.startswith("s3://"): + os.makedirs(output_video_dir, exist_ok=True) + if output_frames: + output_frames_dir = os.path.join(output_video_dir, "interpolated_frames") + os.makedirs(output_frames_dir, exist_ok=True) + + return output_video_dir + + +def _generate_interpolated_frames( + input_video, + interpolator, + upsample_factor: int, + num_frame_pairs: int, + num_interleaved_frames: int = 0, + prompt: str = None, + guidance: int = -1, + resolution: str = None, + seed: int = 1, + negative_prompt: str = None, +) -> list: + """Generate interpolated frames for consecutive frame pairs. + + Args: + input_video: Input video frames array. + interpolator: Interpolator instance for frame generation. + upsample_factor: Temporal framerate upsampling factor. + num_frame_pairs: Number of consecutive frame pairs to process. + prompt: Optional text prompt for interpolation. + guidance: Classifier-free guidance scale. + resolution: Target resolution as 'H,W'. + seed: Random seed for reproducibility. + negative_prompt: Custom negative prompt for classifier-free guidance. + + Returns: + List of interpolated frames as numpy arrays. + """ + interpolated_frames = [] + + if num_interleaved_frames > 0: + actual_num_pairs = (len(input_video) - 1) // num_interleaved_frames + else: + actual_num_pairs = len(input_video) - 1 + if num_frame_pairs > 0: + actual_num_pairs = min(num_frame_pairs, actual_num_pairs) + + for frame_idx in range(1, actual_num_pairs + 1): + if num_interleaved_frames > 0: + start_idx = (frame_idx - 1) * num_interleaved_frames + end_idx = start_idx + num_interleaved_frames + 1 + input_frames = input_video[start_idx:end_idx] + zeros = np.zeros_like(input_frames[0]) + concat_frames = [input_frames[0]] + for i in range(1, num_interleaved_frames + 1): + concat_frames.append(zeros) + concat_frames.append(input_frames[i]) + assert len(concat_frames) == 9, f"Only support 9 frames for now, got {len(concat_frames)}" + video_batch = np.stack(concat_frames) + else: + # Get consecutive frame pair + first_frame, last_frame = input_video[frame_idx - 1 : frame_idx + 1] + + # Create interpolation sequence: first frame, zeros, last frame + zeros = np.zeros_like(first_frame) + middle_frames = [zeros] * (upsample_factor - 1) # List of zero frames + video_batch = np.stack([first_frame] + middle_frames + [last_frame]) + + # Convert to tensor and resize + video_batch = numpy2tensor(video_batch[np.newaxis, ...]) + + # Generate interpolated frames + curr_frames = interpolator( + prompt=prompt, + input_video=video_batch, + guidance=guidance, + resolution=resolution, + seed=seed, + negative_prompt=negative_prompt, + ) + + # Convert to numpy and accumulate frames + curr_frames = tensor2numpy(curr_frames)[0] + # Skip first frame for subsequent pairs to avoid duplication + if num_interleaved_frames > 0: # remove input frames + indices = [0] + list(range(1, curr_frames.shape[0], 2)) + [curr_frames.shape[0] - 1] + curr_frames = curr_frames[indices] + curr_frames_ = curr_frames if frame_idx == 1 else curr_frames[1:] + interpolated_frames.extend(curr_frames_) + + return np.stack(interpolated_frames) + + +def main(): + """Main entry point for the interpolator CLI.""" + torch.enable_grad(False) # Disable gradients for inference + args = parse_arguments() + + # Initialize distributed processing if environment is set up for it + if "RANK" in os.environ and "WORLD_SIZE" in os.environ: + with distributed_init(): + distributed.init() + + world_size = distributed.get_world_size() + rank = distributed.get_rank() + + # Initialize the interpolator + interpolator = Interpolator( + args.experiment, args.ckpt_path, args.ckpt_cred, context_parallel_size=args.context_parallel_size + ) + + # Set S3 backend for all ranks (needed for video file access) + set_s3_backend(credentials=args.input_cred) + + # Discover input video files (only rank 0 does the search) + if rank == 0: + filepaths = get_filepaths(args.video_pattern) + + # Ensure we have enough files for all ranks + if len(filepaths) < world_size: + log.error(f"Found {len(filepaths)} files but need at least {world_size} for {world_size} GPUs") + exit(1) + + # Trim to be evenly divisible by world_size + num_files_per_rank = len(filepaths) // world_size + filepaths = filepaths[: num_files_per_rank * world_size] + log.info(f"Processing {len(filepaths)} files ({num_files_per_rank} per rank)") + else: + filepaths = [] + + # Broadcast the file list to all ranks + if world_size > 1: + filepaths_list = [filepaths] + dist.broadcast_object_list(filepaths_list, src=0) + filepaths = filepaths_list[0] + + # Distribute videos across ranks using round-robin + rank_filepaths = filepaths[rank::world_size] + log.info(f"Rank {rank}: Processing {len(rank_filepaths)} videos") + + # Process each video file assigned to this rank + for idx, input_video_filepath in enumerate(rank_filepaths): + log.info(f"Rank {rank}: Processing input video {idx + 1}/{len(rank_filepaths)}: {input_video_filepath}") + + # Load input video and metadata + input_video = read_video(input_video_filepath) + input_fps = getattr(input_video.metadata, "fps", _DEFAULT_FPS) + if args.num_interleaved_frames > 0: + output_fps = input_fps * (args.num_interleaved_frames + 1) / args.num_interleaved_frames + else: + output_fps = input_fps * args.upsample_factor + + # Load optional text prompt + prompt = _read_prompt(input_video_filepath.replace(".mp4", ".txt")) + + # Generate interpolated frames for consecutive frame pairs + interpolated_frames = _generate_interpolated_frames( + input_video=input_video, + interpolator=interpolator, + upsample_factor=args.upsample_factor, + num_frame_pairs=args.num_frame_pairs, + num_interleaved_frames=args.num_interleaved_frames, + prompt=prompt, + guidance=args.guidance, + resolution=args.resolution, + seed=args.seed, + negative_prompt=args.negative_prompt, + ) + + # Save interpolated video + output_video_dir = _get_output_video_dir(input_video_filepath, args.output_dir, args.output_frames) + output_video_path = f"{output_video_dir}/interpolated.mp4" + write_video(output_video_path, interpolated_frames, fps=output_fps) + + # Optionally save individual frames + if args.output_frames: + frames_dir = f"{output_video_dir}/interpolated_frames" + for frame_idx, frame in enumerate(interpolated_frames): + frame_path = f"{frames_dir}/frame_{frame_idx:06d}.jpg" + write_image(frame_path, frame) + + log.info(f"Rank {rank}: Completed output video {idx + 1}/{len(rank_filepaths)}: {output_video_path}") + + log.info(f"Rank {rank}: Finished processing all {len(rank_filepaths)} videos") + + # Synchronize before cleanup + if world_size > 1: + dist.barrier() + + # Clean up distributed resources + interpolator.cleanup() + + +if __name__ == "__main__": + main() diff --git a/REGEN-main/cosmos_policy/_src/predict2/inference/interpolator_lib.py b/REGEN-main/cosmos_policy/_src/predict2/inference/interpolator_lib.py new file mode 100644 index 0000000000000000000000000000000000000000..c1c63af57a4669dc041102a4f0a941ffd56ceb09 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/inference/interpolator_lib.py @@ -0,0 +1,343 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""Library for video frame interpolation using diffusion models. + +This module provides functionality to interpolate frames between two input frames, +effectively increasing the frame rate of videos using trained diffusion models. + +Usage: + interpolator = Interpolator(experiment_name, ckpt_path) + output = interpolator(**input_args) +""" + +import math +from urllib.parse import urlparse + +import torch +import torch.nn as nn +import torchvision +from einops import rearrange +from loguru import logger +from megatron.core import parallel_state + +from cosmos_policy._src.imaginaire.utils import distributed +from cosmos_policy._src.imaginaire.utils.s3_utils import load_from_s3_with_cache +from cosmos_policy._src.predict2.inference.get_t5_emb import get_text_embedding +from cosmos_policy._src.predict2.utils.model_loader import load_model_from_checkpoint + +_CONFIG_FILE = "cosmos_policy/_src/predict2/configs/frame_interpolation/config.py" +_NEG_PROMPT_EMBEDDINGS_S3_PATH = "s3://bucket/projects/edify_video/v4/video_neg_prompt_embeddings_v0.pt" +_UINT8_MAX_F = float(torch.iinfo(torch.uint8).max) +_T5_MAX_LENGTH = 512 +_T5_HIDDEN_DIM = 1024 +_DEFAULT_NEGATIVE_PROMPT = "The video captures a series of frames showing ugly scenes, static with no motion, motion blur, over-saturation, shaky footage, low resolution, grainy texture, pixelated images, poorly lit areas, underexposed and overexposed scenes, poor color balance, washed out colors, choppy sequences, jerky movements, low frame rate, artifacting, color banding, unnatural transitions, outdated special effects, fake elements, unconvincing visuals, poorly edited content, jump cuts, visual noise, and flickering. Overall, the video is of poor quality." + + +def resize_video_spatially(video: torch.Tensor, resolution: list[int]) -> torch.Tensor: + """Resize and center-crop video to target resolution while preserving aspect ratio. + + Args: + video: Input video tensor of shape (B, C, T, H, W). + resolution: Target resolution [H, W]. + + Returns: + Resized and cropped video tensor of shape (B, C, T, target_H, target_W). + """ + b, _, _, h, w = video.shape + # Reshape all frames into a batch of images for efficient processing + image_batch = rearrange(video, "b c t h w -> (b t) c h w") + + target_h, target_w = resolution + # Scale to ensure the smaller dimension matches target while preserving aspect ratio + scaling_ratio = max((target_w / w), (target_h / h)) + resizing_shape = (int(math.ceil(scaling_ratio * h)), int(math.ceil(scaling_ratio * w))) + + # Apply resize and center crop operations + image_resized = torchvision.transforms.functional.resize(image_batch, resizing_shape) + image_resized = torchvision.transforms.functional.center_crop(image_resized, resolution) + + # Reshape back to video tensor format + return rearrange(image_resized, "(b t) c h w -> b c t h w", b=b) + + +class Interpolator(nn.Module): + """Video frame interpolation inference handler using diffusion models. + + Supports both single-GPU and multi-GPU inference with context parallelism. + Loads trained diffusion models and generates interpolated frames between + input frame pairs using optional text conditioning. + """ + + def __init__(self, experiment_name: str, ckpt_path: str, s3_credential_path: str, context_parallel_size: int = 1): + """Initialize the interpolator inference handler. + + Args: + experiment_name: Name of the experiment configuration. + ckpt_path: Path to the model checkpoint (local or S3). + s3_credential_path: Path to S3 credentials file. + context_parallel_size: Number of GPUs for context parallelism. + """ + super().__init__() + self.experiment_name = experiment_name + self.ckpt_path = ckpt_path + self.s3_credential_path = s3_credential_path + self.context_parallel_size = context_parallel_size + self.process_group = None + + # Initialize distributed processing for multi-GPU setups + if self.context_parallel_size > 1: + self._init_distributed() + + # Load diffusion model and configuration + model, config = load_model_from_checkpoint( + experiment_name=self.experiment_name, + s3_checkpoint_dir=self.ckpt_path, + config_file=_CONFIG_FILE, + load_ema_to_reg=True, + experiment_opts=[ + f"checkpoint.load_from_object_store.credentials={self.s3_credential_path}", + f"checkpoint.save_to_object_store.credentials={self.s3_credential_path}", + f"checkpoint.load_from_object_store.bucket={urlparse(self.ckpt_path).netloc or 'dummy_bucket'}", + f"checkpoint.save_to_object_store.bucket={urlparse(self.ckpt_path).netloc or 'dummy_bucket'}", + f"model.config.tokenizer.s3_credential_path={self.s3_credential_path}", + ], + ) + + # Enable context parallelism for multi-GPU inference + if self.context_parallel_size > 1: + model.net.enable_context_parallel(self.process_group) + + self.model = model + self.model_config = config.model.config + self.precision = getattr(config.model, "precision", torch.bfloat16) + self.neg_t5_embeddings = None + + def _init_distributed(self): + """Initialize distributed processing for context parallelism.""" + # Setup distributed environment + distributed.init() + + # Configure model parallel states for context parallelism + parallel_state.initialize_model_parallel( + context_parallel_size=self.context_parallel_size, + ) + + # Obtain process group for context parallel communication + self.process_group = parallel_state.get_context_parallel_group() + + logger.info(f"Initialized context parallel with size {self.context_parallel_size}") + logger.info(f"Current rank: {distributed.get_rank()}, World size: {distributed.get_world_size()}") + + def _get_data_batch_input( + self, + video: torch.Tensor, + prompt: str | None, + negative_prompt: str | None = None, + use_neg_prompt: bool = True, + ) -> dict: + """Prepare input data batch for the diffusion model. + + Args: + video: Input video tensor (B, C, T, H, W). + prompt: Text prompt for conditioning (optional). + negative_prompt: Custom negative prompt (optional). + use_neg_prompt: Whether to include negative prompt embeddings. + + Returns: + Dictionary containing the prepared data batch with proper device and dtype. + """ + B, _, _, H, W = video.shape + + # Construct base data batch with required model inputs + data_batch = { + "dataset_name": "video_data", + "video": video, + "fps": torch.randint(16, 32, (B,)).float(), # Random FPS for model conditioning + "padding_mask": torch.zeros(B, 1, H, W), # No padding mask needed + "num_conditional_frames": 1, # Number of conditioning frames + } + + # # Add positive prompt embeddings if provided + # if prompt is not None: + # data_batch["t5_text_embeddings"] = get_text_embedding(prompt) + + # Compute text embeddings + if self.model.text_encoder is not None: + data_batch["ai_caption"] = [prompt] + data_batch["t5_text_embeddings"] = self.model.text_encoder.compute_text_embeddings_online( + data_batch={"ai_caption": [prompt], "images": None}, + input_caption_key="ai_caption", + ) + if use_neg_prompt: + if negative_prompt is None: + negative_prompt = _DEFAULT_NEGATIVE_PROMPT + data_batch["neg_t5_text_embeddings"] = self.model.text_encoder.compute_text_embeddings_online( + data_batch={"ai_caption": [negative_prompt], "images": None}, + input_caption_key="ai_caption", + ) + else: + data_batch["t5_text_embeddings"] = get_text_embedding(prompt) + if use_neg_prompt: + if negative_prompt is not None: + # Use custom negative prompt embeddings + logger.info(f"Using custom negative prompt: {negative_prompt}") + data_batch["neg_t5_text_embeddings"] = ( + get_text_embedding(negative_prompt).cuda().to(dtype=self.precision) + ) + else: + # Load default negative embeddings from S3 + if self.neg_t5_embeddings is None: + self._load_default_negative_embeddings() + + # Create zero-padded tensor for negative embeddings + zeros_t5 = torch.zeros([1, _T5_MAX_LENGTH, _T5_HIDDEN_DIM], dtype=self.precision).cuda() + length = min(_T5_MAX_LENGTH, self.neg_t5_embeddings.shape[0]) + zeros_t5[0, :length] = self.neg_t5_embeddings.to(dtype=self.precision).cuda()[:length] + data_batch["neg_t5_text_embeddings"] = zeros_t5 + + # Move floating-point tensors to GPU with model precision + for k, v in data_batch.items(): + if isinstance(v, torch.Tensor) and torch.is_floating_point(v): + data_batch[k] = v.cuda().to(dtype=self.precision) + + # # Configure negative prompts for classifier-free guidance + # if use_neg_prompt: + # if negative_prompt is not None: + # # Use custom negative prompt embeddings + # logger.info(f"Using custom negative prompt: {negative_prompt}") + # data_batch["neg_t5_text_embeddings"] = ( + # get_text_embedding(negative_prompt).cuda().to(dtype=self.precision) + # ) + # else: + # # Load default negative embeddings from S3 + # if self.neg_t5_embeddings is None: + # self._load_default_negative_embeddings() + + # # Create zero-padded tensor for negative embeddings + # zeros_t5 = torch.zeros([1, _T5_MAX_LENGTH, _T5_HIDDEN_DIM], dtype=self.precision).cuda() + # length = min(_T5_MAX_LENGTH, self.neg_t5_embeddings.shape[0]) + # zeros_t5[0, :length] = self.neg_t5_embeddings.to(dtype=self.precision).cuda()[:length] + # data_batch["neg_t5_text_embeddings"] = zeros_t5 + + # # Use zero embeddings for positive prompt if none provided + # if prompt is None: + # data_batch["t5_text_embeddings"] = zeros_t5 + # else: + # zeros_t5 = torch.zeros([B, _T5_MAX_LENGTH, _T5_HIDDEN_DIM], dtype=self.precision).cuda() + # data_batch["neg_t5_text_embeddings"] = zeros_t5 + # data_batch["t5_text_embeddings"] = zeros_t5 + + return data_batch + + def synchronize(self): + """Synchronize all processes in distributed mode.""" + if self.context_parallel_size > 1: + import torch.distributed as dist + + dist.barrier() + + def cleanup(self): + """Clean up distributed resources.""" + # Add synchronization before cleanup + self.synchronize() + + if self.context_parallel_size > 1: + import torch.distributed as dist + from megatron.core import parallel_state + + if parallel_state.is_initialized(): + parallel_state.destroy_model_parallel() + dist.destroy_process_group() + + def _load_default_negative_embeddings(self): + """Load default negative text embeddings from S3.""" + backend_args = { + "backend": "s3", + "path_mapping": None, + "s3_credential_path": self.s3_credential_path, + } + self.neg_t5_embeddings = load_from_s3_with_cache( + _NEG_PROMPT_EMBEDDINGS_S3_PATH, + easy_io_kwargs={"map_location": torch.device(torch.cuda.current_device())}, + backend_args=backend_args, + ) + + def forward( + self, + prompt: str, + input_video: torch.Tensor, + guidance: int = -1, + resolution: str = "1072,1920", + seed: int = 1, + negative_prompt: str = None, + ) -> torch.Tensor: + """Generate interpolated frames between input frame pair. + + Args: + prompt: Text prompt for conditioning the interpolation. + input_video: Input video batch of layout (B, C, T, H, W), range [-1..1], + with first and last temporal frames being the conditioning frames, + while the in-betweens to be interpolated are initialized as zeros. + guidance: Classifier-free guidance scale. + resolution: Target resolution as "H,W" string. + seed: Random seed for reproducibility. + negative_prompt: Custom negative prompt (optional). + + Returns: + Generated video tensor (B, C, T, H, W) in range [-1, 1]. + """ + # Validate input tensor dimensions and temporal frames + assert input_video.ndim == 5, "Input video must be a 5D tensor of layout (B, C, T, H, W)" + assert input_video.shape[-3] == self.model_config.state_t, ( + "The number of temporal frames in the input video must match the state_t in the model config" + ) + + # Determine target resolution for processing + if resolution is not None: + video_resolution = tuple(int(x) for x in resolution.split(",")) + else: + video_resolution = self.model.get_video_height_width() + input_video = resize_video_spatially(input_video, video_resolution) + + # Determine if we should use CFG + use_cfg = prompt is not None and guidance != -1 + + # Convert from [-1,1] float range to [0,255] uint8 range expected by model + input_video_uint8 = (_UINT8_MAX_F * (input_video + 1.0) / 2.0).to(dtype=torch.uint8) + + # Prepare model input data batch + data_batch = self._get_data_batch_input( + input_video_uint8, + prompt, + negative_prompt=negative_prompt, + use_neg_prompt=use_cfg, # Only use negative prompts when doing CFG + ) + + # Log current GPU memory usage + mem_bytes = torch.cuda.memory_allocated(device=torch.device("cuda" if torch.cuda.is_available() else "cpu")) + logger.info(f"GPU memory usage after preparing data batch: {mem_bytes / (1024**3):.2f} GB") + + # Generate latent samples using diffusion model + sample = self.model.generate_samples_from_batch( + data_batch, + n_sample=1, + guidance=guidance, + seed=seed, + is_negative_prompt=use_cfg, # Consistent with use_neg_prompt + ) + + # Decode latent samples back to video tensor + return self.model.decode(sample) diff --git a/REGEN-main/cosmos_policy/_src/predict2/inference/text2image.py b/REGEN-main/cosmos_policy/_src/predict2/inference/text2image.py new file mode 100644 index 0000000000000000000000000000000000000000..c3f34ebc897a21aa2829ce252d910a4d99dd74f0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/inference/text2image.py @@ -0,0 +1,111 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +PYTHONPATH=. streamlit run cosmos_policy/_src/predict2/inference/text2image.py --server.port 2222 +""" + +import torch + +from cosmos_policy._src.predict2.datasets.utils import IMAGE_RES_SIZE_INFO +from cosmos_policy._src.predict2.inference.get_t5_emb import get_text_embedding +from cosmos_policy._src.predict2.utils.model_loader import load_model_from_checkpoint + +torch.enable_grad(False) + + +def get_sample_batch( + resolution: str = "1024", + aspect_ratio: str = "16,9", + batch_size: int = 1, +) -> torch.Tensor: + w, h = IMAGE_RES_SIZE_INFO[resolution][aspect_ratio] + data_batch = { + "dataset_name": "image_data", + "images": torch.randn(batch_size, 3, h, w).cuda(), + "t5_text_embeddings": torch.randn(batch_size, 512, 1024).cuda(), + "fps": torch.randint(16, 32, (batch_size,)).cuda(), + "padding_mask": torch.zeros(batch_size, 1, h, w).cuda(), + } + + for k, v in data_batch.items(): + if isinstance(v, torch.Tensor) and torch.is_floating_point(data_batch[k]): + data_batch[k] = v.cuda().to(dtype=torch.bfloat16) + + return data_batch + + +class Text2ImageInference: + def __init__(self, experiment_name: str, ckpt_path: str, s3_credential_path: str): + self.experiment_name = experiment_name + self.ckpt_path = ckpt_path + self.s3_credential_path = s3_credential_path + + model, config = load_model_from_checkpoint( + experiment_name=experiment_name, + config_file="cosmos_policy/_src/predict2/configs/text2world/config.py", + s3_checkpoint_dir=ckpt_path, + enable_fsdp=False, + load_ema_to_reg=True, + ) + self.model = model + self.config = config + self.resolution = str(self.model.config.resolution) # Store resolution from loaded model + + def generate_image( + self, prompt: str, neg_prompt: str, guidance: int = 7, aspect_ratio: str = "16,9", num_samples: int = 1 + ): + data_batch = get_sample_batch( + resolution=self.resolution, # Use resolution from loaded model + aspect_ratio=aspect_ratio, + batch_size=num_samples, + ) + + # modify the batch if prompt is provided + if self.model.text_encoder is not None: + # Text encoder is defined in the model class. Use it + if prompt: + data_batch["ai_caption"] = [prompt] + data_batch["t5_text_embeddings"] = self.model.text_encoder.compute_text_embeddings_online( + data_batch={"ai_caption": [prompt], "images": None}, + input_caption_key="ai_caption", + ) + if neg_prompt: + data_batch["neg_t5_text_embeddings"] = self.model.text_encoder.compute_text_embeddings_online( + data_batch={"ai_caption": [neg_prompt], "images": None}, + input_caption_key="ai_caption", + ) + else: + if prompt: + text_emb = get_text_embedding(prompt) + data_batch["t5_text_embeddings"] = text_emb.to(dtype=torch.bfloat16).cuda() + if neg_prompt: + text_emb = get_text_embedding(neg_prompt) + data_batch["neg_t5_text_embeddings"] = text_emb.to(dtype=torch.bfloat16).cuda() + + # generate samples + sample = self.model.generate_samples_from_batch( + data_batch, + guidance=guidance, + seed=torch.randint(0, 10000, (1,)).item(), # Use random seed for variation + is_negative_prompt=bool(neg_prompt), # Only set true if neg_prompt provided + ) + out_samples = self.model.decode(sample) + out_samples = (1.0 + out_samples) / 2 # Convert from [-1, 1] to [0, 1] + out_samples = out_samples.clamp(0, 1) # Clamp values + out_samples = out_samples.squeeze(2) # Convert the video tensor to image tensor + + # Now reshape + return out_samples diff --git a/REGEN-main/cosmos_policy/_src/predict2/inference/utils.py b/REGEN-main/cosmos_policy/_src/predict2/inference/utils.py new file mode 100644 index 0000000000000000000000000000000000000000..f2c739475dd5a5573ffd6534d205ff773efb2c65 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/inference/utils.py @@ -0,0 +1,410 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import fnmatch +import math +import os +from glob import glob + +import mediapy as media +import numpy as np +import torch +from loguru import logger +from mediapy import _VideoArray +from PIL import Image +from torchvision.transforms import functional as F + +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.predict2.datasets.utils import IMAGE_RES_SIZE_INFO + +_CREDENTIAL, _BACKEND = "credentials/pdx_cosmos_base.secret", "s3" +_DTYPE, _DEVICE = torch.bfloat16, "cuda" +_UINT8_MAX_F = float(torch.iinfo(torch.uint8).max) + +_PROMPT_EXTENSIONS = [".txt"] +_IMAGE_EXTENSIONS = [".png", ".jpg", ".jpeg", ".webp"] +_VIDEO_EXTENSIONS = [".mp4"] + +_DEFAULT_NEGATIVE_PROMPT = "The video captures a series of frames showing ugly scenes, static with no motion, motion blur, over-saturation, shaky footage, low resolution, grainy texture, pixelated images, poorly lit areas, underexposed and overexposed scenes, poor color balance, washed out colors, choppy sequences, jerky movements, low frame rate, artifacting, color banding, unnatural transitions, outdated special effects, fake elements, unconvincing visuals, poorly edited content, jump cuts, visual noise, and flickering. Overall, the video is of poor quality." + + +def get_sample_batch( + resolution: str = "1024", + aspect_ratio: str = "16,9", + batch_size: int = 1, +): + w, h = IMAGE_RES_SIZE_INFO[resolution][aspect_ratio] + data_batch = { + "dataset_name": "image_data", + "images": torch.randn(batch_size, 3, h, w).cuda(), + "t5_text_embeddings": torch.randn(batch_size, 512, 1024).cuda(), + "fps": torch.randint(16, 32, (batch_size,)).cuda(), + "padding_mask": torch.zeros(batch_size, 1, h, w).cuda(), + } + + for k, v in data_batch.items(): + if isinstance(v, torch.Tensor) and torch.is_floating_point(data_batch[k]): + data_batch[k] = v.cuda().to(dtype=torch.bfloat16) + + return data_batch + + +def resize_input(video: torch.Tensor, resolution: tuple[int, int]): + r""" + Resizes and crops the input video tensor while preserving aspect ratio. + + The video is first resized so that the smaller dimension matches the target resolution, + preserving the aspect ratio. Then, it's center-cropped to the target resolution. + + Args: + video (torch.Tensor): Input video tensor of shape (T, C, H, W). + resolution (list[int]): Target resolution [H, W]. + + Returns: + torch.Tensor: Resized and cropped video tensor of shape (T, C, target_H, target_W). + """ + + orig_h, orig_w = video.shape[2], video.shape[3] + target_h, target_w = resolution + + scaling_ratio = max((target_w / orig_w), (target_h / orig_h)) + resizing_shape = (int(math.ceil(scaling_ratio * orig_h)), int(math.ceil(scaling_ratio * orig_w))) + video_resized = F.resize(video, list(resizing_shape)) + video_cropped = F.center_crop(video_resized, list(resolution)) + return video_cropped + + +def read_and_process_image(img_path: str, resolution: tuple[int, int], num_video_frames: int, resize: bool = True): + """ + Reads an image, converts it to a video tensor, and processes it for model input. + + The image is loaded, converted to a tensor, and replicated to match the + `num_video_frames`. It's then optionally resized and permuted to the + standard video format (B, C, T, H, W). + + Args: + img_path (str): Path to the input image file. + resolution (list[int]): Target resolution [H, W] for resizing. + num_video_frames (int): Number of frames needed by the model (should equal model.tokenizer.get_pixel_num_frames(model.config.state_t)). + resize (bool, optional): Whether to resize the image to the target resolution. Defaults to True. + + Returns: + torch.Tensor: Processed video tensor of shape (1, C, T, H, W). + + Raises: + ValueError: If the image extension is not one of the supported types. + """ + ext = os.path.splitext(img_path)[1] + if ext not in _IMAGE_EXTENSIONS: + raise ValueError(f"Invalid image extension: {ext}") + + # Read the image + img = Image.open(img_path) + + # Convert to tensor + img = F.to_tensor(img) + # Create a video tensor by repeating the first frame + vid_input = img.unsqueeze(0) # Add temporal dimension T=1 + + # Repeat the first frame to match the desired number of video frames + # Note: The actual content for frames > 0 will be generated by the model. + vid_input = torch.cat([vid_input, torch.zeros_like(vid_input).repeat(num_video_frames - 1, 1, 1, 1)], dim=0) + vid_input = (vid_input * 255.0).to(torch.uint8) # Convert to uint8 range if needed (might depend on model) + if resize: + # Resize and crop to the target resolution + vid_input = resize_input(vid_input, resolution) + + # Convert to {B, C, T, H, W} format expected by the model + vid_input = vid_input.unsqueeze(0).permute(0, 2, 1, 3, 4) # Add batch dim B=1 and permute + return vid_input + + +def read_and_process_video( + video_path: str, + resolution: tuple[int, int], + num_video_frames: int, + num_latent_conditional_frames: int = 2, + resize: bool = True, +): + """ + Reads a video, processes it for model input. + + The video is loaded using easy_io, and uses the last 4x(num_latent_conditional_frames - 1) + 1 from the video. + If the video is shorter than num_video_frames, it pads with the last frame repeated. + The first num_latent_conditional_frames are marked as conditioning frames. + + Args: + video_path (str): Path to the input video file. + resolution (list[int]): Target resolution [H, W] for resizing. + num_video_frames (int): Number of frames needed by the model (should equal model.tokenizer.get_pixel_num_frames(model.config.state_t)). + num_latent_conditional_frames (int): Number of latent conditional frames from the input video (1 or 2). + resize (bool, optional): Whether to resize the video to the target resolution. Defaults to True. + + Returns: + torch.Tensor: Processed video tensor of shape (1, C, T, H, W) where T equals num_video_frames. + + Raises: + ValueError: If the video extension is not supported or other validation errors. + + Note: + Uses the last 4x(num_latent_conditional_frames - 1) + 1 frames from the video. If video is shorter, pads with last frame repeated. + """ + ext = os.path.splitext(video_path)[1] + if ext.lower() not in _VIDEO_EXTENSIONS: + raise ValueError(f"Invalid video extension: {ext}") + + # Load video using easy_io + try: + video_frames, video_metadata = easy_io.load(video_path) # Returns (T, H, W, C) numpy array + logger.info(f"Loaded video with shape {video_frames.shape}, metadata: {video_metadata}") + except Exception as e: + raise ValueError(f"Failed to load video {video_path}: {e}") + + # Convert numpy array to tensor and rearrange dimensions + video_tensor = torch.from_numpy(video_frames).float() / 255.0 # Convert to [0, 1] range + video_tensor = video_tensor.permute(3, 0, 1, 2) # (T, H, W, C) -> (C, T, H, W) + + available_frames = video_tensor.shape[1] + + # Calculate how many frames to extract from input video + frames_to_extract = 4 * (num_latent_conditional_frames - 1) + 1 + logger.info(f"Will extract {frames_to_extract} frames from input video and pad to {num_video_frames}") + + # Validate num_latent_conditional_frames + if num_latent_conditional_frames not in [1, 2]: + raise ValueError(f"num_latent_conditional_frames must be 1 or 2, but got {num_latent_conditional_frames}") + + # Create output tensor with exact num_video_frames + C, _, H, W = video_tensor.shape + full_video = torch.zeros(C, num_video_frames, H, W) + + if available_frames < frames_to_extract: + raise ValueError( + f"Video has only {available_frames} frames but needs at least {frames_to_extract} frames for num_latent_conditional_frames={num_latent_conditional_frames}" + ) + + # Extract the last frames_to_extract from input video + start_idx = available_frames - frames_to_extract + extracted_frames = video_tensor[:, start_idx:, :, :] + full_video[:, :frames_to_extract, :, :] = extracted_frames + logger.info(f"Extracted last {frames_to_extract} frames from video (frames {start_idx} to {available_frames - 1})") + + # Pad remaining frames with the last extracted frame + if frames_to_extract < num_video_frames: + last_frame = extracted_frames[:, -1:, :, :] # (C, 1, H, W) + padding_frames = num_video_frames - frames_to_extract + last_frame_repeated = last_frame.repeat(1, padding_frames, 1, 1) # (C, padding_frames, H, W) + full_video[:, frames_to_extract:, :, :] = last_frame_repeated + logger.info(f"Padded {padding_frames} frames with last extracted frame") + + # Convert to the format expected by the rest of the pipeline + full_video = full_video.permute(1, 0, 2, 3) # (C, T, H, W) -> (T, C, H, W) + full_video = (full_video * 255.0).to(torch.uint8) # Convert to uint8 range + + if resize: + # Resize and crop to the target resolution + full_video = resize_input(full_video, resolution) + + # Convert to {B, C, T, H, W} format expected by the model + full_video = full_video.unsqueeze(0).permute(0, 2, 1, 3, 4) # Add batch dim B=1 and permute + return full_video + + +def set_s3_backend(backend: str = _BACKEND, credentials: str = _CREDENTIAL) -> None: + """Set the backend with the proper credentials.""" + credentials = credentials or _CREDENTIAL + easy_io.set_s3_backend( + backend_args={ + "backend": backend, + "s3_credential_path": credentials, + } + ) + + +def get_filepaths(input_pattern: str) -> list[str]: + """Returns a list of filepaths from a pattern, supporting wildcards.""" + if input_pattern.startswith("s3://"): + return _get_s3_filepaths(input_pattern) + else: + filepaths = glob(str(input_pattern)) + return sorted(list(set(filepaths))) + + +def _get_s3_filepaths(s3_pattern: str) -> list[str]: + """Get S3 filepaths matching a pattern with wildcards.""" + # Parse the pattern to find the base directory and pattern + pattern_parts = s3_pattern.replace("s3://", "").split("/") + + # Find the first part with wildcards + base_parts = [] + pattern_start_index = -1 + + for i, part in enumerate(pattern_parts): + if "*" in part or "?" in part or "[" in part: + pattern_start_index = i + break + base_parts.append(part) + + if pattern_start_index == -1: + # No wildcards, just check if the file exists + if easy_io.exists(s3_pattern): + return [s3_pattern] + else: + return [] + + # Build the base directory path + base_dir = "s3://" + "/".join(base_parts) if base_parts else "s3://" + + # Build the pattern for matching (everything after the base directory) + pattern_suffix = "/".join(pattern_parts[pattern_start_index:]) + + # Use recursive listing to get all files under the base directory + filepaths = [] + try: + for relative_path in easy_io.list_dir_or_file( + base_dir, + list_dir=False, # Only list files, not directories + list_file=True, + recursive=True, # This is the key - recursive listing + ): + # Check if this relative path matches our pattern + if fnmatch.fnmatch(relative_path, pattern_suffix): + full_path = f"{base_dir.rstrip('/')}/{relative_path}" + filepaths.append(full_path) + except Exception: + # If listing fails, return empty list + pass + + return sorted(list(set(filepaths))) + + +def read_video(filepath: str) -> np.ndarray: + """Reads a video from a filepath in S3 or local. + + Args: + filepath: The filepath to the video. (local or S3) + Returns: + The video as a numpy array, layout TxHxWxC, range [0..255], uint8 dtype. + """ + if filepath.startswith("s3://"): + video_data, metadata = easy_io.load(filepath) + video = _VideoArray(video_data, metadata) + else: + video = media.read_video(filepath) + # convert the grey scale image to RGB + # since our tokenizers always assume 3-channel RGB image + if video.ndim == 3: + video = np.stack([video] * 3, axis=-1) + # convert RGBA to RGB + if video.shape[-1] == 4: + video = video[..., :3] + return video + + +def _pad_to_even(video: np.ndarray) -> np.ndarray: + """Pads video frames to even height and width if necessary. + + Args: + video: A numpy array of shape (T, H, W, C) in range [0..255], uint8 dtype. + Returns: + A numpy array of shape (T, H, W, C) in range [0..255], uint8 dtype. + """ + H, W = video.shape[-3:-1] + pad_h = H % 2 + pad_w = W % 2 + if pad_h == 0 and pad_w == 0: + return video + pad = ((0, 0), (0, pad_h), (0, pad_w), (0, 0)) + return np.pad(video, pad_width=pad, mode="edge") + + +def write_video(filepath: str, video: np.ndarray, fps: int = 24, lossless: bool = True) -> None: + """Writes a video to a filepath in S3 or local. + + Args: + filepath: A string filepath to save the video. For S3, the filepath should start with s3://. + video: A numpy array of shape (T, H, W, C) in range [0..255], uint8 dtype. + fps: The frames per second of the video. + lossless: Whether to use lossless compression. + """ + video = _pad_to_even(video) + if lossless: + ffmpeg_params = [ + "-c:v", + "libx264", # Use H.264 codec + "-preset", + "veryslow", # Slowest preset = best compression + "-qp", + "0", # Quantization parameter 0 = lossless + "-crf", + "0", # Constant Rate Factor 0 = lossless + ] + else: + ffmpeg_params = [ + "-c:v", + "libx264", + "-preset", + "veryslow", + "-crf", + "23", # Reasonable quality–compression tradeoff + ] + easy_io.dump(video, filepath, fps=fps, quality=None, ffmpeg_params=ffmpeg_params) + + +def write_image(filepath: str, image: np.ndarray, quality: int = 85) -> None: + """Writes an image to a filepath in S3 or local. + + Args: + filepath: A string filepath to save the image. For S3, the filepath should start with s3://. + image: A numpy array of shape (H, W, C) in range [0..255], uint8 dtype. + quality: The quality of the image, on a scale from 0 (worst) to 95 (best), default=85. + https://pillow.readthedocs.io/en/stable/handbook/image-file-formats.html#jpeg + """ + pil_image = Image.fromarray(image) + easy_io.dump(pil_image, filepath, quality=quality) + + +def numpy2tensor( + input_image: np.ndarray, dtype: torch.dtype = _DTYPE, device: str = _DEVICE, range_min: int = -1 +) -> torch.Tensor: + """Converts image(dtype=np.uint8) to `dtype` in range [0..255]. + + Args: + input_image: A batch of images in range [0..255], BxHxWx3 layout. + Returns: + A torch.Tensor of layout Bx3xHxW in range [-1..1], dtype. + """ + ndim = input_image.ndim + indices = list(range(1, ndim))[-1:] + list(range(1, ndim))[:-1] + image = input_image.transpose((0,) + tuple(indices)) / _UINT8_MAX_F + if range_min == -1: + image = 2.0 * image - 1.0 + return torch.from_numpy(image).to(dtype).to(device) + + +def tensor2numpy(input_tensor: torch.Tensor, range_min: int = -1) -> np.ndarray: + """Converts tensor in [-1,1] to image(dtype=np.uint8) in range [0..255]. + + Args: + input_tensor: Input image tensor of Bx3xHxW layout, range [-1..1]. + Returns: + A numpy image of layout BxHxWx3, range [0..255], uint8 dtype. + """ + if range_min == -1: + input_tensor = (input_tensor.float() + 1.0) / 2.0 + ndim = input_tensor.ndim + output_image = input_tensor.clamp(0, 1).cpu().numpy() + output_image = output_image.transpose((0,) + tuple(range(2, ndim)) + (1,)) + return (output_image * _UINT8_MAX_F + 0.5).astype(np.uint8) diff --git a/REGEN-main/cosmos_policy/_src/predict2/inference/video2world.py b/REGEN-main/cosmos_policy/_src/predict2/inference/video2world.py new file mode 100644 index 0000000000000000000000000000000000000000..026c2b85c6a66765347c7234aa9a8e558630454b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/inference/video2world.py @@ -0,0 +1,866 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +# Script for generating I2W videos in s3 +PYTHONPATH=. python cosmos_policy/_src/predict2/inference/video2world.py --experiment=Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4 --ckpt_path s3://bucket/cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000 --save_root results/cli_debug_from_s3 --input_root /project/cosmos/ybalaji/data/internal_val_set_clean + +# Script for text2world generation +export EXPERIMENT=Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-T2V_high_sigma_loss_reweighted +CUDA_VISIBLE_DEVICES=0 PYTHONPATH=. python cosmos_policy/_src/predict2/inference/video2world.py \ +--experiment=${EXPERIMENT} \ +--ckpt_path s3://bucket/cosmos_diffusion_v2/official_runs_vid2vid/${EXPERIMENT}/checkpoints/iter_000025000 \ +--save_root results/base_model/${EXPERIMENT}_025k_seed0_t2w \ +--num_latent_conditional_frames=0 --seed=0 \ +--input_root /project/cosmos/fangyinw/data/pbench/v0 + +# I2W with context parallel with 8 GPUs: +PYTHONPATH=. torchrun --nproc_per_node=8 cosmos_policy/_src/predict2/inference/video2world.py --experiment=Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4 --ckpt_path s3://bucket/cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000 --save_root results/cli_debug_from_s3 --input_root /project/cosmos/ybalaji/data/internal_val_set_clean --context_parallel_size 8 + +# V2W with context parallel with 8 GPUs: +PYTHONPATH=. torchrun --nproc_per_node=8 cosmos_policy/_src/predict2/inference/video2world.py --experiment=Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4 --ckpt_path s3://bucket/cosmos_diffusion_v2/official_runs_vid2vid/Stage-c_pt_4-reason_embeddings-Index-26-Size-2B-Res-720-Fps-16-Note-HQ_V6_from_22_qwen_concat_resume4/checkpoints/iter_000045000 --save_root results/cli_debug_from_s3 --input_root pbench_upsampled_prompts --num_latent_conditional_frames=2 --context_parallel_size=8 + + +Folder structure: +We assume the input root contains images and prompts in the following format: +input_root/ + ├── image_1.jpg + ├── image_1.txt + ├── image_2.jpg + └── image_2.txt + └── ... + +or videos and prompts in the following format: +input_root/ + ├── video_1.mp4 + ├── video_1.txt + ├── video_2.mp4 + └── video_2.txt + └── ... +""" + +import math +import os +from typing import TYPE_CHECKING + +import torch +import torchvision +from megatron.core import parallel_state +from PIL import Image + +from cosmos_policy._src.imaginaire.flags import INTERNAL +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.predict2.inference.get_t5_emb import get_text_embedding +from cosmos_policy._src.predict2.utils.model_loader import load_model_from_checkpoint + +_IMAGE_EXTENSIONS = [".png", ".jpg", ".jpeg", ".webp"] +_VIDEO_EXTENSIONS = [".mp4"] + +_DEFAULT_NEGATIVE_PROMPT = "The video captures a series of frames showing ugly scenes, static with no motion, motion blur, over-saturation, shaky footage, low resolution, grainy texture, pixelated images, poorly lit areas, underexposed and overexposed scenes, poor color balance, washed out colors, choppy sequences, jerky movements, low frame rate, artifacting, color banding, unnatural transitions, outdated special effects, fake elements, unconvincing visuals, poorly edited content, jump cuts, visual noise, and flickering. Overall, the video is of poor quality." + + +def resize_input(video: torch.Tensor, resolution: list[int]): + r""" + Resizes and crops the input video tensor while preserving aspect ratio. + + The video is first resized so that the smaller dimension matches the target resolution, + preserving the aspect ratio. Then, it's center-cropped to the target resolution. + + Args: + video (torch.Tensor): Input video tensor of shape (T, C, H, W). + resolution (list[int]): Target resolution [H, W]. + + Returns: + torch.Tensor: Resized and cropped video tensor of shape (T, C, target_H, target_W). + """ + + orig_h, orig_w = video.shape[2], video.shape[3] + target_h, target_w = resolution + + scaling_ratio = max((target_w / orig_w), (target_h / orig_h)) + resizing_shape = (int(math.ceil(scaling_ratio * orig_h)), int(math.ceil(scaling_ratio * orig_w))) + video_resized = torchvision.transforms.functional.resize(video, resizing_shape) + video_cropped = torchvision.transforms.functional.center_crop(video_resized, resolution) + return video_cropped + + +def read_and_process_image(img_path: str, resolution: list[int], num_video_frames: int, resize: bool = True): + """ + Reads an image, converts it to a video tensor, and processes it for model input. + + The image is loaded, converted to a tensor, and replicated to match the + `num_video_frames`. It's then optionally resized and permuted to the + standard video format (B, C, T, H, W). + + Args: + img_path (str): Path to the input image file. + resolution (list[int]): Target resolution [H, W] for resizing. + num_video_frames (int): The number of frames the output video tensor should have. + resize (bool, optional): Whether to resize the image to the target resolution. Defaults to True. + + Returns: + torch.Tensor: Processed video tensor of shape (1, C, T, H, W). + + Raises: + ValueError: If the image extension is not one of the supported types. + """ + ext = os.path.splitext(img_path)[1] + if ext not in _IMAGE_EXTENSIONS: + raise ValueError(f"Invalid image extension: {ext}") + + # Read the image + img = Image.open(img_path) + + # Convert to tensor + img = torchvision.transforms.functional.to_tensor(img) + # Create a video tensor by repeating the first frame + vid_input = img.unsqueeze(0) # Add temporal dimension T=1 + + # Repeat the first frame to match the desired number of video frames + # Note: The actual content for frames > 0 will be generated by the model. + vid_input = torch.cat([vid_input, torch.zeros_like(vid_input).repeat(num_video_frames - 1, 1, 1, 1)], dim=0) + vid_input = (vid_input * 255.0).to(torch.uint8) # Convert to uint8 range if needed (might depend on model) + if resize: + # Resize and crop to the target resolution + vid_input = resize_input(vid_input, resolution) + + # Convert to {B, C, T, H, W} format expected by the model + vid_input = vid_input.unsqueeze(0).permute(0, 2, 1, 3, 4) # Add batch dim B=1 and permute + return vid_input + + +def read_and_process_video( + video_path: str, + resolution: list[int], + num_video_frames: int, + num_latent_conditional_frames: int = 2, + resize: bool = True, +): + """ + Reads a video, processes it for model input. + + The video is loaded using easy_io, and uses the last 4x(num_latent_conditional_frames - 1) + 1 from the video. + If the video is shorter than num_video_frames, it pads with the last frame repeated. + The first num_latent_conditional_frames are marked as conditioning frames. + + Args: + video_path (str): Path to the input video file. + resolution (list[int]): Target resolution [H, W] for resizing. + num_video_frames (int): Number of frames needed by the model (should equal model.tokenizer.get_pixel_num_frames(model.config.state_t)). + num_latent_conditional_frames (int): Number of latent conditional frames from the input video (1 or 2). + resize (bool, optional): Whether to resize the video to the target resolution. Defaults to True. + + Returns: + torch.Tensor: Processed video tensor of shape (1, C, T, H, W) where T equals num_video_frames. + + Raises: + ValueError: If the video extension is not supported or other validation errors. + + Note: + Uses the last 4x(num_latent_conditional_frames - 1) + 1 frames from the video. If video is shorter, pads with last frame repeated. + """ + ext = os.path.splitext(video_path)[1] + if ext.lower() not in _VIDEO_EXTENSIONS: + raise ValueError(f"Invalid video extension: {ext}") + + # Load video using easy_io + try: + video_frames, video_metadata = easy_io.load(video_path) # Returns (T, H, W, C) numpy array + log.info(f"Loaded video with shape {video_frames.shape}, metadata: {video_metadata}") + except Exception as e: + raise ValueError(f"Failed to load video {video_path}: {e}") + + # Convert numpy array to tensor and rearrange dimensions + video_tensor = torch.from_numpy(video_frames).float() / 255.0 # Convert to [0, 1] range + video_tensor = video_tensor.permute(3, 0, 1, 2) # (T, H, W, C) -> (C, T, H, W) + + available_frames = video_tensor.shape[1] + + # Calculate how many frames to extract from input video + frames_to_extract = 4 * (num_latent_conditional_frames - 1) + 1 + log.info(f"Will extract {frames_to_extract} frames from input video and pad to {num_video_frames}") + + # Validate num_latent_conditional_frames + if num_latent_conditional_frames not in [1, 2]: + raise ValueError(f"num_latent_conditional_frames must be 1 or 2, but got {num_latent_conditional_frames}") + + # Create output tensor with exact num_video_frames + C, _, H, W = video_tensor.shape + full_video = torch.zeros(C, num_video_frames, H, W) + + if available_frames < frames_to_extract: + raise ValueError( + f"Video has only {available_frames} frames but needs at least {frames_to_extract} frames for num_latent_conditional_frames={num_latent_conditional_frames}" + ) + + # Extract the last frames_to_extract from input video + start_idx = available_frames - frames_to_extract + extracted_frames = video_tensor[:, start_idx:, :, :] + full_video[:, :frames_to_extract, :, :] = extracted_frames + log.info(f"Extracted last {frames_to_extract} frames from video (frames {start_idx} to {available_frames - 1})") + + # Pad remaining frames with the last extracted frame + if frames_to_extract < num_video_frames: + last_frame = extracted_frames[:, -1:, :, :] # (C, 1, H, W) + padding_frames = num_video_frames - frames_to_extract + last_frame_repeated = last_frame.repeat(1, padding_frames, 1, 1) # (C, padding_frames, H, W) + full_video[:, frames_to_extract:, :, :] = last_frame_repeated + log.info(f"Padded {padding_frames} frames with last extracted frame") + + # Convert to the format expected by the rest of the pipeline + full_video = full_video.permute(1, 0, 2, 3) # (C, T, H, W) -> (T, C, H, W) + full_video = (full_video * 255.0).to(torch.uint8) # Convert to uint8 range + + if resize: + # Resize and crop to the target resolution + full_video = resize_input(full_video, resolution) + + # Convert to {B, C, T, H, W} format expected by the model + full_video = full_video.unsqueeze(0).permute(0, 2, 1, 3, 4) # Add batch dim B=1 and permute + return full_video + + +class Video2WorldInference: + """ + Handles the Video2World inference process, including model loading, data preparation, + and video generation from an image/video and text prompt. Now supports context parallelism. + """ + + def __init__( + self, + experiment_name: str, + ckpt_path: str, + s3_credential_path: str, + context_parallel_size: int = 1, + config_file: str = "cosmos_policy/_src/predict2/configs/video2world/config.py", + offload_diffusion_model: bool = False, + offload_text_encoder: bool = False, + offload_tokenizer: bool = False, + ): + """ + Initializes the Video2WorldInference class. + + Loads the diffusion model and its configuration based on the provided + experiment name and checkpoint path. Sets up distributed processing if needed. + + Args: + experiment_name (str): Name of the experiment configuration. + ckpt_path (str): Path to the model checkpoint (local or S3). + s3_credential_path (str): Path to S3 credentials file (if loading from S3). + context_parallel_size (int): Number of GPUs for context parallelism. + """ + self.experiment_name = experiment_name + self.ckpt_path = ckpt_path + self.s3_credential_path = s3_credential_path + self.context_parallel_size = context_parallel_size + self.process_group = None + + self.offload_diffusion_model = offload_diffusion_model + self.offload_text_encoder = offload_text_encoder + self.offload_tokenizer = offload_tokenizer + + # If no offloading is specified, instruct model loader to move the model to GPU + model_device = None if offload_diffusion_model else "cuda" + + # Initialize distributed processing if context parallel size > 1 + if self.context_parallel_size > 1: + self._init_distributed() + + # Load the model and config + experiment_opts = [] + if not INTERNAL: + experiment_opts.append("~data_train") + + # LazyConfig interference is not available yet + # Use envvar to control whether DiT should be offloaded immediately after ctor + if self.offload_diffusion_model: + os.environ["COSMOS_PREDICT2_OFFLOAD_DIT"] = "1" + + model, config = load_model_from_checkpoint( + experiment_name=self.experiment_name, + s3_checkpoint_dir=self.ckpt_path, + config_file=config_file, + load_ema_to_reg=True, + experiment_opts=experiment_opts, + to_device=model_device, + ) + + # By default, everything will be constructed directly on the GPU (except DiT) + # Handle offloading options at inference entry + + # [On-entry offloading part 1]: DiT was offloaded as default by the lazy ctor + # Offload or reload according to setup + if self.offload_diffusion_model: + log.info("[Memory Optimization] Offloading DiT conditioner to CPU") + if hasattr(model, "conditioner") and model.conditioner is not None: + model.conditioner = model.conditioner.to("cpu") + else: + # Move everything to the GPU (marginal overhead) + model.net.to("cuda") + + # [On-entry offloading part 2]: Tokenizer + if self.offload_tokenizer: + log.info("[Memory Optimization] Offloading tokenizer encoder & decoder to CPU") + if hasattr(model.tokenizer, "encoder") and model.tokenizer.encoder is not None: + model.tokenizer.encoder = model.tokenizer.encoder.to("cpu") + if hasattr(model.tokenizer, "decoder") and model.tokenizer.decoder is not None: + model.tokenizer.decoder = model.tokenizer.decoder.to("cpu") + torch.cuda.empty_cache() + + # [On-entry offloading part 3]: Text encoder + if self.offload_text_encoder: + # Text encoder is the first module in the pipeline. + # Rather offload it **during** DiT run. + pass + + if TYPE_CHECKING: + from cosmos_policy._src.predict2.models.video2world_model_rectified_flow import ( + Video2WorldModelRectifiedFlow, + ) + + model: Video2WorldModelRectifiedFlow = model + + # Enable context parallel on the model if using context parallelism + if self.context_parallel_size > 1: + model.net.enable_context_parallel(self.process_group) + + self.model = model + self.config = config + self.batch_size = 1 + self.neg_t5_embeddings = None + + def _init_distributed(self): + """Initialize distributed processing for context parallelism.""" + + # Initialize distributed environment + distributed.init() + + # Initialize model parallel states + parallel_state.initialize_model_parallel( + context_parallel_size=self.context_parallel_size, + ) + + # Get the process group for context parallel + self.process_group = parallel_state.get_context_parallel_group() + + log.info(f"Initialized context parallel with size {self.context_parallel_size}") + log.info(f"Current rank: {distributed.get_rank()}, World size: {distributed.get_world_size()}") + + def _get_data_batch_input( + self, + video: torch.Tensor, + prompt: str, + num_conditional_frames: int = 1, + negative_prompt: str = _DEFAULT_NEGATIVE_PROMPT, + use_neg_prompt: bool = True, + camera: torch.Tensor | None = None, + action: torch.Tensor | None = None, + ): + """ + Prepares the input data batch for the diffusion model. + + Constructs a dictionary containing the video tensor, text embeddings, + and other necessary metadata required by the model's forward pass. + Optionally includes negative text embeddings. + + Args: + video (torch.Tensor): The input video tensor (B, C, T, H, W). + prompt (str): The text prompt for conditioning. + num_conditional_frames (int): Number of conditional frames to use. + negative_prompt (str, optional): Custom negative prompt. + use_neg_prompt (bool, optional): Whether to include negative prompt embeddings. Defaults to True. + camera: (torch.Tensor, optional) Target camera extrinsics and intrinsics for the K output videos, must be provided for camera conditioned model. + action: (torch.Tensor, optional) Target robot action for the K output videos, must be provided for action conditioned model. + + Returns: + dict: A dictionary containing the prepared data batch, moved to the correct device and dtype. + """ + B, C, T, H, W = video.shape + + data_batch = { + "dataset_name": "video_data", + "video": video, + "camera": camera, + "action": action.unsqueeze(0) if action is not None else None, + "fps": torch.randint(16, 32, (self.batch_size,)).float(), # Random FPS (might be used by model) + "padding_mask": torch.zeros(self.batch_size, 1, H, W), # Padding mask (assumed no padding here) + "num_conditional_frames": num_conditional_frames, # Specify number of conditional frames + } + + if use_neg_prompt: + assert negative_prompt is not None, "Negative prompt is required when use_neg_prompt is True" + + # Compute text embeddings + if self.model.text_encoder is not None: + data_batch["ai_caption"] = [prompt] + data_batch["t5_text_embeddings"] = self.model.text_encoder.compute_text_embeddings_online( + data_batch={"ai_caption": [prompt], "images": None}, + input_caption_key="ai_caption", + ) + if use_neg_prompt: + data_batch["neg_t5_text_embeddings"] = self.model.text_encoder.compute_text_embeddings_online( + data_batch={"ai_caption": [negative_prompt], "images": None}, + input_caption_key="ai_caption", + ) + else: + data_batch["t5_text_embeddings"] = get_text_embedding(prompt) + if use_neg_prompt: + data_batch["neg_t5_text_embeddings"] = get_text_embedding(negative_prompt) + + # Move tensors to GPU and convert to bfloat16 if they are floating point + for k, v in data_batch.items(): + if isinstance(v, torch.Tensor) and torch.is_floating_point(data_batch[k]): + data_batch[k] = v.cuda().to(dtype=torch.bfloat16) + + return data_batch + + def generate_vid2world( + self, + prompt: str, + input_path: str | torch.Tensor | None, + guidance: int = 7, + num_video_frames: int = 77, + num_latent_conditional_frames: int = 1, + num_input_video: int = 1, + num_output_video: int = 1, + resolution: str = "192,320", + seed: int = 1, + negative_prompt: str = _DEFAULT_NEGATIVE_PROMPT, + camera: torch.Tensor | None = None, + action: torch.Tensor | None = None, + num_steps: int = 35, + ): + """ + Generates a video based on an input image or video and text prompt. + + Processes the input, prepares the data batch, runs the diffusion + model sampling, and decodes the result into a video tensor. + + Args: + prompt: The text prompt describing the desired video content/style. + input_path: Path to the input image or video file or a torch.Tensor. + guidance: Classifier-free guidance scale. Defaults to 7. + num_video_frames: Number of video frames to generate. Defaults to 77. + num_latent_conditional_frames : Number of latent conditional frames. Defaults to 1. + resolution: Target video resolution in "H,W" format. Defaults to "192,320". + seed: Random seed for reproducibility. Defaults to 1. + negative_prompt: Custom negative prompt. Defaults to the predefined default negative prompt. + camera: Target camera extrinsics and intrinsics for the K output videos. Must be provided if model is camera conditioned. + action: Target robot action for the K output videos. Must be provided if model is action conditioned. + num_steps: Number of generation steps. Defaults to 35. + offload_diffusion_model: If True, offload diffusion model to CPU to save GPU memory. Defaults to False. + offload_text_encoder: If True, offload text encoder to CPU to save GPU memory. Defaults to False. + offload_tokenizer: If True, offload tokenizer to CPU to save GPU memory. Defaults to False. + + Returns: + torch.Tensor: The generated video tensor (B, C, T, H, W) in the range [-1, 1]. + """ + assert camera is not None or action is not None or num_input_video == 1 and num_output_video == 1, ( + "expected num_output_video==1 and num_output_video==1 for no camera conditioning or action conditioning" + ) + + # Parse resolution string into tuple of integers + if resolution == "none": + h, w = self.model.get_video_height_width() + video_resolution = (h, w) + else: + video_resolution = resolution.split(",") + video_resolution = tuple([int(x) for x in video_resolution]) + assert len(video_resolution) == 2, "Resolution must be in 'H,W' format" + + # Get the correct number of frames needed by the model + model_required_frames = self.model.tokenizer.get_pixel_num_frames(self.model.config.state_t) + + # Determine if input is image or video and process accordingly + if input_path is None or num_latent_conditional_frames == 0: + vid_input = torch.zeros(1, 3, model_required_frames, video_resolution[0], video_resolution[1]).to( + torch.uint8 + ) + elif isinstance(input_path, str): + ext = os.path.splitext(input_path)[1].lower() + if ext in _IMAGE_EXTENSIONS: + log.info(f"Processing image input: {input_path}") + vid_input = read_and_process_image( + img_path=input_path, + resolution=video_resolution, + num_video_frames=model_required_frames, + resize=True, + ) + elif ext in _VIDEO_EXTENSIONS: + log.info(f"Processing video input: {input_path}") + vid_input = read_and_process_video( + video_path=input_path, + resolution=video_resolution, + num_video_frames=model_required_frames, + num_latent_conditional_frames=num_latent_conditional_frames, + resize=True, + ) + else: + raise ValueError( + f"Unsupported file extension: {ext}. Supported extensions: {_IMAGE_EXTENSIONS + _VIDEO_EXTENSIONS}" + ) + elif isinstance(input_path, torch.Tensor): + vid_input = input_path + else: + raise ValueError(f"Unsupported input_path type: {type(input_path)}") + + # Prepare the data batch with text embeddings + # Note: TextEncoder.compute_text_embeddings_online() will automatically move its model to GPU + data_batch = self._get_data_batch_input( + video=vid_input, + prompt=prompt, + camera=camera, + action=action, + num_conditional_frames=num_latent_conditional_frames, + negative_prompt=negative_prompt, + use_neg_prompt=True, + ) + + mem_bytes = torch.cuda.memory_allocated(device=torch.device("cuda" if torch.cuda.is_available() else "cpu")) + log.info(f"GPU memory usage after getting data_batch: {mem_bytes / (1024**3):.2f} GB") + + # Memory Optimization Step 1: Offload Text Encoder + # Offload text encoder after computing embeddings to free memory + if self.offload_text_encoder and self.model.text_encoder is not None: + log.info("[Memory Optimization] Offloading text encoder to CPU") + # TextEncoder is a wrapper class with self.model (the actual neural network) + if hasattr(self.model.text_encoder, "model") and self.model.text_encoder.model is not None: + self.model.text_encoder.model = self.model.text_encoder.model.to("cpu") + torch.cuda.empty_cache() + + # Memory Optimization Step 2: Tokenizer Encoder + # Load tokenizer encoder to GPU for encoding input video + if self.offload_tokenizer: + log.info("[Memory Optimization] Loading tokenizer encoder to GPU") + if hasattr(self.model.tokenizer, "encoder") and self.model.tokenizer.encoder is not None: + self.model.tokenizer.encoder = self.model.tokenizer.encoder.to("cuda") + torch.cuda.empty_cache() + + # Memory Optimization Step 3: Diffusion Network + # Load the main diffusion network to GPU for sampling + if self.offload_diffusion_model: + log.info("[Memory Optimization] Loading diffusion network to GPU") + self.model.net = self.model.net.to("cuda") + # Also load conditioner if it exists + if hasattr(self.model, "conditioner") and self.model.conditioner is not None: + self.model.conditioner = self.model.conditioner.to("cuda") + torch.cuda.empty_cache() + + extra_kwargs = {} + if camera is not None: + extra_kwargs = { + "num_input_video": num_input_video, + "num_output_video": num_output_video, + } + + # Generate latent samples using the diffusion model + # Video should be of shape torch.Size([1, 3, 93, 192, 320]) # Note: Shape check comment + log.info("[Memory Optimization] Starting latent sample generation") + if self.model.config.use_lora: + generate_samples = self.model.generate_samples_from_batch_lora + else: + generate_samples = self.model.generate_samples_from_batch + sample = generate_samples( + data_batch, + n_sample=1, # Generate one sample + guidance=guidance, + seed=seed, # Fixed seed for reproducibility + is_negative_prompt=True, # Use classifier-free guidance + num_steps=num_steps, + **extra_kwargs, + ) + + # Memory Optimization Step 4: Offload Diffusion Network + # Offload diffusion network after sampling to make room for decoder + if self.offload_diffusion_model: + log.info("[Memory Optimization] Offloading diffusion network to CPU") + self.model.net = self.model.net.to("cpu") + if hasattr(self.model, "conditioner") and self.model.conditioner is not None: + self.model.conditioner = self.model.conditioner.to("cpu") + + if self.offload_tokenizer: + # Also offload encoder since we only need decoder now + if hasattr(self.model.tokenizer, "encoder") and self.model.tokenizer.encoder is not None: + self.model.tokenizer.encoder = self.model.tokenizer.encoder.to("cpu") + torch.cuda.empty_cache() + + # Memory Optimization Step 5: Load Decoder + # Load tokenizer decoder to GPU for decoding latents + if self.offload_tokenizer: + log.info("[Memory Optimization] Loading tokenizer decoder to GPU") + if hasattr(self.model.tokenizer, "decoder") and self.model.tokenizer.decoder is not None: + self.model.tokenizer.decoder = self.model.tokenizer.decoder.to("cuda") + torch.cuda.empty_cache() + + # Decode the latent samples + if isinstance(sample, list): + # Decode the latent sample into a video tensor + video_list = [] + for sample_chunk in sample: + video_chunk = self.model.decode(sample_chunk) + video_list.append(video_chunk) + video = torch.cat(video_list, dim=3) + else: + # Decode the latent sample into a video tensor + video = self.model.decode(sample) + + # Memory Optimization Step 6: Final Cleanup + # Offload decoder after decoding & reload the tokenizer for the next inference call + if self.offload_tokenizer: + log.info("[Memory Optimization] Offloading tokenizer decoder to CPU") + if hasattr(self.model.tokenizer, "decoder") and self.model.tokenizer.decoder is not None: + self.model.tokenizer.decoder = self.model.tokenizer.decoder.to("cpu") + torch.cuda.empty_cache() + + if self.offload_text_encoder and self.model.text_encoder is not None: + log.info("[Memory Optimization] Load text encoder to GPU") + # TextEncoder is a wrapper class with self.model (the actual neural network) + if hasattr(self.model.text_encoder, "model") and self.model.text_encoder.model is not None: + self.model.text_encoder.model = self.model.text_encoder.model.to("cuda") + torch.cuda.empty_cache() + + return video + + def generate_autoregressive_from_batch( + self, + prompt: str, + input_path: str | torch.Tensor | None, + num_output_frames: int, + chunk_size: int, + chunk_overlap: int, + guidance: int = 7, + num_latent_conditional_frames: int = 1, + resolution: str = "192,320", + seed: int = 1, + negative_prompt: str = _DEFAULT_NEGATIVE_PROMPT, + camera: torch.Tensor | None = None, + action: torch.Tensor | None = None, + num_steps: int = 35, + ) -> torch.Tensor: + """ + Generate video using autoregressive sliding window approach. + + Args: + prompt: The text prompt describing the desired video content/style. + input_path: Path to the input image or video file or a torch.Tensor. + num_output_frames: Total number of frames to generate in the final output. + chunk_size: Number of frames per chunk (model's native capacity). + chunk_overlap: Number of overlapping frames between chunks. + guidance: Classifier-free guidance scale. + num_latent_conditional_frames: Number of latent conditional frames. + resolution: Target video resolution in "H,W" format. + seed: Random seed for reproducibility. + negative_prompt: Custom negative prompt. + camera: Target camera extrinsics and intrinsics for the K output videos. + action: Target robot action for the K output videos. + num_steps: Number of generation steps. + + Returns: + torch.Tensor: The generated video tensor (B, C, T, H, W) in the range [-1, 1]. + """ + # Parse resolution string into tuple of integers + if resolution == "none": + h, w = self.model.get_video_height_width() + video_resolution = (h, w) + else: + video_resolution = resolution.split(",") + video_resolution = tuple([int(x) for x in video_resolution]) + assert len(video_resolution) == 2, "Resolution must be in 'H,W' format" + + # Get the correct number of frames needed by the model + model_required_frames = self.model.tokenizer.get_pixel_num_frames(self.model.config.state_t) + + # Load and process the full input video/image + if input_path is None or num_latent_conditional_frames == 0: + # For text2world, create a full length zero video + full_input_video = torch.zeros(1, 3, num_output_frames, video_resolution[0], video_resolution[1]).to( + torch.uint8 + ) + elif isinstance(input_path, str): + ext = os.path.splitext(input_path)[1].lower() + if ext in _IMAGE_EXTENSIONS: + log.info(f"Processing image input for autoregressive: {input_path}") + # For image input, create full video with first frame as image, rest zeros + img = Image.open(input_path) + img = torchvision.transforms.functional.to_tensor(img) + img = img.unsqueeze(0) # Add temporal dimension T=1 + img = (img * 255.0).to(torch.uint8) + if video_resolution: + img = resize_input(img, video_resolution) + # Create full length video with first frame as image + full_input_video = torch.cat([img, torch.zeros_like(img).repeat(num_output_frames - 1, 1, 1, 1)], dim=0) + full_input_video = full_input_video.unsqueeze(0).permute(0, 2, 1, 3, 4) + elif ext in _VIDEO_EXTENSIONS: + log.info(f"Processing video input for autoregressive: {input_path}") + # Load video and extend to full length if needed + video_frames, _ = easy_io.load(input_path) + video_tensor = torch.from_numpy(video_frames).float() / 255.0 + video_tensor = video_tensor.permute(3, 0, 1, 2) # (T, H, W, C) -> (C, T, H, W) + available_frames = video_tensor.shape[1] + + # Calculate frames to extract + frames_to_extract = 4 * (num_latent_conditional_frames - 1) + 1 + if available_frames < frames_to_extract: + raise ValueError(f"Video has only {available_frames} frames but needs at least {frames_to_extract}") + + # Extract last frames_to_extract + start_idx = available_frames - frames_to_extract + extracted_frames = video_tensor[:, start_idx:, :, :] + + # Create full length tensor + C, _, H, W = video_tensor.shape + full_video = torch.zeros(C, num_output_frames, H, W) + full_video[:, :frames_to_extract, :, :] = extracted_frames + + # Pad with last frame + if frames_to_extract < num_output_frames: + last_frame = extracted_frames[:, -1:, :, :] + padding_frames = num_output_frames - frames_to_extract + last_frame_repeated = last_frame.repeat(1, padding_frames, 1, 1) + full_video[:, frames_to_extract:, :, :] = last_frame_repeated + + full_video = full_video.permute(1, 0, 2, 3) # (C, T, H, W) -> (T, C, H, W) + full_video = (full_video * 255.0).to(torch.uint8) + if video_resolution: + full_video = resize_input(full_video, video_resolution) + full_input_video = full_video.unsqueeze(0).permute(0, 2, 1, 3, 4) + else: + raise ValueError(f"Unsupported file extension: {ext}") + elif isinstance(input_path, torch.Tensor): + # If tensor, extend to full length + full_input_video = input_path + if full_input_video.shape[2] < num_output_frames: + # Pad with zeros + padding_frames = num_output_frames - full_input_video.shape[2] + padding = torch.zeros( + full_input_video.shape[0], + full_input_video.shape[1], + padding_frames, + full_input_video.shape[3], + full_input_video.shape[4], + ).to(full_input_video.dtype) + full_input_video = torch.cat([full_input_video, padding], dim=2) + else: + raise ValueError(f"Unsupported input_path type: {type(input_path)}") + + # Initialize output + generated_chunks = [] + + # Calculate number of chunks + # Note: All chunks generate chunk_size frames, we store all of chunk 0 and (chunk_size - chunk_overlap) from others + # Total stored = chunk_size + (num_chunks - 1) * (chunk_size - chunk_overlap) >= num_output_frames + effective_chunk_size = chunk_size - chunk_overlap + + # Solve for num_chunks: chunk_size + (num_chunks - 1) * effective_chunk_size >= num_output_frames + remaining_after_first = num_output_frames - chunk_size + if remaining_after_first <= 0: + num_chunks = 1 + else: + # Ceiling division to ensure we have enough frames for the last chunk. + num_chunks = 1 + (remaining_after_first + effective_chunk_size - 1) // effective_chunk_size + + log.info( + f"Generating {num_chunks} chunks with chunk_size={chunk_size}, chunk_overlap={chunk_overlap} " + f"for {num_output_frames} total frames" + ) + + # Generate chunks + current_input_video = full_input_video.clone() + + for chunk_idx in range(num_chunks): + # Calculate frame range for this chunk + # All chunks are positioned with stride (chunk_size - chunk_overlap) + start_frame = chunk_idx * effective_chunk_size + end_frame = min(start_frame + chunk_size, num_output_frames) + actual_chunk_size = end_frame - start_frame + + if start_frame >= num_output_frames: + break + + log.info(f"Processing chunk {chunk_idx + 1}/{num_chunks}, frames {start_frame}-{end_frame}") + + # Extract chunk from current input + chunk_input = current_input_video[:, :, start_frame:end_frame, :, :] + + # Pad to model_required_frames if needed + if actual_chunk_size < model_required_frames: + padding_frames = model_required_frames - actual_chunk_size + padding = torch.zeros( + chunk_input.shape[0], + chunk_input.shape[1], + padding_frames, + chunk_input.shape[3], + chunk_input.shape[4], + ).to(chunk_input.dtype) + chunk_input = torch.cat([chunk_input, padding], dim=2) + + # Determine num_conditional_frames for this chunk + if chunk_idx == 0: + chunk_num_conditional = num_latent_conditional_frames + else: + chunk_num_conditional = chunk_overlap + + # Generate chunk + chunk_video = self.generate_vid2world( + prompt=prompt, + input_path=chunk_input, + guidance=guidance, + num_video_frames=model_required_frames, + num_latent_conditional_frames=chunk_num_conditional, + resolution=resolution, + seed=seed + chunk_idx, + negative_prompt=negative_prompt, + camera=camera, + action=action, + num_steps=num_steps, + ) # Returns (1, C, T, H, W) + + # Extract only the actual generated frames (remove padding) + chunk_video = chunk_video[:, :, :actual_chunk_size, :, :] + + # Store generated chunk + if chunk_idx == 0: + generated_chunks.append(chunk_video) + else: + # Remove overlap frames from the beginning + generated_chunks.append(chunk_video[:, :, chunk_overlap:, :, :]) + + # Update input for next iteration using generated frames + if chunk_idx < num_chunks - 1: + # Convert generated chunk from [-1, 1] to [0, 255] uint8 range + chunk_video_uint8 = ((chunk_video / 2.0 + 0.5).clamp(0.0, 1.0) * 255.0).to(torch.uint8) + # Update the input video with generated frames for conditioning next chunk + update_start = start_frame + chunk_num_conditional + update_end = end_frame + current_input_video[:, :, update_start:update_end, :, :] = chunk_video_uint8[ + :, :, chunk_num_conditional:, :, : + ] + + # Concatenate all chunks along time dimension + final_video = torch.cat(generated_chunks, dim=2) + + log.info(f"Generated final video with shape {final_video.shape}") + return final_video + + def cleanup(self): + """Clean up distributed resources.""" + if self.context_parallel_size > 1: + import torch.distributed as dist + from megatron.core import parallel_state + + if parallel_state.is_initialized(): + parallel_state.destroy_model_parallel() + dist.destroy_process_group() diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/models/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/denoise_prediction.py b/REGEN-main/cosmos_policy/_src/predict2/models/denoise_prediction.py new file mode 100644 index 0000000000000000000000000000000000000000..c35d8bbe1b2fd51920ea3855ed644aa04f8293fc --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/denoise_prediction.py @@ -0,0 +1,27 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from dataclasses import dataclass +from typing import Optional + +import torch + + +@dataclass +class DenoisePrediction: + x0: Optional[torch.Tensor] = None # clean data prediction + F: Optional[torch.Tensor] = None # F prediction in TrigFlow + velocity: Optional[torch.Tensor] = None # velocity prediction if using RF + intermediate_features: Optional[list[torch.Tensor]] = None diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/fm_solvers_unipc.py b/REGEN-main/cosmos_policy/_src/predict2/models/fm_solvers_unipc.py new file mode 100644 index 0000000000000000000000000000000000000000..65f4dabdf108618915b43b177cb4d75083352859 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/fm_solvers_unipc.py @@ -0,0 +1,766 @@ +# Copied from https://github.com/huggingface/diffusers/blob/v0.31.0/src/diffusers/schedulers/scheduling_unipc_multistep.py +# Convert unipc for flow matching +# Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved. + +import math +from typing import List, Optional, Tuple, Union + +import numpy as np +import torch +from diffusers.configuration_utils import ConfigMixin, register_to_config +from diffusers.schedulers.scheduling_utils import KarrasDiffusionSchedulers, SchedulerMixin, SchedulerOutput +from diffusers.utils import deprecate + + +class FlowUniPCMultistepScheduler(SchedulerMixin, ConfigMixin): + """ + `UniPCMultistepScheduler` is a training-free framework designed for the fast sampling of diffusion models. + + This model inherits from [`SchedulerMixin`] and [`ConfigMixin`]. Check the superclass documentation for the generic + methods the library implements for all schedulers such as loading and saving. + + Args: + num_train_timesteps (`int`, defaults to 1000): + The number of diffusion steps to train the model. + solver_order (`int`, default `2`): + The UniPC order which can be any positive integer. The effective order of accuracy is `solver_order + 1` + due to the UniC. It is recommended to use `solver_order=2` for guided sampling, and `solver_order=3` for + unconditional sampling. + prediction_type (`str`, defaults to "flow_prediction"): + Prediction type of the scheduler function; must be `flow_prediction` for this scheduler, which predicts + the flow of the diffusion process. + thresholding (`bool`, defaults to `False`): + Whether to use the "dynamic thresholding" method. This is unsuitable for latent-space diffusion models such + as Stable Diffusion. + dynamic_thresholding_ratio (`float`, defaults to 0.995): + The ratio for the dynamic thresholding method. Valid only when `thresholding=True`. + sample_max_value (`float`, defaults to 1.0): + The threshold value for dynamic thresholding. Valid only when `thresholding=True` and `predict_x0=True`. + predict_x0 (`bool`, defaults to `True`): + Whether to use the updating algorithm on the predicted x0. + solver_type (`str`, default `bh2`): + Solver type for UniPC. It is recommended to use `bh1` for unconditional sampling when steps < 10, and `bh2` + otherwise. + lower_order_final (`bool`, default `True`): + Whether to use lower-order solvers in the final steps. Only valid for < 15 inference steps. This can + stabilize the sampling of DPMSolver for steps < 15, especially for steps <= 10. + disable_corrector (`list`, default `[]`): + Decides which step to disable the corrector to mitigate the misalignment between `epsilon_theta(x_t, c)` + and `epsilon_theta(x_t^c, c)` which can influence convergence for a large guidance scale. Corrector is + usually disabled during the first few steps. + solver_p (`SchedulerMixin`, default `None`): + Any other scheduler that if specified, the algorithm becomes `solver_p + UniC`. + use_karras_sigmas (`bool`, *optional*, defaults to `False`): + Whether to use Karras sigmas for step sizes in the noise schedule during the sampling process. If `True`, + the sigmas are determined according to a sequence of noise levels {σi}. + use_exponential_sigmas (`bool`, *optional*, defaults to `False`): + Whether to use exponential sigmas for step sizes in the noise schedule during the sampling process. + timestep_spacing (`str`, defaults to `"linspace"`): + The way the timesteps should be scaled. Refer to Table 2 of the [Common Diffusion Noise Schedules and + Sample Steps are Flawed](https://huggingface.co/papers/2305.08891) for more information. + steps_offset (`int`, defaults to 0): + An offset added to the inference steps, as required by some model families. + final_sigmas_type (`str`, defaults to `"zero"`): + The final `sigma` value for the noise schedule during the sampling process. If `"sigma_min"`, the final + sigma is the same as the last sigma in the training schedule. If `zero`, the final sigma is set to 0. + """ + + _compatibles = [e.name for e in KarrasDiffusionSchedulers] + order = 1 + + @register_to_config + def __init__( + self, + num_train_timesteps: int = 1000, + solver_order: int = 2, + prediction_type: str = "flow_prediction", + shift: Optional[float] = 1.0, + use_dynamic_shifting=False, + thresholding: bool = False, + dynamic_thresholding_ratio: float = 0.995, + sample_max_value: float = 1.0, + predict_x0: bool = True, + solver_type: str = "bh2", + lower_order_final: bool = True, + disable_corrector: List[int] = [], + solver_p: SchedulerMixin = None, + timestep_spacing: str = "linspace", + steps_offset: int = 0, + final_sigmas_type: Optional[str] = "zero", # "zero", "sigma_min" + ): + if solver_type not in ["bh1", "bh2"]: + if solver_type in ["midpoint", "heun", "logrho"]: + self.register_to_config(solver_type="bh2") + else: + raise NotImplementedError(f"{solver_type} is not implemented for {self.__class__}") + + self.predict_x0 = predict_x0 + # setable values + self.num_inference_steps = None + alphas = np.linspace(1, 1 / num_train_timesteps, num_train_timesteps)[::-1].copy() + sigmas = 1.0 - alphas + sigmas = torch.from_numpy(sigmas).to(dtype=torch.float32) + + if not use_dynamic_shifting: + # when use_dynamic_shifting is True, we apply the timestep shifting on the fly based on the image resolution + sigmas = shift * sigmas / (1 + (shift - 1) * sigmas) # pyright: ignore + + self.sigmas = sigmas + self.timesteps = sigmas * num_train_timesteps + + self.model_outputs = [None] * solver_order + self.timestep_list = [None] * solver_order + self.lower_order_nums = 0 + self.disable_corrector = disable_corrector + self.solver_p = solver_p + self.last_sample = None + self._step_index = None + self._begin_index = None + + self.sigmas = self.sigmas.to("cpu") # to avoid too much CPU/GPU communication + self.sigma_min = self.sigmas[-1].item() + self.sigma_max = self.sigmas[0].item() + + @property + def step_index(self): + """ + The index counter for current timestep. It will increase 1 after each scheduler step. + """ + return self._step_index + + @property + def begin_index(self): + """ + The index for the first timestep. It should be set from pipeline with `set_begin_index` method. + """ + return self._begin_index + + # Copied from diffusers.schedulers.scheduling_dpmsolver_multistep.DPMSolverMultistepScheduler.set_begin_index + def set_begin_index(self, begin_index: int = 0): + """ + Sets the begin index for the scheduler. This function should be run from pipeline before the inference. + + Args: + begin_index (`int`): + The begin index for the scheduler. + """ + self._begin_index = begin_index + + # Modified from diffusers.schedulers.scheduling_flow_match_euler_discrete.FlowMatchEulerDiscreteScheduler.set_timesteps + def set_timesteps( + self, + num_inference_steps: Union[int, None] = None, + device: Union[str, torch.device] = None, + sigmas: Optional[List[float]] = None, + mu: Optional[Union[float, None]] = None, + shift: Optional[Union[float, None]] = None, + use_kerras_sigma: bool = False, + ): + """ + Sets the discrete timesteps used for the diffusion chain (to be run before inference). + Args: + num_inference_steps (`int`): + Total number of the spacing of the time steps. + device (`str` or `torch.device`, *optional*): + The device to which the timesteps should be moved to. If `None`, the timesteps are not moved. + """ + if self.config.use_dynamic_shifting and mu is None: + raise ValueError(" you have to pass a value for `mu` when `use_dynamic_shifting` is set to be `True`") + + if use_kerras_sigma: + # force to use the exact sigma used in edm sampler + sigma_max = 200 + sigma_min = 0.01 + rho = 7 + sigmas = np.arange(num_inference_steps + 1) / num_inference_steps + min_inv_rho = sigma_min ** (1 / rho) + max_inv_rho = sigma_max ** (1 / rho) + sigmas = (max_inv_rho + sigmas * (min_inv_rho - max_inv_rho)) ** rho + sigmas = sigmas / (1 + sigmas) + else: + if sigmas is None: + sigmas = np.linspace(self.sigma_max, self.sigma_min, num_inference_steps + 1).copy()[:-1] # pyright: ignore + + if self.config.use_dynamic_shifting: + sigmas = self.time_shift(mu, 1.0, sigmas) # pyright: ignore + else: + if shift is None: + shift = self.config.shift + sigmas = shift * sigmas / (1 + (shift - 1) * sigmas) # pyright: ignore + + if self.config.final_sigmas_type == "sigma_min": + sigma_last = ((1 - self.alphas_cumprod[0]) / self.alphas_cumprod[0]) ** 0.5 + elif self.config.final_sigmas_type == "zero": + sigma_last = 0 + else: + raise ValueError( + f"`final_sigmas_type` must be one of 'zero', or 'sigma_min', but got {self.config.final_sigmas_type}" + ) + + timesteps = sigmas * self.config.num_train_timesteps + sigmas = np.concatenate([sigmas, [sigma_last]]).astype(np.float32) # pyright: ignore + + self.sigmas = torch.from_numpy(sigmas) + self.timesteps = torch.from_numpy(timesteps).to(device=device, dtype=torch.int64) + + self.num_inference_steps = len(timesteps) + + self.model_outputs = [ + None, + ] * self.config.solver_order + self.lower_order_nums = 0 + self.last_sample = None + if self.solver_p: + self.solver_p.set_timesteps(self.num_inference_steps, device=device) + + # add an index counter for schedulers that allow duplicated timesteps + self._step_index = None + self._begin_index = None + self.sigmas = self.sigmas.to("cpu") # to avoid too much CPU/GPU communication + + # Copied from diffusers.schedulers.scheduling_ddpm.DDPMScheduler._threshold_sample + def _threshold_sample(self, sample: torch.Tensor) -> torch.Tensor: + """ + "Dynamic thresholding: At each sampling step we set s to a certain percentile absolute pixel value in xt0 (the + prediction of x_0 at timestep t), and if s > 1, then we threshold xt0 to the range [-s, s] and then divide by + s. Dynamic thresholding pushes saturated pixels (those near -1 and 1) inwards, thereby actively preventing + pixels from saturation at each step. We find that dynamic thresholding results in significantly better + photorealism as well as better image-text alignment, especially when using very large guidance weights." + + https://arxiv.org/abs/2205.11487 + """ + dtype = sample.dtype + batch_size, channels, *remaining_dims = sample.shape + + if dtype not in (torch.float32, torch.float64): + sample = sample.float() # upcast for quantile calculation, and clamp not implemented for cpu half + + # Flatten sample for doing quantile calculation along each image + sample = sample.reshape(batch_size, channels * np.prod(remaining_dims)) + + abs_sample = sample.abs() # "a certain percentile absolute pixel value" + + s = torch.quantile(abs_sample, self.config.dynamic_thresholding_ratio, dim=1) + s = torch.clamp( + s, min=1, max=self.config.sample_max_value + ) # When clamped to min=1, equivalent to standard clipping to [-1, 1] + s = s.unsqueeze(1) # (batch_size, 1) because clamp will broadcast along dim=0 + sample = torch.clamp(sample, -s, s) / s # "we threshold xt0 to the range [-s, s] and then divide by s" + + sample = sample.reshape(batch_size, channels, *remaining_dims) + sample = sample.to(dtype) + + return sample + + # Copied from diffusers.schedulers.scheduling_flow_match_euler_discrete.FlowMatchEulerDiscreteScheduler._sigma_to_t + def _sigma_to_t(self, sigma): + return sigma * self.config.num_train_timesteps + + def _sigma_to_alpha_sigma_t(self, sigma): + return 1 - sigma, sigma + + # Copied from diffusers.schedulers.scheduling_flow_match_euler_discrete.set_timesteps + def time_shift(self, mu: float, sigma: float, t: torch.Tensor): + return math.exp(mu) / (math.exp(mu) + (1 / t - 1) ** sigma) + + def convert_model_output( + self, + model_output: torch.Tensor, + *args, + sample: torch.Tensor = None, + **kwargs, + ) -> torch.Tensor: + r""" + Convert the model output to the corresponding type the UniPC algorithm needs. + + Args: + model_output (`torch.Tensor`): + The direct output from the learned diffusion model. + timestep (`int`): + The current discrete timestep in the diffusion chain. + sample (`torch.Tensor`): + A current instance of a sample created by the diffusion process. + + Returns: + `torch.Tensor`: + The converted model output. + """ + timestep = args[0] if len(args) > 0 else kwargs.pop("timestep", None) + if sample is None: + if len(args) > 1: + sample = args[1] + else: + raise ValueError("missing `sample` as a required keyward argument") + if timestep is not None: + deprecate( + "timesteps", + "1.0.0", + "Passing `timesteps` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`", + ) + + sigma = self.sigmas[self.step_index] + alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma) + + # print("sigma_t ==>", self.step_index, sigma, sigma_t, alpha_t, sample.shape, model_output.shape) + if self.predict_x0: + if self.config.prediction_type == "flow_prediction": + sigma_t = self.sigmas[self.step_index] + x0_pred = sample - sigma_t * model_output + else: + raise ValueError( + f"prediction_type given as {self.config.prediction_type} must be one of `epsilon`, `sample`," + " `v_prediction` or `flow_prediction` for the UniPCMultistepScheduler." + ) + + if self.config.thresholding: + x0_pred = self._threshold_sample(x0_pred) + # print("self.config.thresholding", self.config.thresholding) + return x0_pred + else: + if self.config.prediction_type == "flow_prediction": + sigma_t = self.sigmas[self.step_index] + epsilon = sample - (1 - sigma_t) * model_output + else: + raise ValueError( + f"prediction_type given as {self.config.prediction_type} must be one of `epsilon`, `sample`," + " `v_prediction` or `flow_prediction` for the UniPCMultistepScheduler." + ) + + if self.config.thresholding: + sigma_t = self.sigmas[self.step_index] + x0_pred = sample - sigma_t * model_output + x0_pred = self._threshold_sample(x0_pred) + epsilon = model_output + x0_pred + + return epsilon + + def multistep_uni_p_bh_update( + self, + model_output: torch.Tensor, + *args, + sample: torch.Tensor = None, + order: int = None, # pyright: ignore + **kwargs, + ) -> torch.Tensor: + """ + One step for the UniP (B(h) version). Alternatively, `self.solver_p` is used if is specified. + + Args: + model_output (`torch.Tensor`): + The direct output from the learned diffusion model at the current timestep. + prev_timestep (`int`): + The previous discrete timestep in the diffusion chain. + sample (`torch.Tensor`): + A current instance of a sample created by the diffusion process. + order (`int`): + The order of UniP at this timestep (corresponds to the *p* in UniPC-p). + + Returns: + `torch.Tensor`: + The sample tensor at the previous timestep. + """ + prev_timestep = args[0] if len(args) > 0 else kwargs.pop("prev_timestep", None) + if sample is None: + if len(args) > 1: + sample = args[1] + else: + raise ValueError(" missing `sample` as a required keyward argument") + if order is None: + if len(args) > 2: + order = args[2] + else: + raise ValueError(" missing `order` as a required keyward argument") + if prev_timestep is not None: + deprecate( + "prev_timestep", + "1.0.0", + "Passing `prev_timestep` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`", + ) + model_output_list = self.model_outputs + + s0 = self.timestep_list[-1] + m0 = model_output_list[-1] + x = sample + + if self.solver_p: + x_t = self.solver_p.step(model_output, s0, x).prev_sample + return x_t + + sigma_t, sigma_s0 = self.sigmas[self.step_index + 1], self.sigmas[self.step_index] # pyright: ignore + alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma_t) + alpha_s0, sigma_s0 = self._sigma_to_alpha_sigma_t(sigma_s0) + + lambda_t = torch.log(alpha_t) - torch.log(sigma_t) + lambda_s0 = torch.log(alpha_s0) - torch.log(sigma_s0) + + h = lambda_t - lambda_s0 + device = sample.device + + rks = [] + D1s = [] + for i in range(1, order): + si = self.step_index - i # pyright: ignore + mi = model_output_list[-(i + 1)] + alpha_si, sigma_si = self._sigma_to_alpha_sigma_t(self.sigmas[si]) + lambda_si = torch.log(alpha_si) - torch.log(sigma_si) + rk = (lambda_si - lambda_s0) / h + rks.append(rk) + D1s.append((mi - m0) / rk) # pyright: ignore + + rks.append(1.0) + rks = torch.tensor(rks, device=device) + + R = [] + b = [] + + hh = -h if self.predict_x0 else h + h_phi_1 = torch.expm1(hh) # h\phi_1(h) = e^h - 1 + h_phi_k = h_phi_1 / hh - 1 + + factorial_i = 1 + + if self.config.solver_type == "bh1": + B_h = hh + elif self.config.solver_type == "bh2": + B_h = torch.expm1(hh) + else: + raise NotImplementedError() + + for i in range(1, order + 1): + R.append(torch.pow(rks, i - 1)) + b.append(h_phi_k * factorial_i / B_h) + factorial_i *= i + 1 + h_phi_k = h_phi_k / hh - 1 / factorial_i + + R = torch.stack(R) + b = torch.tensor(b, device=device) + + if len(D1s) > 0: + D1s = torch.stack(D1s, dim=1) # (B, K) + # for order 2, we use a simplified version + if order == 2: + rhos_p = torch.tensor([0.5], dtype=x.dtype, device=device) + else: + rhos_p = torch.linalg.solve(R[:-1, :-1], b[:-1]).to(device).to(x.dtype) + else: + D1s = None + + if self.predict_x0: + x_t_ = sigma_t / sigma_s0 * x - alpha_t * h_phi_1 * m0 + if D1s is not None: + pred_res = torch.einsum("k,bkc...->bc...", rhos_p, D1s) # pyright: ignore + else: + pred_res = 0 + x_t = x_t_ - alpha_t * B_h * pred_res + else: + x_t_ = alpha_t / alpha_s0 * x - sigma_t * h_phi_1 * m0 + if D1s is not None: + pred_res = torch.einsum("k,bkc...->bc...", rhos_p, D1s) # pyright: ignore + else: + pred_res = 0 + x_t = x_t_ - sigma_t * B_h * pred_res + + x_t = x_t.to(x.dtype) + return x_t + + def multistep_uni_c_bh_update( + self, + this_model_output: torch.Tensor, + *args, + last_sample: torch.Tensor = None, + this_sample: torch.Tensor = None, + order: int = None, # pyright: ignore + **kwargs, + ) -> torch.Tensor: + """ + One step for the UniC (B(h) version). + + Args: + this_model_output (`torch.Tensor`): + The model outputs at `x_t`. + this_timestep (`int`): + The current timestep `t`. + last_sample (`torch.Tensor`): + The generated sample before the last predictor `x_{t-1}`. + this_sample (`torch.Tensor`): + The generated sample after the last predictor `x_{t}`. + order (`int`): + The `p` of UniC-p at this step. The effective order of accuracy should be `order + 1`. + + Returns: + `torch.Tensor`: + The corrected sample tensor at the current timestep. + """ + this_timestep = args[0] if len(args) > 0 else kwargs.pop("this_timestep", None) + if last_sample is None: + if len(args) > 1: + last_sample = args[1] + else: + raise ValueError(" missing`last_sample` as a required keyward argument") + if this_sample is None: + if len(args) > 2: + this_sample = args[2] + else: + raise ValueError(" missing`this_sample` as a required keyward argument") + if order is None: + if len(args) > 3: + order = args[3] + else: + raise ValueError(" missing`order` as a required keyward argument") + if this_timestep is not None: + deprecate( + "this_timestep", + "1.0.0", + "Passing `this_timestep` is deprecated and has no effect as model output conversion is now handled via an internal counter `self.step_index`", + ) + + model_output_list = self.model_outputs + + m0 = model_output_list[-1] + x = last_sample + x_t = this_sample + model_t = this_model_output + + sigma_t, sigma_s0 = self.sigmas[self.step_index], self.sigmas[self.step_index - 1] # pyright: ignore + alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma_t) + alpha_s0, sigma_s0 = self._sigma_to_alpha_sigma_t(sigma_s0) + + lambda_t = torch.log(alpha_t) - torch.log(sigma_t) + lambda_s0 = torch.log(alpha_s0) - torch.log(sigma_s0) + + h = lambda_t - lambda_s0 + device = this_sample.device + + rks = [] + D1s = [] + for i in range(1, order): + si = self.step_index - (i + 1) # pyright: ignore + mi = model_output_list[-(i + 1)] + alpha_si, sigma_si = self._sigma_to_alpha_sigma_t(self.sigmas[si]) + lambda_si = torch.log(alpha_si) - torch.log(sigma_si) + rk = (lambda_si - lambda_s0) / h + rks.append(rk) + D1s.append((mi - m0) / rk) # pyright: ignore + + rks.append(1.0) + rks = torch.tensor(rks, device=device) + + R = [] + b = [] + + hh = -h if self.predict_x0 else h + h_phi_1 = torch.expm1(hh) # h\phi_1(h) = e^h - 1 + h_phi_k = h_phi_1 / hh - 1 + + factorial_i = 1 + + if self.config.solver_type == "bh1": + B_h = hh + elif self.config.solver_type == "bh2": + B_h = torch.expm1(hh) + else: + raise NotImplementedError() + + for i in range(1, order + 1): + R.append(torch.pow(rks, i - 1)) + b.append(h_phi_k * factorial_i / B_h) + factorial_i *= i + 1 + h_phi_k = h_phi_k / hh - 1 / factorial_i + + R = torch.stack(R) + b = torch.tensor(b, device=device) + + if len(D1s) > 0: + D1s = torch.stack(D1s, dim=1) + else: + D1s = None + + # for order 1, we use a simplified version + if order == 1: + rhos_c = torch.tensor([0.5], dtype=x.dtype, device=device) + else: + rhos_c = torch.linalg.solve(R, b).to(device).to(x.dtype) + + if self.predict_x0: + x_t_ = sigma_t / sigma_s0 * x - alpha_t * h_phi_1 * m0 + if D1s is not None: + corr_res = torch.einsum("k,bkc...->bc...", rhos_c[:-1], D1s) + else: + corr_res = 0 + D1_t = model_t - m0 + x_t = x_t_ - alpha_t * B_h * (corr_res + rhos_c[-1] * D1_t) + else: + x_t_ = alpha_t / alpha_s0 * x - sigma_t * h_phi_1 * m0 + if D1s is not None: + corr_res = torch.einsum("k,bkc...->bc...", rhos_c[:-1], D1s) + else: + corr_res = 0 + D1_t = model_t - m0 + x_t = x_t_ - sigma_t * B_h * (corr_res + rhos_c[-1] * D1_t) + x_t = x_t.to(x.dtype) + return x_t + + def index_for_timestep(self, timestep, schedule_timesteps=None): + if schedule_timesteps is None: + schedule_timesteps = self.timesteps + + indices = (schedule_timesteps == timestep).nonzero() + + # The sigma index that is taken for the **very** first `step` + # is always the second index (or the last index if there is only 1) + # This way we can ensure we don't accidentally skip a sigma in + # case we start in the middle of the denoising schedule (e.g. for image-to-image) + pos = 1 if len(indices) > 1 else 0 + + return indices[pos].item() + + # Copied from diffusers.schedulers.scheduling_dpmsolver_multistep.DPMSolverMultistepScheduler._init_step_index + def _init_step_index(self, timestep): + """ + Initialize the step_index counter for the scheduler. + """ + + if self.begin_index is None: + if isinstance(timestep, torch.Tensor): + timestep = timestep.to(self.timesteps.device) + self._step_index = self.index_for_timestep(timestep) + else: + self._step_index = self._begin_index + + def step( + self, + model_output: torch.Tensor, + timestep: Union[int, torch.Tensor], + sample: torch.Tensor, + return_dict: bool = True, + generator=None, + ) -> Union[SchedulerOutput, Tuple]: + """ + Predict the sample from the previous timestep by reversing the SDE. This function propagates the sample with + the multistep UniPC. + + Args: + model_output (`torch.Tensor`): + The direct output from learned diffusion model. + timestep (`int`): + The current discrete timestep in the diffusion chain. + sample (`torch.Tensor`): + A current instance of a sample created by the diffusion process. + return_dict (`bool`): + Whether or not to return a [`~schedulers.scheduling_utils.SchedulerOutput`] or `tuple`. + + Returns: + [`~schedulers.scheduling_utils.SchedulerOutput`] or `tuple`: + If return_dict is `True`, [`~schedulers.scheduling_utils.SchedulerOutput`] is returned, otherwise a + tuple is returned where the first element is the sample tensor. + + """ + if self.num_inference_steps is None: + raise ValueError( + "Number of inference steps is 'None', you need to run 'set_timesteps' after creating the scheduler" + ) + + if self.step_index is None: + self._init_step_index(timestep) + + # print("self.step_index ==> ", self.step_index) + + use_corrector = ( + self.step_index > 0 and self.step_index - 1 not in self.disable_corrector and self.last_sample is not None # pyright: ignore + ) + + model_output_convert = self.convert_model_output(model_output, sample=sample) + + if use_corrector: + sample = self.multistep_uni_c_bh_update( + this_model_output=model_output_convert, + last_sample=self.last_sample, + this_sample=sample, + order=self.this_order, + ) + + for i in range(self.config.solver_order - 1): + self.model_outputs[i] = self.model_outputs[i + 1] + self.timestep_list[i] = self.timestep_list[i + 1] + + self.model_outputs[-1] = model_output_convert + self.timestep_list[-1] = timestep # pyright: ignore + + if self.config.lower_order_final: + this_order = min(self.config.solver_order, len(self.timesteps) - self.step_index) # pyright: ignore + else: + this_order = self.config.solver_order + + self.this_order = min(this_order, self.lower_order_nums + 1) # warmup for multistep + assert self.this_order > 0 + + self.last_sample = sample + prev_sample = self.multistep_uni_p_bh_update( + model_output=model_output, # pass the original non-converted model output, in case solver-p is used + sample=sample, + order=self.this_order, + ) + + if self.lower_order_nums < self.config.solver_order: + self.lower_order_nums += 1 + + # upon completion increase step index by one + self._step_index += 1 # pyright: ignore + + if not return_dict: + return (prev_sample, model_output_convert) + + return SchedulerOutput(prev_sample=prev_sample) + + def scale_model_input(self, sample: torch.Tensor, *args, **kwargs) -> torch.Tensor: + """ + Ensures interchangeability with schedulers that need to scale the denoising model input depending on the + current timestep. + + Args: + sample (`torch.Tensor`): + The input sample. + + Returns: + `torch.Tensor`: + A scaled input sample. + """ + return sample + + # Copied from diffusers.schedulers.scheduling_dpmsolver_multistep.DPMSolverMultistepScheduler.add_noise + def add_noise( + self, + original_samples: torch.Tensor, + noise: torch.Tensor, + timesteps: torch.IntTensor, + ) -> torch.Tensor: + # Make sure sigmas and timesteps have the same device and dtype as original_samples + sigmas = self.sigmas.to(device=original_samples.device, dtype=original_samples.dtype) + if original_samples.device.type == "mps" and torch.is_floating_point(timesteps): + # mps does not support float64 + schedule_timesteps = self.timesteps.to(original_samples.device, dtype=torch.float32) + timesteps = timesteps.to(original_samples.device, dtype=torch.float32) + else: + schedule_timesteps = self.timesteps.to(original_samples.device) + timesteps = timesteps.to(original_samples.device) + + # begin_index is None when the scheduler is used for training or pipeline does not implement set_begin_index + if self.begin_index is None: + step_indices = [self.index_for_timestep(t, schedule_timesteps) for t in timesteps] + elif self.step_index is not None: + # add_noise is called after first denoising step (for inpainting) + step_indices = [self.step_index] * timesteps.shape[0] + else: + # add noise is called before first denoising step to create initial latent(img2img) + step_indices = [self.begin_index] * timesteps.shape[0] + + sigma = sigmas[step_indices].flatten() + while len(sigma.shape) < len(original_samples.shape): + sigma = sigma.unsqueeze(-1) + + alpha_t, sigma_t = self._sigma_to_alpha_sigma_t(sigma) + noisy_samples = alpha_t * original_samples + sigma_t * noise + return noisy_samples + + def __len__(self): + return self.config.num_train_timesteps diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/interpolator_model.py b/REGEN-main/cosmos_policy/_src/predict2/models/interpolator_model.py new file mode 100644 index 0000000000000000000000000000000000000000..fd56af9caff48222ff5e0a40d002d75a1d4bcfa0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/interpolator_model.py @@ -0,0 +1,434 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Callable, Dict, Optional, Tuple + +import attrs +import torch +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor + +from cosmos_policy._src.imaginaire.modules.res_sampler import COMMON_SOLVER_OPTIONS +from cosmos_policy._src.imaginaire.utils import misc +from cosmos_policy._src.imaginaire.utils.context_parallel import cat_outputs_cp, split_inputs_cp +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.configs.frame_interpolation.conditioner import ( + InterpolatorCondition, # type: ignore[missing-import] +) +from cosmos_policy._src.predict2.models.text2world_model_rectified_flow import ( + IS_PREPROCESSED_KEY, + DenoisePrediction, + Text2WorldCondition, +) +from cosmos_policy._src.predict2.models.video2world_model import ( + NUM_CONDITIONAL_FRAMES_KEY, + ConditioningStrategy, + Video2WorldConfig, + Video2WorldModel, +) + + +@attrs.define(slots=False) +class InterpolatorConfig(Video2WorldConfig): + """Configuration for interpolator model with frame interpolation specific settings.""" + + sigma_conditional: float = 0.0001 # Noise level used for conditional frames + frame_wise_encoding: bool = True # Whether to use frame-wise encoding (True) or causal video encoding (False) + interleaved_conditioning: bool = False # Whether to use interleaved conditioning + + def __attrs_post_init__(self): + super().__attrs_post_init__() + + +class InterpolatorModel(Video2WorldModel): + """ + Interpolator model that extends Vid2VidModel with frame interpolation capabilities. + + This model inherits from Vid2VidModel and only overrides the methods that have + been specifically adapted for frame interpolation functionality, while reusing + all other functionality from the parent classes. + """ + + def get_data_and_condition( + self, data_batch: dict[str, torch.Tensor] + ) -> Tuple[torch.Tensor, torch.Tensor, InterpolatorCondition]: + # generate random number of conditional frames for training + raw_state, latent_state, condition = super().get_data_and_condition(data_batch) + condition = condition.set_video_condition( + gt_frames=latent_state.to(**self.tensor_kwargs), + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=data_batch.get(NUM_CONDITIONAL_FRAMES_KEY, None), + interleaved_conditioning=self.config.interleaved_conditioning, + ) + return raw_state, latent_state, condition + + @torch.no_grad() + def encode(self, state: torch.Tensor) -> torch.Tensor: + """ + Encode input video frames to latent space with frame-by-frame processing. + + Args: + state: Input video tensor of shape (B, C, T, H, W) + + Returns: + Encoded latent tensor of shape (B, C, T, H, W) + """ + if not self.config.frame_wise_encoding: + # Use causal video encoding. + return super().encode(state) + + # Use frame-wise encoding. + input_state = rearrange(state, "b c t h w -> (b t) c 1 h w") + encoded_state = [self.tokenizer.encode(one_state.unsqueeze(0)) * self.sigma_data for one_state in input_state] + encoded_state = torch.cat(encoded_state, dim=0) + return rearrange(encoded_state, "(b t) c 1 h w -> b c t h w", b=state.shape[0]) + + @torch.no_grad() + def decode(self, latent: torch.Tensor) -> torch.Tensor: + """ + Decode latent representations back to video frames with frame-by-frame processing. + + Args: + latent: Latent tensor of shape (B, C, T, H, W) + + Returns: + Decoded video tensor of shape (B, C, T, H, W) + """ + if not self.config.frame_wise_encoding: + # Use causal video decoding. + return super().decode(latent) + + # Use frame-wise decoding. + latent_batch = rearrange(latent, "b c t h w -> (b t) c 1 h w") + decoded_batch = [ + self.tokenizer.decode(one_latent.unsqueeze(0) / self.sigma_data) for one_latent in latent_batch + ] + decoded_batch = torch.cat(decoded_batch, dim=0) + return rearrange(decoded_batch, "(b t) c 1 h w -> b c t h w", b=latent.shape[0]) + + def _normalize_video_databatch_inplace(self, data_batch: dict[str, Tensor], input_key: str = None) -> None: + """ + Normalizes video data in-place on a CUDA device to reduce data loading overhead. + + This function modifies the video data tensor within the provided data_batch dictionary + in-place, scaling the uint8 data from the range [0, 255] to the normalized range [-1, 1]. + + Warning: + A warning is issued if the data has not been previously normalized. + + Args: + data_batch (dict[str, Tensor]): A dictionary containing the video data under a specific key. + This tensor is expected to be on a CUDA device and have dtype of torch.uint8. + + Side Effects: + Modifies the 'input_data_key' tensor within the 'data_batch' dictionary in-place. + + Note: + This operation is performed directly on the CUDA device to avoid the overhead associated + with moving data to/from the GPU. Ensure that the tensor is already on the appropriate device + and has the correct dtype (torch.uint8) to avoid unexpected behaviors. + """ + input_key = self.input_data_key if input_key is None else input_key + # only handle video batch + if input_key in data_batch: + # Check if the data has already been normalized and avoid re-normalizing + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert torch.is_floating_point(data_batch[input_key]), "Video data is not in float format." + assert torch.all((data_batch[input_key] >= -1.0001) & (data_batch[input_key] <= 1.0001)), ( + f"Video data is not in the range [-1, 1]. get data range [{data_batch[input_key].min()}, {data_batch[input_key].max()}]" + ) + else: + assert data_batch[input_key].dtype == torch.uint8, "Video data is not in uint8 format." + data_batch[input_key] = data_batch[input_key].to(**self.tensor_kwargs) / 127.5 - 1.0 + data_batch[IS_PREPROCESSED_KEY] = True + + expected_length = self.config.state_t + original_length = data_batch[input_key].shape[2] + assert original_length == expected_length, ( + f"Input video length doesn't match expected length specified by state_t: {original_length} != {expected_length}" + ) + + def denoise( + self, xt_B_C_T_H_W: torch.Tensor, sigma: torch.Tensor, condition: Text2WorldCondition + ) -> DenoisePrediction: + """ + Performs denoising on the input noise data, noise level, and condition with interpolation-specific + noise handling for conditional frames. + + Args: + xt (torch.Tensor): The input noise data. + sigma (torch.Tensor): The noise level. + condition (Text2WorldCondition): conditional information, generated from self.conditioner + + Returns: + DenoisePrediction: The denoised prediction, it includes clean data predicton (x0), \ + noise prediction (eps_pred). + """ + + if sigma.ndim == 1: + sigma_B_T = rearrange(sigma, "b -> b 1") + elif sigma.ndim == 2: + sigma_B_T = sigma + else: + raise ValueError(f"sigma shape {sigma.shape} is not supported") + + sigma_B_1_T_1_1 = rearrange(sigma_B_T, "b t -> b 1 t 1 1") + # get precondition for the network + c_skip_B_1_T_1_1, c_out_B_1_T_1_1, c_in_B_1_T_1_1, c_noise_B_1_T_1_1 = self.scaling(sigma=sigma_B_1_T_1_1) + + net_state_in_B_C_T_H_W = xt_B_C_T_H_W * c_in_B_1_T_1_1 + + if condition.is_video: + condition_state_in_B_C_T_H_W = condition.gt_frames.type_as(net_state_in_B_C_T_H_W) / self.config.sigma_data + if not condition.use_video_condition: + # When using random dropout, we zero out the ground truth frames + condition_state_in_B_C_T_H_W = condition_state_in_B_C_T_H_W * 0 + + _, C, _, _, _ = xt_B_C_T_H_W.shape + condition_video_mask = condition.condition_video_input_mask_B_C_T_H_W.repeat(1, C, 1, 1, 1).type_as( + net_state_in_B_C_T_H_W + ) + + if self.config.conditioning_strategy == str(ConditioningStrategy.FRAME_REPLACE): + # In case of frame replacement strategy, replace the first few frames of the video with the conditional frames + # ADD ACTUAL NOISE to conditional frames to match what we tell the model about noise levels (v1-style fix) + + # Add actual noise to conditional frames. + condition_noise = torch.randn_like(condition_state_in_B_C_T_H_W) * self.config.sigma_conditional + condition_state_in_B_C_T_H_W = condition_state_in_B_C_T_H_W + condition_noise + + # Make the first few frames of x_t be the (now properly noisy) ground truth frames + net_state_in_B_C_T_H_W = ( + condition_state_in_B_C_T_H_W * condition_video_mask + + net_state_in_B_C_T_H_W * (1 - condition_video_mask) + ) + # Adjust c_noise for the conditional frames + sigma_cond_B_1_T_1_1 = torch.ones_like(sigma_B_1_T_1_1) * self.config.sigma_conditional + _, _, _, c_noise_cond_B_1_T_1_1 = self.scaling(sigma=sigma_cond_B_1_T_1_1) + condition_video_mask_B_1_T_1_1 = condition_video_mask.mean(dim=[1, 3, 4], keepdim=True) + c_noise_B_1_T_1_1 = c_noise_cond_B_1_T_1_1 * condition_video_mask_B_1_T_1_1 + c_noise_B_1_T_1_1 * ( + 1 - condition_video_mask_B_1_T_1_1 + ) + elif self.config.conditioning_strategy == str(ConditioningStrategy.CHANNEL_CONCAT): + # In case of channel concatenation strategy, concatenate the conditional frames in the channel dimension + condition_state_in_masked_B_C_T_H_W = condition_state_in_B_C_T_H_W * condition_video_mask + net_state_in_B_C_T_H_W = torch.cat([net_state_in_B_C_T_H_W, condition_state_in_masked_B_C_T_H_W], dim=1) + + else: + # In case of image batch, simply concatenate the 0 frames when channel concat strategy is used + if self.config.conditioning_strategy == str(ConditioningStrategy.CHANNEL_CONCAT): + net_state_in_B_C_T_H_W = torch.cat( + [net_state_in_B_C_T_H_W, torch.zeros_like(net_state_in_B_C_T_H_W)], dim=1 + ) + + # forward pass through the network + net_output_B_C_T_H_W = self.net( + x_B_C_T_H_W=net_state_in_B_C_T_H_W.to( + **self.tensor_kwargs + ), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + timesteps_B_T=c_noise_B_1_T_1_1.squeeze(dim=[1, 3, 4]).to( + **self.tensor_kwargs + ), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + **condition.to_dict(), + ).float() + + x0_pred_B_C_T_H_W = c_skip_B_1_T_1_1 * xt_B_C_T_H_W + c_out_B_1_T_1_1 * net_output_B_C_T_H_W + if condition.is_video and self.config.denoise_replace_gt_frames: + # Set the first few frames to the ground truth frames. This will ensure that the loss is not computed for the first few frames. + x0_pred_B_C_T_H_W = condition.gt_frames.type_as( + x0_pred_B_C_T_H_W + ) * condition_video_mask + x0_pred_B_C_T_H_W * (1 - condition_video_mask) + + # get noise prediction based on sde + eps_pred_B_C_T_H_W = (xt_B_C_T_H_W - x0_pred_B_C_T_H_W) / sigma_B_1_T_1_1 + + return DenoisePrediction(x0_pred_B_C_T_H_W, eps_pred_B_C_T_H_W, None) + + def generate_samples_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + is_negative_prompt: bool = False, + num_steps: int = 35, + solver_option: COMMON_SOLVER_OPTIONS = "2ab", + x_sigma_max: Optional[torch.Tensor] = None, + sigma_max: float | None = None, + ) -> torch.Tensor: + """ + Generate interpolated samples from the batch with interpolation-specific logic. + + Args: + data_batch: Raw data batch from the training data loader + guidance: Guidance weight for classifier-free guidance + seed: Random seed for reproducible generation + state_shape: Shape of the state, defaults to data batch if not provided + n_sample: Number of samples to generate + is_negative_prompt: Whether to use negative prompt in unconditioning + num_steps: Number of diffusion steps + solver_option: Differential equation solver option + x_sigma_max: Initial noise tensor + sigma_max: Maximum sigma value for diffusion + + Returns: + Generated latent samples + """ + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + state_shape = [ + self.config.state_ch, + _T, + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + # Get interpolation-specific x0 function + x0_fn = self.get_x0_fn_from_batch(data_batch, guidance, is_negative_prompt=is_negative_prompt) + + if x_sigma_max is None: + x_sigma_max = ( + misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + * self.sde.sigma_max + ) + + # Handle context parallelism for interpolation + if self.net.is_context_parallel_enabled: + x_sigma_max = split_inputs_cp(x=x_sigma_max, seq_dim=2, cp_group=self.get_context_parallel_group()) + + if sigma_max is None: + sigma_max = self.sde.sigma_max + + # Generate samples using interpolation-aware sampling + samples = self.sampler( + x0_fn, + x_sigma_max, + num_steps=num_steps, + sigma_max=sigma_max, + sigma_min=self.sde.sigma_min, + solver_option=solver_option, + ) + + if self.net.is_context_parallel_enabled: + samples = cat_outputs_cp(samples, seq_dim=2, cp_group=self.get_context_parallel_group()) + + return samples + + def get_x0_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generate x0 function with interpolation-specific conditioning logic. + + This method provides a clean, self-contained implementation that uses + our custom denoise method with proper noise handling for conditional frames. + + Args: + data_batch: Input data batch + guidance: Classifier-free guidance scale + is_negative_prompt: Whether to use negative prompts + + Returns: + Function that generates x0 predictions for interpolation + """ + # Set up conditioning logic + if NUM_CONDITIONAL_FRAMES_KEY in data_batch: + num_conditional_frames = data_batch[NUM_CONDITIONAL_FRAMES_KEY] + else: + num_conditional_frames = 1 + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + is_image_batch = self.is_image_batch(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + _, x0, _ = self.get_data_and_condition(data_batch) + + # Set up both conditions with proper gt_frames for interpolation + condition = condition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + interleaved_conditioning=self.config.interleaved_conditioning, + ) + uncondition = uncondition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + interleaved_conditioning=self.config.interleaved_conditioning, + ) + + # condition = condition.edit_for_inference(is_cfg_conditional=True, num_conditional_frames=num_conditional_frames) + # uncondition = uncondition.edit_for_inference( + # is_cfg_conditional=False, num_conditional_frames=num_conditional_frames + # ) + + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def interpolation_x0_fn(noise_x: torch.Tensor, sigma: torch.Tensor) -> torch.Tensor: + """ + Clean interpolation x0 function using our custom denoise method. + """ + if guidance == -1: + # Unconditional generation - use our denoise method + raw_x0 = self.denoise(noise_x, sigma, uncondition).x0 + else: + # Classifier-free guidance - use our denoise method for both paths + cond_x0 = self.denoise(noise_x, sigma, condition).x0 + uncond_x0 = self.denoise(noise_x, sigma, uncondition).x0 + raw_x0 = cond_x0 + guidance * (cond_x0 - uncond_x0) + + # Apply guided interpolation if masks are provided + if "guided_image" in data_batch: + assert "guided_mask" in data_batch, "guided_mask should be in data_batch if guided_image is present" + guide_image = data_batch["guided_image"] + guide_mask = data_batch["guided_mask"] + raw_x0 = guide_mask * guide_image + (1 - guide_mask) * raw_x0 + + return raw_x0 + + return interpolation_x0_fn diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/interpolator_model_rectified_flow.py b/REGEN-main/cosmos_policy/_src/predict2/models/interpolator_model_rectified_flow.py new file mode 100644 index 0000000000000000000000000000000000000000..0733f11b732d70e92674a9767befeb316148963e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/interpolator_model_rectified_flow.py @@ -0,0 +1,269 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Callable, Dict, Tuple + +import attrs +import torch +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor + +from cosmos_policy._src.imaginaire.utils import misc +from cosmos_policy._src.imaginaire.utils.context_parallel import broadcast_split_tensor, cat_outputs_cp +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.configs.frame_interpolation.conditioner import ( + InterpolatorCondition, # type: ignore[missing-import] +) +from cosmos_policy._src.predict2.models.video2world_model_rectified_flow import ( + NUM_CONDITIONAL_FRAMES_KEY, + Video2WorldModelRectifiedFlow, + Video2WorldModelRectifiedFlowConfig, +) + + +@attrs.define(slots=False) +class InterpolatorModelRectifiedFlowConfig(Video2WorldModelRectifiedFlowConfig): + """Configuration for interpolator model with frame interpolation specific settings for rectified flow.""" + + conditional_frame_timestep: float = 0.0001 # Noise level used for conditional frames + frame_wise_encoding: bool = True # Whether to use frame-wise encoding (True) or causal video encoding (False) + interleaved_conditioning: bool = False # Whether to use interleaved conditioning (every other frame) + + def __attrs_post_init__(self): + super().__attrs_post_init__() + + +class InterpolatorModelRectifiedFlow(Video2WorldModelRectifiedFlow): + """ + Interpolator model that extends Video2WorldModelRectifiedFlow with frame interpolation capabilities. + + This model inherits from Video2WorldModelRectifiedFlow and only overrides the methods that have + been specifically adapted for frame interpolation functionality with rectified flow, + while reusing all other functionality from the parent classes. + """ + + def get_data_and_condition( + self, data_batch: dict[str, torch.Tensor] + ) -> Tuple[Tensor, Tensor, InterpolatorCondition]: + # generate random number of conditional frames for training + raw_state, latent_state, condition = super().get_data_and_condition(data_batch) + condition = condition.set_video_condition( + gt_frames=latent_state.to(**self.tensor_kwargs), + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=data_batch.get(NUM_CONDITIONAL_FRAMES_KEY, None), + interleaved_conditioning=self.config.interleaved_conditioning, + ) + return raw_state, latent_state, condition + + @torch.no_grad() + def encode(self, state: torch.Tensor) -> torch.Tensor: + """ + Encode input video frames to latent space with frame-by-frame processing. + + Args: + state: Input video tensor of shape (B, C, T, H, W) + + Returns: + Encoded latent tensor of shape (B, C, T, H, W) + """ + if not self.config.frame_wise_encoding: + # Use causal video encoding. + return super().encode(state) + + # Use frame-wise encoding. + input_state = rearrange(state, "b c t h w -> (b t) c 1 h w") + encoded_state = [self.tokenizer.encode(one_state.unsqueeze(0)) for one_state in input_state] + encoded_state = torch.cat(encoded_state, dim=0) + return rearrange(encoded_state, "(b t) c 1 h w -> b c t h w", b=state.shape[0]) + + @torch.no_grad() + def decode(self, latent: torch.Tensor) -> torch.Tensor: + """ + Decode latent representations back to video frames with frame-by-frame processing. + + Args: + latent: Latent tensor of shape (B, C, T, H, W) + + Returns: + Decoded video tensor of shape (B, C, T, H, W) + """ + if not self.config.frame_wise_encoding: + # Use causal video decoding. + return super().decode(latent) + + # Use frame-wise decoding. + latent_batch = rearrange(latent, "b c t h w -> (b t) c 1 h w") + decoded_batch = [self.tokenizer.decode(one_latent.unsqueeze(0)) for one_latent in latent_batch] + decoded_batch = torch.cat(decoded_batch, dim=0) + return rearrange(decoded_batch, "(b t) c 1 h w -> b c t h w", b=latent.shape[0]) + + @torch.no_grad() + def generate_samples_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + is_negative_prompt: bool = False, + num_steps: int = 35, + shift: float = 5.0, + **kwargs, + ) -> torch.Tensor: + """ + Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples. + Args: + data_batch (dict): raw data batch draw from the training data loader. + iteration (int): Current iteration number. + guidance (float): guidance weights + seed (int): random seed + state_shape (tuple): shape of the state, default to data batch if not provided + n_sample (int): number of samples to generate + is_negative_prompt (bool): use negative prompt t5 in uncondition if true + num_steps (int): number of steps for the diffusion process + """ + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + state_shape = [ + self.config.state_ch, + _T, + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + noise = misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + + seed_g = torch.Generator(device=self.tensor_kwargs["device"]) + seed_g.manual_seed(seed) + + self.sample_scheduler.set_timesteps( + num_steps, + device=self.tensor_kwargs["device"], + shift=shift, + use_kerras_sigma=self.config.use_kerras_sigma_at_inference, + ) + + timesteps = self.sample_scheduler.timesteps + + velocity_fn = self.get_velocity_fn_from_batch(data_batch, guidance, is_negative_prompt=is_negative_prompt) + if self.net.is_context_parallel_enabled: + noise = broadcast_split_tensor(tensor=noise, seq_dim=2, process_group=self.get_context_parallel_group()) + latents = noise + + for _, t in enumerate(timesteps): + latent_model_input = latents + timestep = [t] + + timestep = torch.stack(timestep) + + velocity_pred = velocity_fn(noise, latent_model_input, timestep.unsqueeze(0)) + temp_x0 = self.sample_scheduler.step( + velocity_pred.unsqueeze(0), t, latents[0].unsqueeze(0), return_dict=False, generator=seed_g + )[0] + latents = temp_x0.squeeze(0) + + if self.net.is_context_parallel_enabled: + latents = cat_outputs_cp(latents, seq_dim=2, cp_group=self.get_context_parallel_group()) + + return latents + + def get_velocity_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `x0_fn` based on the provided data batch and guidance factor. + + This function first processes the input data batch through a conditioning workflow (`conditioner`) to obtain conditioned and unconditioned states. It then defines a nested function `x0_fn` which applies a denoising operation on an input `noise_x` at a given noise level `sigma` using both the conditioned and unconditioned states. + + Args: + - data_batch (Dict): A batch of data used for conditioning. The format and content of this dictionary should align with the expectations of the `self.conditioner` + - guidance (float, optional): A scalar value that modulates the influence of the conditioned state relative to the unconditioned state in the output. Defaults to 1.5. + - is_negative_prompt (bool): use negative prompt t5 in uncondition if true + + Returns: + - Callable: A function `x0_fn(noise_x, sigma)` that takes two arguments, `noise_x` and `sigma`, and return velocity predictoin + + The returned function is suitable for use in scenarios where a denoised state is required based on both conditioned and unconditioned inputs, with an adjustable level of guidance influence. + """ + + if NUM_CONDITIONAL_FRAMES_KEY in data_batch: + num_conditional_frames = data_batch[NUM_CONDITIONAL_FRAMES_KEY] + else: + num_conditional_frames = 1 + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + is_image_batch = self.is_image_batch(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + _, x0, _ = self.get_data_and_condition(data_batch) + # override condition with inference mode; num_conditional_frames used Here! + condition = condition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + interleaved_conditioning=self.config.interleaved_conditioning, + ) + uncondition = uncondition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + interleaved_conditioning=self.config.interleaved_conditioning, + ) + # condition = condition.edit_for_inference(is_cfg_conditional=True, num_conditional_frames=num_conditional_frames) + # uncondition = uncondition.edit_for_inference( + # is_cfg_conditional=False, num_conditional_frames=num_conditional_frames + # ) + + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def velocity_fn(noise: torch.Tensor, noise_x: torch.Tensor, timestep: torch.Tensor) -> torch.Tensor: + uncond_v = self.denoise(noise, noise_x, timestep, uncondition) + if guidance == -1: + return uncond_v + cond_v = self.denoise(noise, noise_x, timestep, condition) + velocity_pred = cond_v + guidance * (cond_v - uncond_v) + return velocity_pred + + return velocity_fn diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/model_fsdp2_test.py b/REGEN-main/cosmos_policy/_src/predict2/models/model_fsdp2_test.py new file mode 100644 index 0000000000000000000000000000000000000000..cafc014d94e6accf35e172b97613212988d0448c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/model_fsdp2_test.py @@ -0,0 +1,142 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import torch +from einops import repeat + +import cosmos_policy._src.imaginaire.utils.distributed +from cosmos_policy._src.imaginaire.utils import misc +from cosmos_policy._src.imaginaire.utils.config_helper import override +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.predict2.configs.text2world.config import make_config +from cosmos_policy._src.predict2.models.text2world_model import DiffusionModel + +""" +torchrun --nproc_per_node=2 -m projects.cosmos.diffusion.v2.models.model_fsdp2_test +""" + + +def image_batch(): + batch_size = 1 + num_frame = 17 + image_batch_size = batch_size * num_frame // 2 + data_batch = { + "dataset_name": "image_data", + "images": torch.randn(batch_size * num_frame // 2, 3, 1024, 1024, dtype=torch.float32), + "t5_text_embeddings": torch.randn(image_batch_size, 512, 1024, dtype=torch.float32), + "t5_text_mask": torch.randint(0, 2, (image_batch_size, 512), dtype=torch.int64), + "fps": torch.randint(16, 32, (image_batch_size,)).float(), + "num_frames": torch.ones(image_batch_size) * 1.0, + "image_size": repeat( + torch.tensor([1024, 1024, 1024, 1024]), + "... -> b ...", + b=image_batch_size, + ), + "padding_mask": repeat( + torch.zeros(size=(1, 1024, 1024)), + "... -> b ...", + b=image_batch_size, + ), + } + return data_batch + + +def video_batch(): + batch_size = 1 + num_frame = 17 + # video batch + data_batch = { + "dataset_name": "video_data", + "video": (torch.randn(batch_size, 3, num_frame, 1024, 1024) * 255).to(dtype=torch.uint8), + "t5_text_embeddings": torch.randn(batch_size, 512, 1024, dtype=torch.float32), + "t5_text_mask": torch.randint(0, 2, (batch_size, 512), dtype=torch.int64), + "fps": torch.randint(16, 32, (batch_size,)).float(), + "num_frames": torch.ones(batch_size) * num_frame, + "image_size": repeat( + torch.tensor([1024, 1024, 1024, 1024]), + "... -> b ...", + b=batch_size, + ), + "padding_mask": repeat( + torch.zeros(size=(1, 1024, 1024)), + "... -> b ...", + b=batch_size, + ), + } + return data_batch + + +def model_init_test(): + cosmos_policy._src.imaginaire.utils.distributed.init() + config = make_config() + config = override(config, ["--", "experiment=error-free_fsdp_mock-data_base-cb"]) + easy_io.set_s3_backend( + backend_args={ + "backend": "s3", + "path_mapping": { + "s3://rundir/": f"s3://{config.checkpoint.save_to_object_store.bucket}/{config.job.path}/", + }, + "s3_credential_path": config.checkpoint.save_to_object_store.credentials, + } + ) + misc.set_random_seed(seed=config.trainer.seed, by_rank=True) + model = DiffusionModel(config.model.config).cuda() + model.on_train_start(torch.preserve_format) + + optim = torch.optim.AdamW(model.parameters(), lr=1e-8) + return model, optim + + +def model_forward_test(): + model, optim = model_init_test() + model.on_train_start(torch.preserve_format) + + rank = torch.distributed.get_rank() + + image_batch_data = image_batch() + video_batch_data = video_batch() + + for k, v in video_batch_data.items(): + _v = v + if isinstance(v, torch.Tensor): + _v = _v.cuda() + if torch.is_floating_point(v): + _v = _v.to(**model.tensor_kwargs) + video_batch_data[k] = _v + + output_batch, loss = model.training_step(video_batch_data, 1) + loss.backward() + optim.step() + print(f"rank {rank} Video loss: {loss.item()}") + model.on_before_zero_grad(None, None, iteration=1) + + for k, v in image_batch_data.items(): + _v = v + if isinstance(v, torch.Tensor): + _v = _v.cuda() + if torch.is_floating_point(v): + _v = _v.to(**model.tensor_kwargs) + image_batch_data[k] = _v + output_batch, loss = model.training_step(image_batch_data, 2) + loss.backward() + model.clip_grad_norm_(1.0) + print(f"rank {rank} Image loss: {loss.item()}") + + with model.ema_scope(): + print(f"rank {rank} ema works!") + + +if __name__ == "__main__": + model_forward_test() diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/text2world_model.py b/REGEN-main/cosmos_policy/_src/predict2/models/text2world_model.py new file mode 100644 index 0000000000000000000000000000000000000000..0bc8b6b5eccf1b555beb05d8d099c61e42cba8c8 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/text2world_model.py @@ -0,0 +1,1128 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import collections +import math +import os +from contextlib import contextmanager +from typing import Any, Callable, Dict, Mapping, Optional, Tuple + +import attrs +import numpy as np +import torch +import tqdm +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor +from torch.distributed._composable.fsdp import FSDPModule, fully_shard +from torch.distributed._tensor.api import DTensor +from torch.distributed.device_mesh import DeviceMesh +from torch.nn.modules.module import _IncompatibleKeys +from torch.nn.utils.clip_grad import clip_grad_norm_ + +from cosmos_policy._src.imaginaire.flags import INTERNAL +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.lazy_config import instantiate as lazy_instantiate +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.modules.denoiser_scaling import EDMScaling, RectifiedFlowScaling +from cosmos_policy._src.imaginaire.modules.edm_sde import EDMSDE +from cosmos_policy._src.imaginaire.modules.res_sampler import COMMON_SOLVER_OPTIONS, Sampler +from cosmos_policy._src.imaginaire.utils import log, misc +from cosmos_policy._src.imaginaire.utils.checkpointer import non_strict_load_model +from cosmos_policy._src.imaginaire.utils.context_parallel import ( + broadcast, + broadcast_split_tensor, + cat_outputs_cp, + find_split, +) +from cosmos_policy._src.imaginaire.utils.count_params import count_params +from cosmos_policy._src.imaginaire.utils.denoise_prediction import DenoisePrediction +from cosmos_policy._src.imaginaire.utils.ema import FastEmaModelUpdater +from cosmos_policy._src.imaginaire.utils.fsdp_helper import hsdp_device_mesh +from cosmos_policy._src.imaginaire.utils.optim_instantiate import get_base_scheduler +from cosmos_policy._src.predict2.conditioner import DataType, Text2WorldCondition +from cosmos_policy._src.predict2.datasets.utils import VIDEO_RES_SIZE_INFO +from cosmos_policy._src.predict2.models.fm_solvers_unipc import FlowUniPCMultistepScheduler +from cosmos_policy._src.predict2.networks.model_weights_stats import WeightTrainingStat +from cosmos_policy._src.predict2.text_encoders.text_encoder import TextEncoder, TextEncoderConfig +from cosmos_policy._src.predict2.tokenizers.base_vae import BaseVAE +from cosmos_policy._src.predict2.utils.dtensor_helper import DTensorFastEmaModelUpdater, broadcast_dtensor_model_states + +IS_PREPROCESSED_KEY = "is_preprocessed" + + +@attrs.define(slots=False) +class EMAConfig: + """ + Config for the EMA. + """ + + enabled: bool = True + rate: float = 0.1 + iteration_shift: int = 0 + + +@attrs.define(slots=False) +class Text2WorldModelConfig: + """ + Config for [DiffusionModel][projects.cosmos.diffusion.v2.models.text2world_model.DiffusionModel]. + """ + + tokenizer: LazyDict = None + conditioner: LazyDict = None + net: LazyDict = None + ema: EMAConfig = EMAConfig() + sde: LazyDict = L(EDMSDE)( + p_mean=0.0, + p_std=1.0, + sigma_max=80, + sigma_min=0.0002, + ) + fsdp_shard_size: int = 1 + sigma_data: float = 0.5 + precision: str = "bfloat16" + input_data_key: str = "video" # key to fetch input data from data_batch + input_image_key: str = "images" # key to fetch input image from data_batch + input_caption_key: str = "ai_caption" # Key used to fetch input captions + loss_reduce: str = "mean" + loss_scale: float = 10.0 + use_torch_compile: bool = False + adjust_video_noise: bool = True # whether or not adjust video noise accroding to the video length + + state_ch: int = 16 # for latent model, ref to the latent channel number + state_t: int = 8 # for latent model, ref to the latent number of frames + resolution: str = "512" + scaling: str = "edm" + rectified_flow_t_scaling_factor: float = 1.0 + rectified_flow_loss_weight_uniform: bool = True + resize_online: bool = False # whether or not resize the video online; usecase: we load a long duration video and resize to fewer frames, simulate low fps video. If true, it use tokenizer and state_t to infer the expected length of the resized video. + text_encoder_class: str = "T5" + text_encoder_config: Optional[TextEncoderConfig] = None + use_lora: bool = False + lora_rank: int = 32 + lora_alpha: int = 32 + lora_target_modules: str = "q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2" + init_lora_weights: bool = True + use_wan_fp32_strategy: bool = False # if True, use WAN FP32 strategy for rectified flow + use_flowunipc_scheduler: bool = False # if True, use FlowUniPCMultistepScheduler for inference. Currently only I2V is supported for this scheduler. + + def __attrs_post_init__(self): + assert self.scaling in ["edm", "rectified_flow"] + assert self.text_encoder_class in ["T5", "umT5", "reason1_2B", "reason1_7B", "reason1p1_7B", "qwen0.5B"] + + +class DiffusionModel(ImaginaireModel): + """ + Diffusion model. + """ + + def __init__(self, config: Text2WorldModelConfig): + super().__init__() + + self.config = config + + self.precision = { + "float32": torch.float32, + "float16": torch.float16, + "bfloat16": torch.bfloat16, + }[config.precision] + self.tensor_kwargs = {"device": "cuda", "dtype": self.precision} + log.warning(f"DiffusionModel: precision {self.precision}") + + # 1. set data keys and data information + self.sigma_data = config.sigma_data + self.setup_data_key() + + # 2. setup up diffusion processing and scaling~(pre-condition), sampler + self.sde = lazy_instantiate(config.sde) + self.sampler = Sampler() + self.scaling = ( + EDMScaling(self.sigma_data) + if config.scaling == "edm" + else RectifiedFlowScaling( + self.sigma_data, config.rectified_flow_t_scaling_factor, config.rectified_flow_loss_weight_uniform + ) + ) + + # 3. tokenizer + with misc.timer("DiffusionModel: set_up_tokenizer"): + self.tokenizer: BaseVAE = lazy_instantiate(config.tokenizer) + assert self.tokenizer.latent_ch == self.config.state_ch, ( + f"latent_ch {self.tokenizer.latent_ch} != state_shape {self.config.state_ch}" + ) + + # 4. Set up loss options, including loss masking, loss reduce and loss scaling + self.loss_reduce = getattr(config, "loss_reduce", "mean") + assert self.loss_reduce in ["mean", "sum"] + self.loss_scale = getattr(config, "loss_scale", 1.0) + log.critical(f"Using {self.loss_reduce} loss reduce with loss scale {self.loss_scale}") + if self.config.adjust_video_noise: + self.video_noise_multiplier = math.sqrt(self.config.state_t) + else: + self.video_noise_multiplier = 1.0 + + # 5. create fsdp mesh if needed + if config.fsdp_shard_size > 1: + self.fsdp_device_mesh = hsdp_device_mesh( + sharding_group_size=config.fsdp_shard_size, + ) + else: + self.fsdp_device_mesh = None + + # 6. diffusion neural networks part + self.set_up_model() + + # 7. text encoder + self.text_encoder = None + if self.config.text_encoder_config is not None and self.config.text_encoder_config.compute_online: + self.text_encoder = TextEncoder(self.config.text_encoder_config) + + # 8. training states + if parallel_state.is_initialized(): + self.data_parallel_size = parallel_state.get_data_parallel_world_size() + else: + self.data_parallel_size = 1 + + def setup_data_key(self) -> None: + self.input_data_key = self.config.input_data_key # by default it is video key for Video diffusion model + self.input_image_key = self.config.input_image_key + self.input_caption_key = self.config.input_caption_key + + def build_net(self): + config = self.config + + init_device = "meta" + with misc.timer("Creating PyTorch model"): + with torch.device(init_device): + net = lazy_instantiate(config.net) + if config.use_lora: + self.add_lora( + net, + lora_rank=config.lora_rank, + lora_alpha=config.lora_alpha, + lora_target_modules=config.lora_target_modules, + init_lora_weights=config.init_lora_weights, + ) + + self._param_count = count_params(net, verbose=False) + + if self.fsdp_device_mesh: + net.fully_shard(mesh=self.fsdp_device_mesh) + net = fully_shard(net, mesh=self.fsdp_device_mesh, reshard_after_forward=True) + + with misc.timer("meta to cuda and broadcast model states"): + net.to_empty(device="cuda") + # IMPORTANT: (qsh) model init should not depends on current tensor shape, or it can handle Dtensor shape. + net.init_weights() + + if self.fsdp_device_mesh: + broadcast_dtensor_model_states(net, self.fsdp_device_mesh) + for name, param in net.named_parameters(): + assert isinstance(param, DTensor), f"param should be DTensor, {name} got {type(param)}" + if int(os.environ.get("COSMOS_PREDICT2_OFFLOAD_DIT", "0")) > 0: + net.cpu() + return net + + @misc.timer("DiffusionModel: set_up_model") + def set_up_model(self): + config = self.config + with misc.timer("Creating PyTorch model and ema if enabled"): + self.conditioner = lazy_instantiate(config.conditioner) + assert sum(p.numel() for p in self.conditioner.parameters() if p.requires_grad) == 0, ( + "conditioner should not have learnable parameters" + ) + self.net = self.build_net() + self._param_count = count_params(self.net, verbose=False) + + if config.ema.enabled: + self.net_ema = self.build_net() + self.net_ema.requires_grad_(False) + + if self.fsdp_device_mesh: + self.net_ema_worker = DTensorFastEmaModelUpdater() + else: + self.net_ema_worker = FastEmaModelUpdater() + + s = config.ema.rate + self.ema_exp_coefficient = np.roots([1, 7, 16 - s**-2, 12 - s**-2]).real.max() + + self.net_ema_worker.copy_to(src_model=self.net, tgt_model=self.net_ema) + torch.cuda.empty_cache() + + def apply_fsdp(self, dp_mesh: DeviceMesh) -> None: + """Apply FSDP to the net and net_ema.""" + # Back-to-back fully_shard calls allow for wrapping submodules and the top-level module. + self.net.fully_shard(mesh=dp_mesh) + self.net = fully_shard(self.net, mesh=dp_mesh, reshard_after_forward=True) + broadcast_dtensor_model_states(self.net, dp_mesh) + if hasattr(self, "net_ema") and self.net_ema: + self.net_ema.fully_shard(mesh=dp_mesh) + self.net_ema = fully_shard(self.net_ema, mesh=dp_mesh, reshard_after_forward=True) + broadcast_dtensor_model_states(self.net_ema, dp_mesh) + self.net_ema_worker = DTensorFastEmaModelUpdater() + # No need to copy weights to EMA when applying FSDP, it is already copied before applying FSDP. + + def init_optimizer_scheduler( + self, optimizer_config: LazyDict, scheduler_config: LazyDict + ) -> tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LRScheduler]: + """Creates the optimizer and scheduler for the model. + + Args: + config_model (ModelConfig): The config object for the model. + + Returns: + optimizer (torch.optim.Optimizer): The model optimizer. + scheduler (torch.optim.lr_scheduler.LRScheduler): The optimization scheduler. + """ + optimizer = lazy_instantiate(optimizer_config, model=self.net) + scheduler = get_base_scheduler(optimizer, self, scheduler_config) + return optimizer, scheduler + + # ------------------------ training hooks ------------------------ + def on_before_zero_grad( + self, optimizer: torch.optim.Optimizer, scheduler: torch.optim.lr_scheduler.LRScheduler, iteration: int + ) -> None: + """ + update the net_ema + """ + del scheduler, optimizer + + if self.config.ema.enabled: + # calculate beta for EMA update + ema_beta = self.ema_beta(iteration) + self.net_ema_worker.update_average(self.net, self.net_ema, beta=ema_beta) + + def on_train_start(self, memory_format: torch.memory_format = torch.preserve_format) -> None: + if self.config.ema.enabled: + self.net_ema.to(dtype=torch.float32) + if hasattr(self.tokenizer, "reset_dtype"): + self.tokenizer.reset_dtype() + self.net = self.net.to(memory_format=memory_format, **self.tensor_kwargs) + + if hasattr(self.config, "use_torch_compile") and self.config.use_torch_compile: # compatible with old config + if torch.__version__ < "2.3": + log.warning( + "torch.compile in Pytorch version older than 2.3 doesn't work well with activation checkpointing.\n" + "It's very likely there will be no significant speedup from torch.compile.\n" + "Please use at least 24.04 Pytorch container, or imaginaire4:v7 container." + ) + # Increasing cache size. It's required because of the model size and dynamic input shapes resulting in + # multiple different triton kernels. For 28 TransformerBlocks, the cache limit of 256 should be enough for + # up to 9 different input shapes, as 28*9 < 256. If you have more Blocks or input shapes, and you observe + # graph breaks at each Block (detectable with torch._dynamo.explain) or warnings about + # exceeding cache limit, you may want to increase this size. + # Starting with 24.05 Pytorch container, the default value is 256 anyway. + # You can read more about it in the comments in Pytorch source code under path torch/_dynamo/cache_size.py. + torch._dynamo.config.accumulated_cache_size_limit = 256 + # dynamic=False means that a separate kernel is created for each shape. It incurs higher compilation costs + # at initial iterations, but can result in more specialized and efficient kernels. + # dynamic=True currently throws errors in pytorch 2.3. + self.net = torch.compile(self.net, dynamic=False, disable=not self.config.use_torch_compile) + + # ------------------------ training ------------------------ + + def training_step( + self, data_batch: dict[str, torch.Tensor], iteration: int + ) -> tuple[dict[str, torch.Tensor], torch.Tensor]: + """ + Performs a single training step for the diffusion model. + + This method is responsible for executing one iteration of the model's training. It involves: + 1. Adding noise to the input data using the SDE process. + 2. Passing the noisy data through the network to generate predictions. + 3. Computing the loss based on the difference between the predictions and the original data, \ + considering any configured loss weighting. + + Args: + data_batch (dict): raw data batch draw from the training data loader. + iteration (int): Current iteration number. + + Returns: + tuple: A tuple containing two elements: + - dict: additional data that used to debug / logging / callbacks + - Tensor: The computed loss for the training step as a PyTorch Tensor. + + Raises: + AssertionError: If the class is conditional, \ + but no number of classes is specified in the network configuration. + + Notes: + - The method handles different types of conditioning + - The method also supports Kendall's loss + """ + self._update_train_stats(data_batch) + + # Obtain text embeddings online + if self.config.text_encoder_config is not None and self.config.text_encoder_config.compute_online: + text_embeddings = self.text_encoder.compute_text_embeddings_online(data_batch, self.input_caption_key) + data_batch["t5_text_embeddings"] = text_embeddings + data_batch["t5_text_mask"] = torch.ones(text_embeddings.shape[0], text_embeddings.shape[1], device="cuda") + + # Get the input data to noise and denoise~(image, video) and the corresponding conditioner. + _, x0_B_C_T_H_W, condition = self.get_data_and_condition(data_batch) + + # Sample pertubation noise levels and N(0, 1) noises + sigma_B_T, epsilon_B_C_T_H_W = self.draw_training_sigma_and_epsilon(x0_B_C_T_H_W.size(), condition) + + # Broadcast and split the input data and condition for model parallelism + x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, sigma_B_T = self.broadcast_split_for_model_parallelsim( + x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, sigma_B_T + ) + output_batch, kendall_loss, _, _ = self.compute_loss_with_epsilon_and_sigma( + x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, sigma_B_T + ) + + if self.loss_reduce == "mean": + kendall_loss = kendall_loss.mean() * self.loss_scale + elif self.loss_reduce == "sum": + kendall_loss = kendall_loss.sum(dim=1).mean() * self.loss_scale + else: + raise ValueError(f"Invalid loss_reduce: {self.loss_reduce}") + + return output_batch, kendall_loss + + @staticmethod + def get_context_parallel_group(): + if parallel_state.is_initialized(): + return parallel_state.get_context_parallel_group() + return None + + def broadcast_split_for_model_parallelsim(self, x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, sigma_B_T): + """ + Broadcast and split the input data and condition for model parallelism. + Currently, we only support context parallelism. + """ + cp_group = self.get_context_parallel_group() + cp_size = 1 if cp_group is None else cp_group.size() + if condition.is_video and cp_size > 1: + use_spatial_split = cp_size > x0_B_C_T_H_W.shape[2] or x0_B_C_T_H_W.shape[2] % cp_size != 0 + after_split_shape = find_split(x0_B_C_T_H_W.shape, cp_size) if use_spatial_split else None + if use_spatial_split: + x0_B_C_T_H_W = rearrange(x0_B_C_T_H_W, "B C T H W -> B C (T H W)") + if epsilon_B_C_T_H_W is not None: + epsilon_B_C_T_H_W = rearrange(epsilon_B_C_T_H_W, "B C T H W -> B C (T H W)") + x0_B_C_T_H_W = broadcast_split_tensor(x0_B_C_T_H_W, seq_dim=2, process_group=cp_group) + epsilon_B_C_T_H_W = broadcast_split_tensor(epsilon_B_C_T_H_W, seq_dim=2, process_group=cp_group) + if use_spatial_split: + x0_B_C_T_H_W = rearrange( + x0_B_C_T_H_W, "B C (T H W) -> B C T H W", T=after_split_shape[0], H=after_split_shape[1] + ) + if epsilon_B_C_T_H_W is not None: + epsilon_B_C_T_H_W = rearrange( + epsilon_B_C_T_H_W, "B C (T H W) -> B C T H W", T=after_split_shape[0], H=after_split_shape[1] + ) + if sigma_B_T is not None: + assert sigma_B_T.ndim == 2, "sigma_B_T should be 2D tensor" + if sigma_B_T.shape[-1] == 1: # single sigma is shared across all frames + sigma_B_T = broadcast(sigma_B_T, cp_group) + else: # different sigma for each frame + sigma_B_T = broadcast_split_tensor(sigma_B_T, seq_dim=1, process_group=cp_group) + if condition is not None: + condition = condition.broadcast(cp_group) + self.net.enable_context_parallel(cp_group) + else: + self.net.disable_context_parallel() + + return x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, sigma_B_T + + def _update_train_stats(self, data_batch: dict[str, torch.Tensor]) -> None: + is_image = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image else self.input_data_key + if isinstance(self.net, WeightTrainingStat): + if is_image: + self.net.accum_image_sample_counter += data_batch[input_key].shape[0] * self.data_parallel_size + else: + self.net.accum_video_sample_counter += data_batch[input_key].shape[0] * self.data_parallel_size + + def draw_training_sigma_and_epsilon(self, x0_size: int, condition: Any) -> torch.Tensor: + batch_size = x0_size[0] + # if use_wan_fp32_strategy, it should be float32. But torch.randn will default to float32 so no need to any change + epsilon = torch.randn(x0_size, device="cuda") + sigma_B = self.sde.sample_t(batch_size).to(device="cuda") + if self.config.use_wan_fp32_strategy: + assert sigma_B.dtype == torch.float32, f"sigma_B dtype is {sigma_B.dtype}, expected float32" + sigma_B_1 = rearrange(sigma_B, "b -> b 1") # add a dimension for T, all frames share the same sigma + is_video_batch = condition.data_type == DataType.VIDEO + + multiplier = self.video_noise_multiplier if is_video_batch else 1 + sigma_B_1 = sigma_B_1 * multiplier + return sigma_B_1, epsilon + + def get_per_sigma_loss_weights(self, sigma: torch.Tensor): + """ + Args: + sigma (tensor): noise level + + Returns: + loss weights per sigma noise level + """ + if "edm" == self.config.scaling: + return (sigma**2 + self.sigma_data**2) / (sigma * self.sigma_data) ** 2 + elif "rectified_flow" == self.config.scaling: + return (1 + sigma) ** 2 / sigma**2 + else: + raise ValueError(f"Invalid scaling: {self.config.scaling}") + + def get_x0_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `x0_fn` based on the provided data batch and guidance factor. + + This function first processes the input data batch through a conditioning workflow (`conditioner`) to obtain conditioned and unconditioned states. It then defines a nested function `x0_fn` which applies a denoising operation on an input `noise_x` at a given noise level `sigma` using both the conditioned and unconditioned states. + + Args: + - data_batch (Dict): A batch of data used for conditioning. The format and content of this dictionary should align with the expectations of the `self.conditioner` + - guidance (float, optional): A scalar value that modulates the influence of the conditioned state relative to the unconditioned state in the output. Defaults to 1.5. + - is_negative_prompt (bool): use negative prompt t5 in uncondition if true + + Returns: + - Callable: A function `x0_fn(noise_x, sigma)` that takes two arguments, `noise_x` and `sigma`, and return x0 predictoin + + The returned function is suitable for use in scenarios where a denoised state is required based on both conditioned and unconditioned inputs, with an adjustable level of guidance influence. + """ + is_image_batch = self.is_image_batch(data_batch) + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(None, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(None, uncondition, None, None) + + # For inference, check if parallel_state is initialized + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def x0_fn(noise_x: torch.Tensor, sigma: torch.Tensor) -> torch.Tensor: + cond_x0 = self.denoise(noise_x, sigma, condition).x0 + uncond_x0 = self.denoise(noise_x, sigma, uncondition).x0 + raw_x0 = cond_x0 + guidance * (cond_x0 - uncond_x0) + if "guided_image" in data_batch: + # replacement trick that enables inpainting with base model + assert "guided_mask" in data_batch, "guided_mask should be in data_batch if guided_image is present" + guide_image = data_batch["guided_image"] + guide_mask = data_batch["guided_mask"] + raw_x0 = guide_mask * guide_image + (1 - guide_mask) * raw_x0 + return raw_x0 + + return x0_fn + + def generate_samples_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + is_negative_prompt: bool = False, + num_steps: int = 35, + solver_option: COMMON_SOLVER_OPTIONS = "2ab", + x_sigma_max: Optional[torch.Tensor] = None, + sigma_max: float | None = None, + **kwargs, + ) -> torch.Tensor: + """ + Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples. + Args: + data_batch (dict): raw data batch draw from the training data loader. + iteration (int): Current iteration number. + guidance (float): guidance weights + seed (int): random seed + state_shape (tuple): shape of the state, default to data batch if not provided + n_sample (int): number of samples to generate + is_negative_prompt (bool): use negative prompt t5 in uncondition if true + num_steps (int): number of steps for the diffusion process + solver_option (str): differential equation solver option, default to "2ab"~(mulitstep solver) + """ + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + state_shape = [ + self.config.state_ch, + self.tokenizer.get_latent_num_frames(_T), + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + x0_fn = self.get_x0_fn_from_batch(data_batch, guidance, is_negative_prompt=is_negative_prompt) + + if self.config.use_flowunipc_scheduler: + sample_scheduler = FlowUniPCMultistepScheduler( + num_train_timesteps=1000, shift=1, use_dynamic_shifting=False + ) + noise = misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + + seed_g = torch.Generator(device=self.tensor_kwargs["device"]) + seed_g.manual_seed(seed) + + sample_scheduler.set_timesteps(num_steps, device=self.tensor_kwargs["device"], shift=5) + + timesteps = sample_scheduler.timesteps + with torch.no_grad(): + x0_fn = self.get_x0_fn_from_batch(data_batch, guidance, is_negative_prompt=is_negative_prompt) + latents = noise + + if self.net.is_context_parallel_enabled: + cp_size = len(torch.distributed.get_process_group_ranks(self.get_context_parallel_group())) + use_spatial_split = cp_size > latents.shape[2] or latents.shape[2] % cp_size != 0 + after_split_shape = find_split(latents.shape, cp_size) if use_spatial_split else None + if use_spatial_split: + latents = rearrange(latents, "b c t h w -> b c (t h w)") + latents = broadcast_split_tensor( + latents, seq_dim=2, process_group=self.get_context_parallel_group() + ) + latents = rearrange( + latents, "b c (t h w) -> b c t h w", t=after_split_shape[0], h=after_split_shape[1] + ) + + if INTERNAL: + timesteps_iter = timesteps + else: + timesteps_iter = tqdm.tqdm(timesteps, desc="Generating samples", total=len(timesteps)) + for _, t in enumerate(timesteps_iter): + latent_model_input = latents + timestep = [t] + + # our model supports 0-1 while the t is 0-1000 + timestep = torch.stack(timestep) / 1000 + noise_pred = x0_fn(latent_model_input, timestep.unsqueeze(0)) + temp_x0 = sample_scheduler.step( + noise_pred.unsqueeze(0), t, latents[0].unsqueeze(0), return_dict=False, generator=seed_g + )[0] + latents = temp_x0.squeeze(0) + + if self.net.is_context_parallel_enabled: + if use_spatial_split: + latents = rearrange(latents, "b c t h w -> b c (t h w)") + latents = cat_outputs_cp(latents, seq_dim=2, cp_group=self.get_context_parallel_group()) + if use_spatial_split: + latents = rearrange(latents, "b c (t h w) -> b c t h w", t=state_shape[1], h=state_shape[2]) + return latents + + if x_sigma_max is None: + x_sigma_max = ( + misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + * self.sde.sigma_max + ) + + use_spatial_split = False + if self.net.is_context_parallel_enabled: + cp_size = len(torch.distributed.get_process_group_ranks(self.get_context_parallel_group())) + use_spatial_split = cp_size > x_sigma_max.shape[2] or x_sigma_max.shape[2] % cp_size != 0 + after_split_shape = None + if use_spatial_split: + if use_spatial_split: + after_split_shape = find_split(x_sigma_max.shape, cp_size) + x_sigma_max = rearrange(x_sigma_max, "b c t h w -> b c (t h w)") + x_sigma_max = broadcast_split_tensor( + x_sigma_max, seq_dim=2, process_group=self.get_context_parallel_group() + ) + if use_spatial_split: + x_sigma_max = rearrange( + x_sigma_max, "b c (t h w) -> b c t h w", t=after_split_shape[0], h=after_split_shape[1] + ) + + if sigma_max is None: + sigma_max = self.sde.sigma_max + samples = self.sampler( + x0_fn, + x_sigma_max, + num_steps=num_steps, + sigma_max=sigma_max, + sigma_min=self.sde.sigma_min, + solver_option=solver_option, + ) + if self.net.is_context_parallel_enabled: + if use_spatial_split: + samples = rearrange(samples, "b c t h w -> b c (t h w)") + samples = cat_outputs_cp(samples, seq_dim=2, cp_group=self.get_context_parallel_group()) + if use_spatial_split: + samples = rearrange(samples, "b c (t h w) -> b c t h w", t=state_shape[1], h=state_shape[2]) + + return samples + + @torch.no_grad() + def validation_step( + self, data: dict[str, torch.Tensor], iteration: int + ) -> tuple[dict[str, torch.Tensor], torch.Tensor]: + """ + Current code does nothing. + """ + raw_data, x0, _ = self.get_data_and_condition(data) + guidance = data["guidance"] + data = misc.to(data, **self.tensor_kwargs) + sample = self.generate_samples_from_batch( + data, + guidance=guidance, + # make sure no mismatch and also works for cp + state_shape=x0.shape[1:], + n_sample=x0.shape[0], + ) + sample = self.decode(sample) + gt = raw_data + caption = data["ai_caption"] + return {"gt": gt, "result": sample, "caption": caption}, torch.tensor([0]).to(**self.tensor_kwargs) + + @torch.no_grad() + def forward(self, xt, t, condition: Text2WorldCondition): + """ + Performs denoising on the input noise data, noise level, and condition + + Args: + xt (torch.Tensor): The input noise data. + sigma (torch.Tensor): The noise level. + condition (Text2WorldCondition): conditional information, generated from self.conditioner + + Returns: + DenoisePrediction: The denoised prediction, it includes clean data predicton (x0), \ + noise prediction (eps_pred). + """ + return self.denoise(xt, t, condition) + + def get_data_and_condition(self, data_batch: dict[str, torch.Tensor]) -> Tuple[Tensor, Tensor, Text2WorldCondition]: + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + + # Latent state + raw_state = data_batch[self.input_image_key if is_image_batch else self.input_data_key] + latent_state = self.encode(raw_state).contiguous().float() + + # Condition + condition = self.conditioner(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + return raw_state, latent_state, condition + + def _normalize_video_databatch_inplace(self, data_batch: dict[str, Tensor], input_key: str = None) -> None: + """ + Normalizes video data in-place on a CUDA device to reduce data loading overhead. + + This function modifies the video data tensor within the provided data_batch dictionary + in-place, scaling the uint8 data from the range [0, 255] to the normalized range [-1, 1]. + + Warning: + A warning is issued if the data has not been previously normalized. + + Args: + data_batch (dict[str, Tensor]): A dictionary containing the video data under a specific key. + This tensor is expected to be on a CUDA device and have dtype of torch.uint8. + + Side Effects: + Modifies the 'input_data_key' tensor within the 'data_batch' dictionary in-place. + + Note: + This operation is performed directly on the CUDA device to avoid the overhead associated + with moving data to/from the GPU. Ensure that the tensor is already on the appropriate device + and has the correct dtype (torch.uint8) to avoid unexpected behaviors. + """ + input_key = self.input_data_key if input_key is None else input_key + # only handle video batch + if input_key in data_batch: + # Check if the data has already been normalized and avoid re-normalizing + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert torch.is_floating_point(data_batch[input_key]), "Video data is not in float format." + assert torch.all((data_batch[input_key] >= -1.0001) & (data_batch[input_key] <= 1.0001)), ( + f"Video data is not in the range [-1, 1]. get data range [{data_batch[input_key].min()}, {data_batch[input_key].max()}]" + ) + else: + assert data_batch[input_key].dtype == torch.uint8, "Video data is not in uint8 format." + data_batch[input_key] = data_batch[input_key].to(**self.tensor_kwargs) / 127.5 - 1.0 + data_batch[IS_PREPROCESSED_KEY] = True + + expected_length = self.tokenizer.get_pixel_num_frames(self.config.state_t) + original_length = data_batch[input_key].shape[2] + assert original_length == expected_length, ( + f"Input video length doesn't match expected length specified by state_t: {original_length} != {expected_length}" + ) + + def _augment_image_dim_inplace(self, data_batch: dict[str, Tensor], input_key: str = None) -> None: + input_key = self.input_image_key if input_key is None else input_key + if input_key in data_batch: + # Check if the data has already been augmented and avoid re-augmenting + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert data_batch[input_key].shape[2] == 1, ( + f"Image data is claimed be augmented while its shape is {data_batch[input_key].shape}" + ) + return + else: + data_batch[input_key] = rearrange(data_batch[input_key], "b c h w -> b c 1 h w").contiguous() + data_batch[IS_PREPROCESSED_KEY] = True + + # ------------------ Checkpointing ------------------ + + def state_dict(self) -> Dict[str, Any]: + net_state_dict = self.net.state_dict(prefix="net.") + if self.config.ema.enabled: + ema_state_dict = self.net_ema.state_dict(prefix="net_ema.") + net_state_dict.update(ema_state_dict) + return net_state_dict + + def load_state_dict(self, state_dict: Mapping[str, Any], strict: bool = True, assign: bool = False): + """ + Loads a state dictionary into the model and optionally its EMA counterpart. + Different from torch strict=False mode, the method will not raise error for unmatched state shape while raise warning. + + Parameters:e + state_dict (Mapping[str, Any]): A dictionary containing separate state dictionaries for the model and + potentially for an EMA version of the model under the keys 'model' and 'ema', respectively. + strict (bool, optional): If True, the method will enforce that the keys in the state dict match exactly + those in the model and EMA model (if applicable). Defaults to True. + assign (bool, optional): If True and in strict mode, will assign the state dictionary directly rather than + matching keys one-by-one. This is typically used when loading parts of state dicts + or using customized loading procedures. Defaults to False. + """ + _reg_state_dict = collections.OrderedDict() + _ema_state_dict = collections.OrderedDict() + for k, v in state_dict.items(): + if k.startswith("net."): + _reg_state_dict[k.replace("net.", "")] = v + elif k.startswith("net_ema."): + _ema_state_dict[k.replace("net_ema.", "")] = v + + state_dict = _reg_state_dict + + if strict: + reg_results: _IncompatibleKeys = self.net.load_state_dict(_reg_state_dict, strict=strict, assign=assign) + + if self.config.ema.enabled: + ema_results: _IncompatibleKeys = self.net_ema.load_state_dict( + _ema_state_dict, strict=strict, assign=assign + ) + + return _IncompatibleKeys( + missing_keys=reg_results.missing_keys + (ema_results.missing_keys if self.config.ema.enabled else []), + unexpected_keys=reg_results.unexpected_keys + + (ema_results.unexpected_keys if self.config.ema.enabled else []), + ) + else: + log.critical("load model in non-strict mode") + log.critical(non_strict_load_model(self.net, _reg_state_dict), rank0_only=False) + if self.config.ema.enabled: + log.critical("load ema model in non-strict mode") + log.critical(non_strict_load_model(self.net_ema, _ema_state_dict), rank0_only=False) + + # ------------------ public methods ------------------ + def ema_beta(self, iteration: int) -> float: + """ + Calculate the beta value for EMA update. + weights = weights * beta + (1 - beta) * new_weights + + Args: + iteration (int): Current iteration number. + + Returns: + float: The calculated beta value. + """ + iteration = iteration + self.config.ema.iteration_shift + if iteration < 1: + return 0.0 + return (1 - 1 / (iteration + 1)) ** (self.ema_exp_coefficient + 1) + + def model_param_stats(self) -> Dict[str, int]: + return {"total_learnable_param_num": self._param_count} + + def is_image_batch(self, data_batch: dict[str, Tensor]) -> bool: + """We hanlde two types of data_batch. One comes from a joint_dataloader where "dataset_name" can be used to differenciate image_batch and video_batch. + Another comes from a dataloader which we by default assumes as video_data for video model training. + """ + is_image = self.input_image_key in data_batch + is_video = self.input_data_key in data_batch + assert is_image != is_video, ( + "Only one of the input_image_key or input_data_key should be present in the data_batch." + ) + return is_image + + def denoise( + self, xt_B_C_T_H_W: torch.Tensor, sigma: torch.Tensor, condition: Text2WorldCondition + ) -> DenoisePrediction: + """ + Performs denoising on the input noise data, noise level, and condition + + Args: + xt (torch.Tensor): The input noise data. + sigma (torch.Tensor): The noise level. + condition (Text2WorldCondition): conditional information, generated from self.conditioner + + Returns: + DenoisePrediction: The denoised prediction, it includes clean data predicton (x0), \ + noise prediction (eps_pred). + """ + if sigma.ndim == 1: + sigma_B_T = rearrange(sigma, "b -> b 1") + elif sigma.ndim == 2: + sigma_B_T = sigma + else: + raise ValueError(f"sigma shape {sigma.shape} is not supported") + sigma_B_1_T_1_1 = rearrange(sigma_B_T, "b t -> b 1 t 1 1") + # get precondition for the network + c_skip_B_1_T_1_1, c_out_B_1_T_1_1, c_in_B_1_T_1_1, c_noise_B_1_T_1_1 = self.scaling(sigma=sigma_B_1_T_1_1) + + # forward pass through the network + net_output_B_C_T_H_W = self.net( + x_B_C_T_H_W=(xt_B_C_T_H_W * c_in_B_1_T_1_1).to( + **self.tensor_kwargs + ), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + timesteps_B_T=c_noise_B_1_T_1_1.squeeze(dim=[1, 3, 4]).to( + **self.tensor_kwargs + ), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + **condition.to_dict(), + ).float() + + x0_pred_B_C_T_H_W = c_skip_B_1_T_1_1 * xt_B_C_T_H_W + c_out_B_1_T_1_1 * net_output_B_C_T_H_W + + # get noise prediction based on sde + eps_pred_B_C_T_H_W = (xt_B_C_T_H_W - x0_pred_B_C_T_H_W) / sigma_B_1_T_1_1 + + return DenoisePrediction(x0_pred_B_C_T_H_W, eps_pred_B_C_T_H_W, None) + + def compute_loss_with_epsilon_and_sigma( + self, + x0_B_C_T_H_W: torch.Tensor, + condition: Text2WorldCondition, + epsilon_B_C_T_H_W: torch.Tensor, + sigma_B_T: torch.Tensor, + ): + """ + Compute loss givee epsilon and sigma + + This method is responsible for computing loss give epsilon and sigma. It involves: + 1. Adding noise to the input data using the SDE process. + 2. Passing the noisy data through the network to generate predictions. + 3. Computing the loss based on the difference between the predictions and the original data, \ + considering any configured loss weighting. + + Args: + data_batch (dict): raw data batch draw from the training data loader. + x0: image/video latent + condition: text condition + epsilon: noise + sigma: noise level + + Returns: + tuple: A tuple containing four elements: + - dict: additional data that used to debug / logging / callbacks + - Tensor 1: kendall loss, + - Tensor 2: MSE loss, + - Tensor 3: EDM loss + + Raises: + AssertionError: If the class is conditional, \ + but no number of classes is specified in the network configuration. + + Notes: + - The method handles different types of conditioning + - The method also supports Kendall's loss + """ + # Get the mean and stand deviation of the marginal probability distribution. + mean_B_C_T_H_W, std_B_T = self.sde.marginal_prob(x0_B_C_T_H_W, sigma_B_T) + # Generate noisy observations + xt_B_C_T_H_W = mean_B_C_T_H_W + epsilon_B_C_T_H_W * rearrange(std_B_T, "b t -> b 1 t 1 1") + # make prediction + model_pred = self.denoise(xt_B_C_T_H_W, sigma_B_T, condition) + # loss weights for different noise levels + weights_per_sigma_B_T = self.get_per_sigma_loss_weights(sigma=sigma_B_T) + # extra loss mask for each sample, for example, human faces, hands + pred_mse_B_C_T_H_W = (x0_B_C_T_H_W - model_pred.x0) ** 2 + edm_loss_B_C_T_H_W = pred_mse_B_C_T_H_W * rearrange(weights_per_sigma_B_T, "b t -> b 1 t 1 1") + + kendall_loss = edm_loss_B_C_T_H_W + output_batch = { + "x0": x0_B_C_T_H_W, + "xt": xt_B_C_T_H_W, + "sigma": sigma_B_T, + "weights_per_sigma": weights_per_sigma_B_T, + "condition": condition, + "model_pred": model_pred, + "mse_loss": pred_mse_B_C_T_H_W.mean(), + "edm_loss": edm_loss_B_C_T_H_W.mean(), + "edm_loss_per_frame": torch.mean(edm_loss_B_C_T_H_W, dim=[1, 3, 4]), + } + return output_batch, kendall_loss, pred_mse_B_C_T_H_W, edm_loss_B_C_T_H_W + + @torch.no_grad() + def encode(self, state: torch.Tensor) -> torch.Tensor: + return self.tokenizer.encode(state) * self.sigma_data + + @torch.no_grad() + def decode(self, latent: torch.Tensor) -> torch.Tensor: + return self.tokenizer.decode(latent / self.sigma_data) + + def get_video_height_width(self) -> Tuple[int, int]: + return VIDEO_RES_SIZE_INFO[self.config.resolution]["9,16"] + + def get_video_latent_height_width(self) -> Tuple[int, int]: + height, width = VIDEO_RES_SIZE_INFO[self.config.resolution]["9,16"] + return height // self.tokenizer.spatial_compression_factor, width // self.tokenizer.spatial_compression_factor + + def get_num_video_latent_frames(self) -> int: + return self.config.state_t + + @property + def text_encoder_class(self) -> str: + return self.config.text_encoder_class + + @contextmanager + def ema_scope(self, context=None, is_cpu=False): + if self.config.ema.enabled: + # https://github.com/pytorch/pytorch/issues/144289 + for module in self.net.modules(): + if isinstance(module, FSDPModule): + module.reshard() + self.net_ema_worker.cache(self.net.parameters(), is_cpu=is_cpu) + self.net_ema_worker.copy_to(src_model=self.net_ema, tgt_model=self.net) + if context is not None: + log.info(f"{context}: Switched to EMA weights") + try: + yield None + finally: + if self.config.ema.enabled: + for module in self.net.modules(): + if isinstance(module, FSDPModule): + module.reshard() + self.net_ema_worker.restore(self.net.parameters()) + if context is not None: + log.info(f"{context}: Restored training weights") + + def clip_grad_norm_( + self, + max_norm: float, + norm_type: float = 2.0, + error_if_nonfinite: bool = False, + foreach: Optional[bool] = None, + ): + return clip_grad_norm_( + self.net.parameters(), + max_norm, + norm_type=norm_type, + error_if_nonfinite=error_if_nonfinite, + foreach=foreach, + ) + + def add_lora( + self, + network: torch.nn.Module, + lora_rank: int = 4, + lora_alpha: int = 4, + lora_target_modules: str = "q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2", + init_lora_weights: bool = True, + ) -> None: + """Add LoRA (Low-Rank Adaptation) adapters to `self.net`. + + This function injects LoRA adapters into specified modules of the network, + enabling parameter-efficient fine-tuning by training only a small number + of additional parameters. + + Args: + lora_rank: The rank of the LoRA adaptation matrices. Higher rank allows + more expressiveness but uses more parameters (default: 4) + lora_alpha: Scaling parameter for LoRA. Controls the magnitude of the + LoRA adaptation (default: 4) + lora_target_modules: Comma-separated string of module names to target + for LoRA adaptation (default: attention and MLP layers) + init_lora_weights: Whether to initialize LoRA weights properly (default: True) + + Raises: + ImportError: If PEFT library is not installed + ValueError: If invalid parameters are provided + RuntimeError: If LoRA injection fails + """ + assert network is not None, "Network is not initialized" + try: + from peft import LoraConfig, inject_adapter_in_model + except ImportError as e: + raise ImportError( + "PEFT library is required for LoRA training. Please install it with: pip install peft" + ) from e + + # Validate parameters + if lora_rank <= 0: + raise ValueError(f"LoRA rank must be positive, got {lora_rank}") + if lora_alpha <= 0: + raise ValueError(f"LoRA alpha must be positive, got {lora_alpha}") + + target_modules_list = [module.strip() for module in lora_target_modules.split(",")] + if not target_modules_list: + raise ValueError("LoRA target_modules cannot be empty") + + # Validate target modules exist in model + model_module_names = set(name for name, _ in network.named_modules()) + invalid_modules = [] + for target_module in target_modules_list: + # Check if any module contains this target pattern + if not any(target_module in module_name for module_name in model_module_names): + invalid_modules.append(target_module) + + if invalid_modules: + log.warning(f"Target modules not found in model: {invalid_modules}") + + # Add LoRA to model + self.lora_alpha = lora_alpha + + log.info(f"Adding LoRA adapters: rank={lora_rank}, alpha={lora_alpha}, targets={target_modules_list}") + + lora_config = LoraConfig( + r=lora_rank, + lora_alpha=lora_alpha, + init_lora_weights=init_lora_weights, + target_modules=target_modules_list, + ) + + try: + network = inject_adapter_in_model(lora_config, network) + except Exception as e: + raise RuntimeError(f"Failed to inject LoRA adapters into model: {e}") from e + + # Count and log LoRA parameters + lora_params = 0 + total_params = 0 + for name, param in network.named_parameters(): + total_params += param.numel() + if param.requires_grad: + lora_params += param.numel() + # Upcast LoRA parameters into fp32 + param.data = param.to(torch.float32) + + log.info( + f"LoRA injection successful: {lora_params:,} trainable parameters out of {total_params:,} total ({100 * lora_params / total_params:.3f}%)" + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/text2world_model_rectified_flow.py b/REGEN-main/cosmos_policy/_src/predict2/models/text2world_model_rectified_flow.py new file mode 100644 index 0000000000000000000000000000000000000000..9f72f3136d244ba4eba0333c394bb7d4e3b3e48e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/text2world_model_rectified_flow.py @@ -0,0 +1,1237 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import collections +import math +import os +from contextlib import contextmanager +from typing import Callable, Dict, Mapping, Optional, Tuple + +import attrs +import numpy as np +import torch +import tqdm +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor +from torch.distributed._composable.fsdp import FSDPModule, fully_shard +from torch.distributed._tensor.api import DTensor +from torch.distributed.device_mesh import DeviceMesh +from torch.nn.modules.module import _IncompatibleKeys +from torch.nn.utils.clip_grad import clip_grad_norm_ + +from cosmos_policy._src.imaginaire.flags import INTERNAL +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.lazy_config import instantiate as lazy_instantiate +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import log, misc +from cosmos_policy._src.imaginaire.utils.checkpointer import non_strict_load_model +from cosmos_policy._src.imaginaire.utils.context_parallel import ( + broadcast, + broadcast_split_tensor, + cat_outputs_cp, + find_split, +) +from cosmos_policy._src.imaginaire.utils.count_params import count_params +from cosmos_policy._src.imaginaire.utils.denoise_prediction import DenoisePrediction +from cosmos_policy._src.imaginaire.utils.ema import FastEmaModelUpdater +from cosmos_policy._src.imaginaire.utils.fsdp_helper import hsdp_device_mesh +from cosmos_policy._src.imaginaire.utils.optim_instantiate import get_base_scheduler +from cosmos_policy._src.predict2.conditioner import DataType, Text2WorldCondition +from cosmos_policy._src.predict2.datasets.utils import VIDEO_RES_SIZE_INFO +from cosmos_policy._src.predict2.models.fm_solvers_unipc import FlowUniPCMultistepScheduler +from cosmos_policy._src.predict2.models.text2world_model import EMAConfig +from cosmos_policy._src.predict2.modules.denoiser_scaling import ( + EDM_sCMWrapper, + RectifiedFlow_sCMWrapper, +) +from cosmos_policy._src.predict2.networks.model_weights_stats import WeightTrainingStat +from cosmos_policy._src.predict2.schedulers.rectified_flow import RectifiedFlow +from cosmos_policy._src.predict2.text_encoders.text_encoder import TextEncoder, TextEncoderConfig +from cosmos_policy._src.predict2.tokenizers.base_vae import BaseVAE +from cosmos_policy._src.predict2.utils.dtensor_helper import DTensorFastEmaModelUpdater, broadcast_dtensor_model_states + +IS_PREPROCESSED_KEY = "is_preprocessed" + + +@attrs.define(slots=False) +class Text2WorldModelRectifiedFlowConfig: + """ + Config for [DiffusionModel][projects.cosmos.diffusion.v2.models.text2world_model.DiffusionModel]. + """ + + tokenizer: LazyDict = None + conditioner: LazyDict = None + net: LazyDict = None + ema: EMAConfig = EMAConfig() + + fsdp_shard_size: int = 1 + precision: str = "bfloat16" + input_data_key: str = "video" # key to fetch input data from data_batch + input_image_key: str = "images" # key to fetch input image from data_batch + input_caption_key: str = "ai_caption" # Key used to fetch input captions + use_torch_compile: bool = False + + state_ch: int = 16 # for latent model, ref to the latent channel number + state_t: int = 8 # for latent model, ref to the latent number of frames + resolution: str = "512" + + text_encoder_class: str = "T5" + text_encoder_config: Optional[TextEncoderConfig] = None + use_lora: bool = False + lora_rank: int = 32 + lora_alpha: int = 32 + use_dora: bool = False + lora_target_modules: str = "q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2" + init_lora_weights: bool = True + + shift: int = 5 + use_dynamic_shift: bool = False + train_time_distribution: str = "logitnormal" + train_time_weight: str = "uniform" + + use_high_sigma_strategy: bool = False # Whether to use high sigma strategy + high_sigma_ratio: float = 0.05 # Ratio of high sigma frames + high_sigma_timesteps_min: int = 980 + high_sigma_timesteps_max: int = 1000 + + use_kerras_sigma_at_inference: bool = False # if True, override unipc's timestep schedule with kerras schedule + + def __attrs_post_init__(self): + assert self.text_encoder_class in ["T5", "umT5", "reason1_2B", "reason1_7B", "reason1p1_7B"] + + +class Text2WorldModelRectifiedFlow(ImaginaireModel): + """ + Diffusion model. + """ + + def __init__(self, config: Text2WorldModelRectifiedFlowConfig): + super().__init__() + + self.config = config + + self.precision = { + "float32": torch.float32, + "float16": torch.float16, + "bfloat16": torch.bfloat16, + }[config.precision] + self.tensor_kwargs = {"device": "cuda", "dtype": self.precision} + self.tensor_kwargs_fp32 = {"device": "cuda", "dtype": torch.float32} + log.warning(f"DiffusionModel: precision {self.precision}") + + # 1. set data keys and data information + scaling = "rectified_flow" + self.sigma_data = 1.0 + self.sigma_conditional = 0.0001 + self.change_time_embed = False + self.scaling_from_time = ( + EDM_sCMWrapper(self.sigma_data) if scaling == "edm" else RectifiedFlow_sCMWrapper(self.sigma_data) + ) + self.setup_data_key() + + # 2. setup up rectified_flow and sampler + self.sample_scheduler = FlowUniPCMultistepScheduler( + num_train_timesteps=1000, shift=1, use_dynamic_shifting=False + ) + + # 3. tokenizer + with misc.timer("DiffusionModel: set_up_tokenizer"): + self.tokenizer: BaseVAE = lazy_instantiate(config.tokenizer) + assert self.tokenizer.latent_ch == self.config.state_ch, ( + f"latent_ch {self.tokenizer.latent_ch} != state_shape {self.config.state_ch}" + ) + + # 4. create fsdp mesh if needed + if config.fsdp_shard_size > 1: + self.fsdp_device_mesh = hsdp_device_mesh( + sharding_group_size=config.fsdp_shard_size, + ) + else: + self.fsdp_device_mesh = None + + # 5. diffusion neural networks part + self.set_up_model() + + # 6. text encoder + self.text_encoder = None + if self.config.text_encoder_config is not None and self.config.text_encoder_config.compute_online: + self.text_encoder = TextEncoder(self.config.text_encoder_config) + + # 7. training states + if parallel_state.is_initialized(): + self.data_parallel_size = parallel_state.get_data_parallel_world_size() + else: + self.data_parallel_size = 1 + + self.rectified_flow = RectifiedFlow( + velocity_field=self.net, + train_time_distribution=config.train_time_distribution, + use_dynamic_shift=config.use_dynamic_shift, + shift=config.shift, + train_time_weight_method=config.train_time_weight, + device=torch.device("cuda"), + dtype=self.tensor_kwargs_fp32["dtype"], + ) + + def setup_data_key(self) -> None: + self.input_data_key = self.config.input_data_key # by default it is video key for Video diffusion model + self.input_image_key = self.config.input_image_key + self.input_caption_key = self.config.input_caption_key + + def build_net(self, keep_on_cpu: bool = False): + config = self.config + + init_device = "meta" + with misc.timer("Creating PyTorch model"): + with torch.device(init_device): + net = lazy_instantiate(config.net) + + self._param_count = count_params(net, verbose=False) + + if keep_on_cpu: + # Move to CPU instead of CUDA to save GPU memory during checkpoint loading + with misc.timer("meta to cpu for deferred GPU materialization"): + net.to_empty(device="cpu") + net.init_weights() + + # Add LoRA after base model init to ensure A~N(0,·), B=0) initialization + if config.use_lora: + net = self.add_lora( + net, + lora_rank=config.lora_rank, + lora_alpha=config.lora_alpha, + lora_target_modules=config.lora_target_modules, + init_lora_weights=config.init_lora_weights, + use_dora=config.use_dora, + ) + else: + with misc.timer("meta to cuda and broadcast model states"): + net.to_empty(device="cuda") + # IMPORTANT: (qsh) model init should not depends on current tensor shape, or it can handle Dtensor shape. + net.init_weights() + + # Add LoRA after base model init to ensure A~N(0,·), B=0) initialization + if config.use_lora: + net = self.add_lora( + net, + lora_rank=config.lora_rank, + lora_alpha=config.lora_alpha, + lora_target_modules=config.lora_target_modules, + init_lora_weights=config.init_lora_weights, + use_dora=config.use_dora, + ) + + if self.fsdp_device_mesh: + net.fully_shard(mesh=self.fsdp_device_mesh) + net = fully_shard(net, mesh=self.fsdp_device_mesh, reshard_after_forward=True) + + broadcast_dtensor_model_states(net, self.fsdp_device_mesh) + for name, param in net.named_parameters(): + assert isinstance(param, DTensor), f"param should be DTensor, {name} got {type(param)}" + if int(os.environ.get("COSMOS_PREDICT2_OFFLOAD_DIT", "0")) > 0: + net.cpu() + return net + + @misc.timer("DiffusionModel: set_up_model") + def set_up_model(self): + config = self.config + with misc.timer("Creating PyTorch model and ema if enabled"): + self.conditioner = lazy_instantiate(config.conditioner) + assert sum(p.numel() for p in self.conditioner.parameters() if p.requires_grad) == 0, ( + "conditioner should not have learnable parameters" + ) + self.net = self.build_net() + self._param_count = count_params(self.net, verbose=False) + + if config.ema.enabled: + # Keep EMA on CPU initially to avoid OOM during model loading from consolidated checkpoint + # It will be moved to GPU and properly initialized in apply_fsdp() + keep_on_cpu = config.fsdp_shard_size == 1 # Keep on CPU if FSDP will be applied later + self.net_ema = self.build_net(keep_on_cpu=keep_on_cpu) + self.net_ema.requires_grad_(False) + + if self.fsdp_device_mesh: + self.net_ema_worker = DTensorFastEmaModelUpdater() + else: + self.net_ema_worker = FastEmaModelUpdater() + + s = config.ema.rate + self.ema_exp_coefficient = np.roots([1, 7, 16 - s**-2, 12 - s**-2]).real.max() + + # Only copy if both models are on the same device (not CPU) + if not keep_on_cpu: + self.net_ema_worker.copy_to(src_model=self.net, tgt_model=self.net_ema) + torch.cuda.empty_cache() + + def apply_fsdp(self, dp_mesh: DeviceMesh) -> None: + """Apply FSDP to the net and net_ema.""" + # Back-to-back fully_shard calls allow for wrapping submodules and the top-level module. + self.net.fully_shard(mesh=dp_mesh) + self.net = fully_shard(self.net, mesh=dp_mesh, reshard_after_forward=True) + broadcast_dtensor_model_states(self.net, dp_mesh) + if hasattr(self, "net_ema") and self.net_ema: + # If net_ema is on CPU, move it to CUDA first + if next(self.net_ema.parameters()).device.type == "cpu": + with misc.timer("Moving EMA model from CPU to CUDA"): + self.net_ema.to(device="cuda") + + self.net_ema.fully_shard(mesh=dp_mesh) + self.net_ema = fully_shard(self.net_ema, mesh=dp_mesh, reshard_after_forward=True) + broadcast_dtensor_model_states(self.net_ema, dp_mesh) + self.net_ema_worker = DTensorFastEmaModelUpdater() + # Copy weights from net to net_ema after both are properly initialized + self.net_ema_worker.copy_to(src_model=self.net, tgt_model=self.net_ema) + + def init_optimizer_scheduler( + self, optimizer_config: LazyDict, scheduler_config: LazyDict + ) -> tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LRScheduler]: + """Creates the optimizer and scheduler for the model. + + Args: + config_model (ModelConfig): The config object for the model. + + Returns: + optimizer (torch.optim.Optimizer): The model optimizer. + scheduler (torch.optim.lr_scheduler.LRScheduler): The optimization scheduler. + """ + optimizer = lazy_instantiate(optimizer_config, model=self.net) + scheduler = get_base_scheduler(optimizer, self, scheduler_config) + return optimizer, scheduler + + # ------------------------ training hooks ------------------------ + def on_before_zero_grad( + self, optimizer: torch.optim.Optimizer, scheduler: torch.optim.lr_scheduler.LRScheduler, iteration: int + ) -> None: + """ + update the net_ema + """ + del scheduler, optimizer + + if self.config.ema.enabled: + # calculate beta for EMA update + ema_beta = self.ema_beta(iteration) + self.net_ema_worker.update_average(self.net, self.net_ema, beta=ema_beta) + + def on_train_start(self, memory_format: torch.memory_format = torch.preserve_format) -> None: + if self.config.ema.enabled: + self.net_ema.to(dtype=torch.float32) + if hasattr(self.tokenizer, "reset_dtype"): + self.tokenizer.reset_dtype() + self.net = self.net.to(memory_format=memory_format, **self.tensor_kwargs) + + if hasattr(self.config, "use_torch_compile") and self.config.use_torch_compile: # compatible with old config + if torch.__version__ < "2.3": + log.warning( + "torch.compile in Pytorch version older than 2.3 doesn't work well with activation checkpointing.\n" + "It's very likely there will be no significant speedup from torch.compile.\n" + "Please use at least 24.04 Pytorch container, or imaginaire4:v7 container." + ) + # Increasing cache size. It's required because of the model size and dynamic input shapes resulting in + # multiple different triton kernels. For 28 TransformerBlocks, the cache limit of 256 should be enough for + # up to 9 different input shapes, as 28*9 < 256. If you have more Blocks or input shapes, and you observe + # graph breaks at each Block (detectable with torch._dynamo.explain) or warnings about + # exceeding cache limit, you may want to increase this size. + # Starting with 24.05 Pytorch container, the default value is 256 anyway. + # You can read more about it in the comments in Pytorch source code under path torch/_dynamo/cache_size.py. + torch._dynamo.config.accumulated_cache_size_limit = 256 + # dynamic=False means that a separate kernel is created for each shape. It incurs higher compilation costs + # at initial iterations, but can result in more specialized and efficient kernels. + # dynamic=True currently throws errors in pytorch 2.3. + self.net = torch.compile(self.net, dynamic=False, disable=not self.config.use_torch_compile) + + # ------------------------ training ------------------------ + + def training_step( + self, data_batch: dict[str, torch.Tensor], iteration: int + ) -> tuple[dict[str, torch.Tensor], torch.Tensor]: + """ + Performs a single training step for the diffusion model. + + This method is responsible for executing one iteration of the model's training. It involves: + 1. Adding noise to the input data using the SDE process. + 2. Passing the noisy data through the network to generate predictions. + 3. Computing the loss based on the difference between the predictions and the original data, \ + considering any configured loss weighting. + + Args: + data_batch (dict): raw data batch draw from the training data loader. + iteration (int): Current iteration number. + + Returns: + tuple: A tuple containing two elements: + - dict: additional data that used to debug / logging / callbacks + - Tensor: The computed loss for the training step as a PyTorch Tensor. + + Raises: + AssertionError: If the class is conditional, \ + but no number of classes is specified in the network configuration. + + Notes: + - The method handles different types of conditioning + - The method also supports Kendall's loss + """ + self._update_train_stats(data_batch) + return self.forward(data_batch) + + @staticmethod + def get_context_parallel_group(): + if parallel_state.is_initialized(): + return parallel_state.get_context_parallel_group() + return None + + def broadcast_split_for_model_parallelsim(self, x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, sigma_B_T): + """ + Broadcast and split the input data and condition for model parallelism. + Currently, we only support context parallelism. + """ + cp_group = self.get_context_parallel_group() + cp_size = 1 if cp_group is None else cp_group.size() + if condition.is_video and cp_size > 1: + # Perform spatial split only when it's required, i.e. temporal split is not enough. + # Refer to "find_split" definition for more details. + use_spatial_split = cp_size > x0_B_C_T_H_W.shape[2] or x0_B_C_T_H_W.shape[2] % cp_size != 0 + after_split_shape = find_split(x0_B_C_T_H_W.shape, cp_size) if use_spatial_split else None + if use_spatial_split: + x0_B_C_T_H_W = rearrange(x0_B_C_T_H_W, "B C T H W -> B C (T H W)") + if epsilon_B_C_T_H_W is not None: + epsilon_B_C_T_H_W = rearrange(epsilon_B_C_T_H_W, "B C T H W -> B C (T H W)") + x0_B_C_T_H_W = broadcast_split_tensor(x0_B_C_T_H_W, seq_dim=2, process_group=cp_group) + epsilon_B_C_T_H_W = broadcast_split_tensor(epsilon_B_C_T_H_W, seq_dim=2, process_group=cp_group) + if use_spatial_split: + x0_B_C_T_H_W = rearrange( + x0_B_C_T_H_W, "B C (T H W) -> B C T H W", T=after_split_shape[0], H=after_split_shape[1] + ) + if epsilon_B_C_T_H_W is not None: + epsilon_B_C_T_H_W = rearrange( + epsilon_B_C_T_H_W, "B C (T H W) -> B C T H W", T=after_split_shape[0], H=after_split_shape[1] + ) + if sigma_B_T is not None: + assert sigma_B_T.ndim == 2, "sigma_B_T should be 2D tensor" + if sigma_B_T.shape[-1] == 1: # single sigma is shared across all frames + sigma_B_T = broadcast(sigma_B_T, cp_group) + else: # different sigma for each frame + sigma_B_T = broadcast_split_tensor(sigma_B_T, seq_dim=1, process_group=cp_group) + if condition is not None: + condition = condition.broadcast(cp_group) + self.net.enable_context_parallel(cp_group) + else: + self.net.disable_context_parallel() + + return x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, sigma_B_T + + def _update_train_stats(self, data_batch: dict[str, torch.Tensor]) -> None: + is_image = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image else self.input_data_key + if isinstance(self.net, WeightTrainingStat): + if is_image: + self.net.accum_image_sample_counter += data_batch[input_key].shape[0] * self.data_parallel_size + else: + self.net.accum_video_sample_counter += data_batch[input_key].shape[0] * self.data_parallel_size + + # ------------------------ Sampling ------------------------ + + def get_velocity_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `velocity_fn` based on the provided data batch and guidance factor. + + This function first processes the input data batch through a conditioning workflow (`conditioner`) to obtain conditioned and unconditioned states. It then defines a nested function `velocity_fn` which applies a denoising operation on an input `noise_x` at a given noise level `sigma` using both the conditioned and unconditioned states. + + Args: + - data_batch (Dict): A batch of data used for conditioning. The format and content of this dictionary should align with the expectations of the `self.conditioner` + - guidance (float, optional): A scalar value that modulates the influence of the conditioned state relative to the unconditioned state in the output. Defaults to 1.5. + - is_negative_prompt (bool): use negative prompt t5 in uncondition if true + + Returns: + - Callable: A function `velocity_fn(noise_x, sigma)` that takes two arguments, `noise_x` and `sigma`, and return velocity predictoin + + The returned function is suitable for use in scenarios where a denoised state is required based on both conditioned and unconditioned inputs, with an adjustable level of guidance influence. + """ + _, x0, _ = self.get_data_and_condition(data_batch) # we need always process the data batch first. + is_image_batch = self.is_image_batch(data_batch) + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + # For inference, check if parallel_state is initialized + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def velocity_fn(noise: torch.Tensor, noise_x: torch.Tensor, timestep: torch.Tensor) -> torch.Tensor: + cond_v = self.denoise(noise, noise_x, timestep, condition) + uncond_v = self.denoise(noise, noise_x, timestep, uncondition) + velocity_pred = uncond_v + guidance * (cond_v - uncond_v) + return velocity_pred + + return velocity_fn + + @torch.no_grad() + def generate_samples_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + is_negative_prompt: bool = False, + num_steps: int = 35, + shift: float = 5.0, + **kwargs, + ) -> torch.Tensor: + """ + Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples. + Args: + data_batch (dict): raw data batch draw from the training data loader. + iteration (int): Current iteration number. + guidance (float): guidance weights + seed (int): random seed + state_shape (tuple): shape of the state, default to data batch if not provided + n_sample (int): number of samples to generate + is_negative_prompt (bool): use negative prompt t5 in uncondition if true + num_steps (int): number of steps for the diffusion process + """ + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + state_shape = [ + self.config.state_ch, + self.tokenizer.get_latent_num_frames(_T), + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + noise = misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + + seed_g = torch.Generator(device=self.tensor_kwargs["device"]) + seed_g.manual_seed(seed) + + self.sample_scheduler.set_timesteps( + num_steps, + device=self.tensor_kwargs["device"], + shift=shift, + use_kerras_sigma=self.config.use_kerras_sigma_at_inference, + ) + + timesteps = self.sample_scheduler.timesteps + + velocity_fn = self.get_velocity_fn_from_batch(data_batch, guidance, is_negative_prompt=is_negative_prompt) + use_spatial_split = False + if self.net.is_context_parallel_enabled: + cp_size = len(torch.distributed.get_process_group_ranks(self.get_context_parallel_group())) + n_views = noise.shape[2] // self.get_num_video_latent_frames() + # Perform spatial split only when it's required, i.e. temporal split is not enough. + # Refer to "find_split" definition for more details. + state_t = noise.shape[2] // n_views + use_spatial_split = cp_size > state_t or state_t % cp_size != 0 + after_split_shape = None + if use_spatial_split: + after_split_shape = find_split(noise.shape, cp_size, view_factor=n_views) + after_split_shape = torch.Size([after_split_shape[0] * n_views, *after_split_shape[1:]]) + noise = rearrange(noise, "b c t h w -> b c (t h w)") + noise = broadcast_split_tensor(tensor=noise, seq_dim=2, process_group=self.get_context_parallel_group()) + if use_spatial_split: + noise = rearrange(noise, "b c (t h w) -> b c t h w", t=after_split_shape[0], h=after_split_shape[1]) + latents = noise + + if INTERNAL: + timesteps_iter = timesteps + else: + timesteps_iter = tqdm.tqdm(timesteps, desc="Generating samples", total=len(timesteps)) + + for _, t in enumerate(timesteps_iter): + latent_model_input = latents + timestep = [t] + + timestep = torch.stack(timestep) + + velocity_pred = velocity_fn(noise, latent_model_input, timestep.unsqueeze(0)) + temp_x0 = self.sample_scheduler.step( + velocity_pred.unsqueeze(0), t, latents[0].unsqueeze(0), return_dict=False, generator=seed_g + )[0] + latents = temp_x0.squeeze(0) + + if self.net.is_context_parallel_enabled: + if use_spatial_split: + latents = rearrange(latents, "b c t h w -> b c (t h w)") + latents = cat_outputs_cp(latents, seq_dim=2, cp_group=self.get_context_parallel_group()) + if use_spatial_split: + latents = rearrange(latents, "b c (t h w) -> b c t h w", t=state_shape[1], h=state_shape[2]) + + return latents + + @torch.no_grad() + def generate_samples_from_batch_lora( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + is_negative_prompt: bool = False, + num_steps: int = 35, + shift: float = 5.0, + disable_lora_at_low_t: bool = False, + lora_disable_threshold_step: int = 900, + adapter_switch_timesteps: list[int] = [], + **kwargs, + ) -> torch.Tensor: + """ + Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples. + Args: + data_batch (dict): raw data batch draw from the training data loader. + iteration (int): Current iteration number. + guidance (float): guidance weights + seed (int): random seed + state_shape (tuple): shape of the state, default to data batch if not provided + n_sample (int): number of samples to generate + is_negative_prompt (bool): use negative prompt t5 in uncondition if true + num_steps (int): number of steps for the diffusion process + disable_lora_at_low_t (bool): if True, disable LoRA modules when timestep < lora_disable_threshold_step + lora_disable_threshold_step (int): step below which LoRA modules are disabled (default: 900) + adapter_switch_timesteps (list[int]): list of timesteps to switch to a different adapter. For example, [900, 700] means using adapter_0 from timestep 1000 until timestep 900, then switch to adapter_1 at timestep 900 and switch to adapter_2 at timestep 700. + """ + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + state_shape = [ + self.config.state_ch, + self.tokenizer.get_latent_num_frames(_T), + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + noise = misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + + seed_g = torch.Generator(device=self.tensor_kwargs["device"]) + seed_g.manual_seed(seed) + + self.sample_scheduler.set_timesteps( + num_steps, + device=self.tensor_kwargs["device"], + shift=shift, + use_kerras_sigma=self.config.use_kerras_sigma_at_inference, + ) + + timesteps = self.sample_scheduler.timesteps + + velocity_fn = self.get_velocity_fn_from_batch(data_batch, guidance, is_negative_prompt=is_negative_prompt) + use_spatial_split = False + if self.net.is_context_parallel_enabled: + cp_size = len(torch.distributed.get_process_group_ranks(self.get_context_parallel_group())) + n_views = noise.shape[2] // self.get_num_video_latent_frames() + # Perform spatial split only when it's required, i.e. temporal split is not enough. + # Refer to "find_split" definition for more details. + state_t = noise.shape[2] // n_views + use_spatial_split = cp_size > state_t or state_t % cp_size != 0 + after_split_shape = None + if use_spatial_split: + after_split_shape = find_split(noise.shape, cp_size, view_factor=n_views) + after_split_shape = torch.Size([after_split_shape[0] * n_views, *after_split_shape[1:]]) + noise = rearrange(noise, "b c t h w -> b c (t h w)") + noise = broadcast_split_tensor(tensor=noise, seq_dim=2, process_group=self.get_context_parallel_group()) + if use_spatial_split: + noise = rearrange(noise, "b c (t h w) -> b c t h w", t=after_split_shape[0], h=after_split_shape[1]) + latents = noise + + if INTERNAL: + timesteps_iter = timesteps + else: + timesteps_iter = tqdm.tqdm(timesteps, desc="Generating samples", total=len(timesteps)) + + lora_disabled = False + if adapter_switch_timesteps: + t_prev = 1000 + adapter_switch_timesteps = [1000] + adapter_switch_timesteps + for iter_idx, t in enumerate(timesteps_iter): + if disable_lora_at_low_t and self.config.use_lora and not lora_disabled and t < lora_disable_threshold_step: + log.info(f"Disabling LoRA adapters at timestep {t} (threshold: {lora_disable_threshold_step})") + self.net.disable_adapter_layers() + lora_disabled = True + if adapter_switch_timesteps: + for adapter_idx, adapter_switch_timestep in enumerate(adapter_switch_timesteps): + if t <= adapter_switch_timestep and t_prev >= adapter_switch_timestep: + adapter_name = f"adapter_{adapter_idx}" + self.net.set_adapter(adapter_name) + log.info(f"Activated {adapter_name} at timestep {t}, step {iter_idx}") + t_prev = t + + latent_model_input = latents + timestep = [t] + + timestep = torch.stack(timestep) + + velocity_pred = velocity_fn(noise, latent_model_input, timestep.unsqueeze(0)) + temp_x0 = self.sample_scheduler.step( + velocity_pred.unsqueeze(0), t, latents[0].unsqueeze(0), return_dict=False, generator=seed_g + )[0] + latents = temp_x0.squeeze(0) + + # Re-enable LoRA if it was disabled + if lora_disabled: + log.info("Re-enabling LoRA adapters after sampling") + self.net.set_adapter("default") + + if self.net.is_context_parallel_enabled: + if use_spatial_split: + latents = rearrange(latents, "b c t h w -> b c (t h w)") + latents = cat_outputs_cp(latents, seq_dim=2, cp_group=self.get_context_parallel_group()) + if use_spatial_split: + latents = rearrange(latents, "b c (t h w) -> b c t h w", t=state_shape[1], h=state_shape[2]) + + return latents + + # ------------------------ Sampling ------------------------ + @torch.no_grad() + def generate_samples_from_batch_dmd2( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + is_negative_prompt: bool = False, + num_steps: int = 35, + # solver_option: COMMON_SOLVER_OPTIONS = "2ab", + init_noise: torch.Tensor = None, + # mid_t: List[float] | None = None, + **kwargs, + ) -> torch.Tensor: + """ + Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples. + his function is only used for distilled (e.g., 4-step) inference of a model that is merged from a Transfer model and a distilled Predict model. + Args: + data_batch (dict): raw data batch draw from the training data loader. + iteration (int): Current iteration number. + guidance (float): guidance weights + seed (int): random seed + state_shape (tuple): shape of the state, default to data batch if not provided + n_sample (int): number of samples to generate + is_negative_prompt (bool): use negative prompt t5 in uncondition if true + num_steps (int): number of steps for the diffusion process + solver_option (str): differential equation solver option, default to "2ab"~(mulitstep solver) + """ + del kwargs + # This function is only used for distilled (e.g., 4-step) inference of a model that is merged from a Transfer model and a distilled Predict model. + # Transfer model has self.net.timestep_scale = 0.001, while the distilled Predict model has self.net.timestep_scale = 1.0. At inference, we load the config from the Transfer model and then overwrite the timestep_scale to 1.0 to align the timestep parameters with the distilled Predict model. + self.net.timestep_scale = 1.0 + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + state_shape = [ + self.config.state_ch, + self.tokenizer.get_latent_num_frames(_T), + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + x0_fn = self.get_x0_fn_from_batch(data_batch, guidance) + + generator = torch.Generator(device=self.tensor_kwargs["device"]) + generator.manual_seed(seed) + + if init_noise is None: + init_noise = torch.randn( + (n_sample,) + tuple(state_shape), + # n_sample, + # *state_shape, + dtype=torch.float32, + device=self.tensor_kwargs["device"], + generator=generator, + ) + # log.info(f"init_noise shape: {init_noise} {(n_sample,) + tuple(state_shape)}") + use_spatial_split = False + if self.net.is_context_parallel_enabled: # type: ignore + cp_size = len(torch.distributed.get_process_group_ranks(self.get_context_parallel_group())) + # Perform spatial split only when it's required, i.e. temporal split is not enough. + # Refer to "find_split" definition for more details. + use_spatial_split = cp_size > init_noise.shape[2] or init_noise.shape[2] % cp_size != 0 + after_split_shape = None + if use_spatial_split: + n_views = init_noise.shape[2] // self.get_num_video_latent_frames() + after_split_shape = find_split(init_noise.shape, cp_size, n_views) + after_split_shape = torch.Size([after_split_shape[0] * n_views, *after_split_shape[1:]]) + init_noise = rearrange(init_noise, "b c t h w -> b c (t h w)") + + init_noise = broadcast_split_tensor(init_noise, seq_dim=2, process_group=self.get_context_parallel_group()) + if use_spatial_split: + init_noise = rearrange( + init_noise, "b c (t h w) -> b c t h w", t=after_split_shape[0], h=after_split_shape[1] + ) + + # Sampling steps + x = init_noise.to(torch.float64) + ones = torch.ones(x.size(0), device=x.device, dtype=x.dtype) + t_steps = self.config.selected_sampling_time[:num_steps] + [ + 0, + ] + for t_cur, t_next in zip(t_steps[:-1], t_steps[1:]): + x = x0_fn(x.float(), t_cur * ones).to(torch.float64) + if t_next > 1e-5: + x = math.cos(t_next) * x / self.sigma_data + math.sin(t_next) * init_noise + samples = x.float() + if self.net.is_context_parallel_enabled: # type: ignore + if use_spatial_split: + samples = rearrange(samples, "b c t h w -> b c (t h w)") + samples = cat_outputs_cp(samples, seq_dim=2, cp_group=self.get_context_parallel_group()) + if use_spatial_split: + samples = rearrange(samples, "b c (t h w) -> b c t h w", t=state_shape[1], h=state_shape[2]) + return torch.nan_to_num(samples) + + @torch.no_grad() + def validation_step( + self, data: dict[str, torch.Tensor], iteration: int + ) -> tuple[dict[str, torch.Tensor], torch.Tensor]: + return self.forward(data) + + def forward(self, data_batch: dict[str, torch.Tensor]) -> tuple[dict[str, torch.Tensor], torch.Tensor]: + # Obtain text embeddings online + if self.config.text_encoder_config is not None and self.config.text_encoder_config.compute_online: + text_embeddings = self.text_encoder.compute_text_embeddings_online(data_batch, self.input_caption_key) + data_batch["t5_text_embeddings"] = text_embeddings + data_batch["t5_text_mask"] = torch.ones(text_embeddings.shape[0], text_embeddings.shape[1], device="cuda") + + # Get the input data to noise and denoise~(image, video) and the corresponding conditioner. + _, x0_B_C_T_H_W, condition = self.get_data_and_condition(data_batch) + + # Sample pertubation noise levels and N(0, 1) noises + epsilon_B_C_T_H_W = torch.randn(x0_B_C_T_H_W.size(), **self.tensor_kwargs_fp32) + batch_size = x0_B_C_T_H_W.size()[0] + t_B = self.rectified_flow.sample_train_time(batch_size).to(**self.tensor_kwargs_fp32) + t_B = rearrange(t_B, "b -> b 1") # add a dimension for T, all frames share the same sigma + + x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, t_B = self.broadcast_split_for_model_parallelsim( + x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, t_B + ) + timesteps = self.rectified_flow.get_discrete_timestamp(t_B, self.tensor_kwargs_fp32) + + if self.config.use_high_sigma_strategy: + # Use high sigma strategy + mask = torch.rand(timesteps.shape, device=timesteps.device) < self.config.high_sigma_ratio + + candidate_timesteps = self.rectified_flow.noise_scheduler.timesteps.to(device=timesteps.device) + candidate_timesteps = candidate_timesteps[ + (candidate_timesteps >= self.config.high_sigma_timesteps_min) + & (candidate_timesteps <= self.config.high_sigma_timesteps_max) + ] + + if len(candidate_timesteps) > 0: + # Sample timesteps.shape values from candidate_timesteps with replacement + new_timesteps = candidate_timesteps[torch.randint(0, len(candidate_timesteps), timesteps.shape)] + timesteps = torch.where(mask, new_timesteps, timesteps) + else: + raise ValueError("No candidate timesteps found for high sigma strategy") + + sigmas = self.rectified_flow.get_sigmas( + timesteps, + self.tensor_kwargs_fp32, + ) + + timesteps = rearrange(timesteps, "b -> b 1") + sigmas = rearrange(sigmas, "b -> b 1") + xt_B_C_T_H_W, vt_B_C_T_H_W = self.rectified_flow.get_interpolation(epsilon_B_C_T_H_W, x0_B_C_T_H_W, sigmas) + + vt_pred_B_C_T_H_W = self.denoise( + noise=epsilon_B_C_T_H_W, + xt_B_C_T_H_W=xt_B_C_T_H_W.to(**self.tensor_kwargs), + timesteps_B_T=timesteps, + condition=condition, + ) + + time_weights_B = self.rectified_flow.train_time_weight(timesteps, self.tensor_kwargs_fp32) + per_instance_loss = torch.mean( + (vt_pred_B_C_T_H_W - vt_B_C_T_H_W) ** 2, dim=list(range(1, vt_pred_B_C_T_H_W.dim())) + ) + + loss = torch.mean(time_weights_B * per_instance_loss) + output_batch = { + "x0": x0_B_C_T_H_W, + "xt": xt_B_C_T_H_W, + "sigma": sigmas, + "condition": condition, + "model_pred": vt_pred_B_C_T_H_W, + "edm_loss": loss, + } + + return output_batch, loss + + def get_data_and_condition(self, data_batch: dict[str, torch.Tensor]) -> Tuple[Tensor, Tensor, Text2WorldCondition]: + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + + # Latent state + raw_state = data_batch[self.input_image_key if is_image_batch else self.input_data_key] + latent_state = self.encode(raw_state).contiguous().float() + + # Condition + condition = self.conditioner(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + return raw_state, latent_state, condition + + def _normalize_video_databatch_inplace(self, data_batch: dict[str, Tensor], input_key: str = None) -> None: + """ + Normalizes video data in-place on a CUDA device to reduce data loading overhead. + + This function modifies the video data tensor within the provided data_batch dictionary + in-place, scaling the uint8 data from the range [0, 255] to the normalized range [-1, 1]. + + Warning: + A warning is issued if the data has not been previously normalized. + + Args: + data_batch (dict[str, Tensor]): A dictionary containing the video data under a specific key. + This tensor is expected to be on a CUDA device and have dtype of torch.uint8. + + Side Effects: + Modifies the 'input_data_key' tensor within the 'data_batch' dictionary in-place. + + Note: + This operation is performed directly on the CUDA device to avoid the overhead associated + with moving data to/from the GPU. Ensure that the tensor is already on the appropriate device + and has the correct dtype (torch.uint8) to avoid unexpected behaviors. + """ + input_key = self.input_data_key if input_key is None else input_key + # only handle video batch + if input_key in data_batch: + # Check if the data has already been normalized and avoid re-normalizing + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert torch.is_floating_point(data_batch[input_key]), "Video data is not in float format." + assert torch.all((data_batch[input_key] >= -1.0001) & (data_batch[input_key] <= 1.0001)), ( + f"Video data is not in the range [-1, 1]. get data range [{data_batch[input_key].min()}, {data_batch[input_key].max()}]" + ) + else: + assert data_batch[input_key].dtype == torch.uint8, "Video data is not in uint8 format." + data_batch[input_key] = data_batch[input_key].to(**self.tensor_kwargs) / 127.5 - 1.0 + data_batch[IS_PREPROCESSED_KEY] = True + + def _augment_image_dim_inplace(self, data_batch: dict[str, Tensor], input_key: str = None) -> None: + input_key = self.input_image_key if input_key is None else input_key + if input_key in data_batch: + # Check if the data has already been augmented and avoid re-augmenting + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert data_batch[input_key].shape[2] == 1, ( + f"Image data is claimed be augmented while its shape is {data_batch[input_key].shape}" + ) + return + else: + data_batch[input_key] = rearrange(data_batch[input_key], "b c h w -> b c 1 h w").contiguous() + data_batch[IS_PREPROCESSED_KEY] = True + + # ------------------ Checkpointing ------------------ + + def state_dict(self) -> Dict[str, Any]: # noqa: F821 + net_state_dict = self.net.state_dict(prefix="net.") + if self.config.ema.enabled: + ema_state_dict = self.net_ema.state_dict(prefix="net_ema.") + net_state_dict.update(ema_state_dict) + return net_state_dict + + def load_state_dict(self, state_dict: Mapping[str, Any], strict: bool = True, assign: bool = False): # noqa: F821 + """ + Loads a state dictionary into the model and optionally its EMA counterpart. + Different from torch strict=False mode, the method will not raise error for unmatched state shape while raise warning. + + Parameters:e + state_dict (Mapping[str, Any]): A dictionary containing separate state dictionaries for the model and + potentially for an EMA version of the model under the keys 'model' and 'ema', respectively. + strict (bool, optional): If True, the method will enforce that the keys in the state dict match exactly + those in the model and EMA model (if applicable). Defaults to True. + assign (bool, optional): If True and in strict mode, will assign the state dictionary directly rather than + matching keys one-by-one. This is typically used when loading parts of state dicts + or using customized loading procedures. Defaults to False. + """ + _reg_state_dict = collections.OrderedDict() + _ema_state_dict = collections.OrderedDict() + for k, v in state_dict.items(): + if k.startswith("net."): + _reg_state_dict[k.replace("net.", "")] = v + elif k.startswith("net_ema."): + _ema_state_dict[k.replace("net_ema.", "")] = v + + state_dict = _reg_state_dict + + if strict: + reg_results: _IncompatibleKeys = self.net.load_state_dict(_reg_state_dict, strict=strict, assign=assign) + + if self.config.ema.enabled: + ema_results: _IncompatibleKeys = self.net_ema.load_state_dict( + _ema_state_dict, strict=strict, assign=assign + ) + + return _IncompatibleKeys( + missing_keys=reg_results.missing_keys + (ema_results.missing_keys if self.config.ema.enabled else []), + unexpected_keys=reg_results.unexpected_keys + + (ema_results.unexpected_keys if self.config.ema.enabled else []), + ) + else: + log.critical("load model in non-strict mode") + log.critical(non_strict_load_model(self.net, _reg_state_dict), rank0_only=False) + if self.config.ema.enabled: + log.critical("load ema model in non-strict mode") + log.critical(non_strict_load_model(self.net_ema, _ema_state_dict), rank0_only=False) + + # ------------------ public methods ------------------ + def ema_beta(self, iteration: int) -> float: + """ + Calculate the beta value for EMA update. + weights = weights * beta + (1 - beta) * new_weights + + Args: + iteration (int): Current iteration number. + + Returns: + float: The calculated beta value. + """ + iteration = iteration + self.config.ema.iteration_shift + if iteration < 1: + return 0.0 + return (1 - 1 / (iteration + 1)) ** (self.ema_exp_coefficient + 1) + + def model_param_stats(self) -> Dict[str, int]: + return {"total_learnable_param_num": self._param_count} + + def is_image_batch(self, data_batch: dict[str, Tensor]) -> bool: + """We hanlde two types of data_batch. One comes from a joint_dataloader where "dataset_name" can be used to differenciate image_batch and video_batch. + Another comes from a dataloader which we by default assumes as video_data for video model training. + """ + is_image = self.input_image_key in data_batch + is_video = self.input_data_key in data_batch + assert is_image != is_video, ( + "Only one of the input_image_key or input_data_key should be present in the data_batch." + ) + return is_image + + def denoise( + self, + noise: torch.Tensor, + xt_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + condition: Text2WorldCondition, + ) -> DenoisePrediction: + """ + Performs denoising on the input noise data, noise level, and condition + + Args: + xt (torch.Tensor): The input noise data. + timesteps_B_T (torch.Tensor): The timestep. + condition (Text2WorldCondition): conditional information, generated from self.conditioner + + Returns: + velocity prediction + """ + del noise + + net_output_B_C_T_H_W = self.net( + x_B_C_T_H_W=(xt_B_C_T_H_W).to(**self.tensor_kwargs), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + timesteps_B_T=timesteps_B_T, # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + **condition.to_dict(), + ).float() + + return net_output_B_C_T_H_W + + @torch.no_grad() + def encode(self, state: torch.Tensor) -> torch.Tensor: + return self.tokenizer.encode(state) + + @torch.no_grad() + def decode(self, latent: torch.Tensor) -> torch.Tensor: + return self.tokenizer.decode(latent) + + def get_video_height_width(self) -> Tuple[int, int]: + return VIDEO_RES_SIZE_INFO[self.config.resolution]["9,16"] + + def get_video_latent_height_width(self) -> Tuple[int, int]: + height, width = VIDEO_RES_SIZE_INFO[self.config.resolution]["9,16"] + return height // self.tokenizer.spatial_compression_factor, width // self.tokenizer.spatial_compression_factor + + def get_num_video_latent_frames(self) -> int: + return self.config.state_t + + @property + def text_encoder_class(self) -> str: + return self.config.text_encoder_class + + @contextmanager + def ema_scope(self, context=None, is_cpu=False): + if self.config.ema.enabled: + # https://github.com/pytorch/pytorch/issues/144289 + for module in self.net.modules(): + if isinstance(module, FSDPModule): + module.reshard() + self.net_ema_worker.cache(self.net.parameters(), is_cpu=is_cpu) + self.net_ema_worker.copy_to(src_model=self.net_ema, tgt_model=self.net) + if context is not None: + log.info(f"{context}: Switched to EMA weights") + try: + yield None + finally: + if self.config.ema.enabled: + for module in self.net.modules(): + if isinstance(module, FSDPModule): + module.reshard() + self.net_ema_worker.restore(self.net.parameters()) + if context is not None: + log.info(f"{context}: Restored training weights") + + def clip_grad_norm_( + self, + max_norm: float, + norm_type: float = 2.0, + error_if_nonfinite: bool = False, + foreach: Optional[bool] = None, + ): + return clip_grad_norm_( + self.net.parameters(), + max_norm, + norm_type=norm_type, + error_if_nonfinite=error_if_nonfinite, + foreach=foreach, + ) + + def add_lora( + self, + network: torch.nn.Module, + lora_rank: int = 4, + lora_alpha: int = 4, + lora_target_modules: str = "q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2", + init_lora_weights: bool = True, + use_dora: bool = False, + ) -> torch.nn.Module: + """Add LoRA (Low-Rank Adaptation) adapters to `self.net`. + + This function injects LoRA adapters into specified modules of the network, + enabling parameter-efficient fine-tuning by training only a small number + of additional parameters. + + Args: + lora_rank: The rank of the LoRA adaptation matrices. Higher rank allows + more expressiveness but uses more parameters (default: 4) + lora_alpha: Scaling parameter for LoRA. Controls the magnitude of the + LoRA adaptation (default: 4) + lora_target_modules: Comma-separated string of module names to target + for LoRA adaptation (default: attention and MLP layers) + init_lora_weights: Whether to initialize LoRA weights properly (default: True) + + Raises: + ImportError: If PEFT library is not installed + ValueError: If invalid parameters are provided + RuntimeError: If LoRA injection fails + """ + assert network is not None, "Network is not initialized" + try: + from peft import LoraConfig, get_peft_model + except ImportError as e: + raise ImportError( + "PEFT library is required for LoRA training. Please install it with: pip install peft" + ) from e + + # Validate parameters + if lora_rank <= 0: + raise ValueError(f"LoRA rank must be positive, got {lora_rank}") + if lora_alpha < 0: + raise ValueError(f"LoRA alpha must be positive, got {lora_alpha}") + + target_modules_list = [module.strip() for module in lora_target_modules.split(",")] + if not target_modules_list: + raise ValueError("LoRA target_modules cannot be empty") + + # Validate target modules exist in model + model_module_names = set(name for name, _ in network.named_modules()) + invalid_modules = [] + for target_module in target_modules_list: + # Check if any module contains this target pattern + if not any(target_module in module_name for module_name in model_module_names): + invalid_modules.append(target_module) + + if invalid_modules: + log.warning(f"Target modules not found in model: {invalid_modules}") + + # Add LoRA to model + self.lora_alpha = lora_alpha + + log.info( + f"Adding LoRA adapters: rank={lora_rank}, alpha={lora_alpha}, targets={target_modules_list}, use_dora={use_dora}" + ) + + lora_config = LoraConfig( + r=lora_rank, + lora_alpha=lora_alpha, + init_lora_weights=init_lora_weights, + target_modules=target_modules_list, + use_dora=use_dora, + ) + + try: + network = get_peft_model(network, lora_config) + except Exception as e: + raise RuntimeError(f"Failed to inject LoRA adapters into model: {e}") from e + + # Count and log LoRA parameters + lora_params = 0 + total_params = 0 + for name, param in network.named_parameters(): + total_params += param.numel() + if param.requires_grad: + lora_params += param.numel() + # Upcast LoRA parameters into fp32 + param.data = param.to(torch.float32) + + log.info( + f"LoRA injection successful: {lora_params:,} trainable parameters out of {total_params:,} total ({100 * lora_params / total_params:.3f}%)" + ) + return network diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/text2world_wan2pt1_model.py b/REGEN-main/cosmos_policy/_src/predict2/models/text2world_wan2pt1_model.py new file mode 100644 index 0000000000000000000000000000000000000000..1f24bbf249f406d89da6947cfdfa398736395a22 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/text2world_wan2pt1_model.py @@ -0,0 +1,751 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import collections +from contextlib import contextmanager +from typing import Any, Callable, Dict, Mapping, Optional, Tuple + +import attrs +import numpy as np +import torch +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor +from torch.distributed._composable.fsdp import FSDPModule, fully_shard +from torch.distributed._tensor.api import DTensor +from torch.distributed.device_mesh import DeviceMesh +from torch.nn.modules.module import _IncompatibleKeys +from torch.nn.utils.clip_grad import clip_grad_norm_ + +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.lazy_config import instantiate as lazy_instantiate +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import log, misc +from cosmos_policy._src.imaginaire.utils.checkpointer import non_strict_load_model +from cosmos_policy._src.imaginaire.utils.context_parallel import ( + broadcast, + broadcast_split_tensor, + cat_outputs_cp, +) +from cosmos_policy._src.imaginaire.utils.count_params import count_params +from cosmos_policy._src.imaginaire.utils.denoise_prediction import DenoisePrediction +from cosmos_policy._src.imaginaire.utils.ema import FastEmaModelUpdater +from cosmos_policy._src.imaginaire.utils.fsdp_helper import hsdp_device_mesh +from cosmos_policy._src.imaginaire.utils.optim_instantiate import get_base_scheduler +from cosmos_policy._src.predict2.conditioner import DataType, Text2WorldCondition +from cosmos_policy._src.predict2.datasets.utils import VIDEO_RES_SIZE_INFO +from cosmos_policy._src.predict2.models.fm_solvers_unipc import FlowUniPCMultistepScheduler +from cosmos_policy._src.predict2.models.text2world_model import EMAConfig +from cosmos_policy._src.predict2.networks.model_weights_stats import WeightTrainingStat +from cosmos_policy._src.predict2.schedulers.rectified_flow import RectifiedFlow +from cosmos_policy._src.predict2.tokenizers.base_vae import BaseVAE +from cosmos_policy._src.predict2.utils.dtensor_helper import DTensorFastEmaModelUpdater, broadcast_dtensor_model_states + +IS_PREPROCESSED_KEY = "is_preprocessed" +NUM_EMBEDDING_PADDING_TOKENS = 512 + + +@attrs.define(slots=False) +class Text2WorldModelWan2pt1Config: + """ + Config for [DiffusionModel][projects.cosmos.diffusion.v2.models.t2v_model.DiffusionModel]. + """ + + tokenizer: LazyDict = None + conditioner: LazyDict = None + net: LazyDict = None + ema: EMAConfig = EMAConfig() + + fsdp_shard_size: int = 1 + precision: str = "bfloat16" + input_data_key: str = "video" # key to fetch input data from data_batch + input_image_key: str = "images" # key to fetch input image from data_batch + input_caption_key: str = "ai_caption" # Key used to fetch input captions + use_torch_compile: bool = False + + state_ch: int = 16 # for latent model, ref to the latent channel number + state_t: int = 8 # for latent model, ref to the latent number of frames + resolution: str = "512" + + shift: int = 5 + use_dynamic_shift: bool = False + train_time_weight: str = "uniform" + + +class WANDiffusionModel(ImaginaireModel): + """ + Diffusion model. + """ + + def __init__(self, config: Text2WorldModelWan2pt1Config): + super().__init__() + + self.config = config + + self.precision = { + "float32": torch.float32, + "float16": torch.float16, + "bfloat16": torch.bfloat16, + }[config.precision] + self.tensor_kwargs = {"device": "cuda", "dtype": self.precision} + self.flow_matching_kwargs = {"device": "cuda", "dtype": torch.float32} + + log.warning(f"WANDiffusionModel: precision {self.precision}") + log.warning(f"Flow Matching: precision {self.flow_matching_kwargs['dtype']}") + + # 1. set data keys and data information + # self.sigma_data = config.sigma_data + self.setup_data_key() + + # 2. setup up diffusion processing and scaling~(pre-condition), sampler + self.sample_scheduler = FlowUniPCMultistepScheduler( + num_train_timesteps=1000, shift=1, use_dynamic_shifting=False + ) + + # 3. tokenizer + with misc.timer("WANDiffusionModel: set_up_tokenizer"): + self.tokenizer: BaseVAE = lazy_instantiate(config.tokenizer) + assert self.tokenizer.latent_ch == self.config.state_ch, ( + f"latent_ch {self.tokenizer.latent_ch} != state_shape {self.config.state_ch}" + ) + + # 5. create fsdp mesh if needed + if config.fsdp_shard_size > 1: + self.fsdp_device_mesh = hsdp_device_mesh( + sharding_group_size=config.fsdp_shard_size, + ) + else: + self.fsdp_device_mesh = None + + # 6. diffusion neural networks part + self.set_up_model() + + # 7. training states + if parallel_state.is_initialized(): + self.data_parallel_size = parallel_state.get_data_parallel_world_size() + else: + self.data_parallel_size = 1 + + # 8. rectified flow + self.rectified_flow = RectifiedFlow( + velocity_field=self.net, + train_time_distribution="logitnormal", + use_dynamic_shift=config.use_dynamic_shift, + shift=config.shift, + train_time_weight_method=config.train_time_weight, + device=torch.device("cuda"), + dtype=self.flow_matching_kwargs["dtype"], + ) + + def setup_data_key(self) -> None: + self.input_data_key = self.config.input_data_key # by default it is video key for Video diffusion model + self.input_image_key = self.config.input_image_key + + def build_net(self): + config = self.config + init_device = "meta" if self.fsdp_device_mesh else "cpu" + with misc.timer("Creating PyTorch model"): + with torch.device(init_device): + net = lazy_instantiate(config.net) + + self._param_count = count_params(net, verbose=False) + + if self.fsdp_device_mesh: + net.fully_shard(mesh=self.fsdp_device_mesh) + net = fully_shard(net, mesh=self.fsdp_device_mesh, reshard_after_forward=True) + + with misc.timer("meta to cuda and broadcast model states"): + net.to_empty(device="cuda") + # IMPORTANT: (qsh) model init should not depends on current tensor shape, or it can handle Dtensor shape. + net.init_weights() + + if self.fsdp_device_mesh: + broadcast_dtensor_model_states(net, self.fsdp_device_mesh) + for name, param in net.named_parameters(): + assert isinstance(param, DTensor), f"param should be DTensor, {name} got {type(param)}" + return net + + @misc.timer("DiffusionModel: set_up_model") + def set_up_model(self): + config = self.config + with misc.timer("Creating PyTorch model and ema if enabled"): + self.conditioner = lazy_instantiate(config.conditioner) + assert sum(p.numel() for p in self.conditioner.parameters() if p.requires_grad) == 0, ( + "conditioner should not have learnable parameters" + ) + self.net = self.build_net() + self._param_count = count_params(self.net, verbose=False) + + if config.ema.enabled: + self.net_ema = self.build_net() + self.net_ema.requires_grad_(False) + + if self.fsdp_device_mesh: + self.net_ema_worker = DTensorFastEmaModelUpdater() + else: + self.net_ema_worker = FastEmaModelUpdater() + + s = config.ema.rate + self.ema_exp_coefficient = np.roots([1, 7, 16 - s**-2, 12 - s**-2]).real.max() + + self.net_ema_worker.copy_to(src_model=self.net, tgt_model=self.net_ema) + torch.cuda.empty_cache() + + def apply_fsdp(self, dp_mesh: DeviceMesh) -> None: + """Apply FSDP to the net and net_ema.""" + # Back-to-back fully_shard calls allow for wrapping submodules and the top-level module. + self.net.fully_shard(mesh=dp_mesh) + self.net = fully_shard(self.net, mesh=dp_mesh, reshard_after_forward=True) + broadcast_dtensor_model_states(self.net, dp_mesh) + if hasattr(self, "net_ema") and self.net_ema: + self.net_ema.fully_shard(mesh=dp_mesh) + self.net_ema = fully_shard(self.net_ema, mesh=dp_mesh, reshard_after_forward=True) + broadcast_dtensor_model_states(self.net_ema, dp_mesh) + self.net_ema_worker = DTensorFastEmaModelUpdater() + # No need to copy weights to EMA when applying FSDP, it is already copied before applying FSDP. + + def init_optimizer_scheduler( + self, optimizer_config: LazyDict, scheduler_config: LazyDict + ) -> tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LRScheduler]: + """Creates the optimizer and scheduler for the model. + + Args: + config_model (ModelConfig): The config object for the model. + + Returns: + optimizer (torch.optim.Optimizer): The model optimizer. + scheduler (torch.optim.lr_scheduler.LRScheduler): The optimization scheduler. + """ + optimizer = lazy_instantiate(optimizer_config, model=self.net) + scheduler = get_base_scheduler(optimizer, self, scheduler_config) + return optimizer, scheduler + + # ------------------------ training hooks ------------------------ + def on_before_zero_grad( + self, optimizer: torch.optim.Optimizer, scheduler: torch.optim.lr_scheduler.LRScheduler, iteration: int + ) -> None: + """ + update the net_ema + """ + del scheduler, optimizer + + if self.config.ema.enabled: + # calculate beta for EMA update + ema_beta = self.ema_beta(iteration) + self.net_ema_worker.update_average(self.net, self.net_ema, beta=ema_beta) + + def on_train_start(self, memory_format: torch.memory_format = torch.preserve_format) -> None: + if self.config.ema.enabled: + self.net_ema.to(dtype=torch.float32) + if hasattr(self.tokenizer, "reset_dtype"): + self.tokenizer.reset_dtype() + self.net = self.net.to(memory_format=memory_format, **self.tensor_kwargs) + + if hasattr(self.config, "use_torch_compile") and self.config.use_torch_compile: # compatible with old config + if torch.__version__ < "2.3": + log.warning( + "torch.compile in Pytorch version older than 2.3 doesn't work well with activation checkpointing.\n" + "It's very likely there will be no significant speedup from torch.compile.\n" + "Please use at least 24.04 Pytorch container, or imaginaire4:v7 container." + ) + # Increasing cache size. It's required because of the model size and dynamic input shapes resulting in + # multiple different triton kernels. For 28 TransformerBlocks, the cache limit of 256 should be enough for + # up to 9 different input shapes, as 28*9 < 256. If you have more Blocks or input shapes, and you observe + # graph breaks at each Block (detectable with torch._dynamo.explain) or warnings about + # exceeding cache limit, you may want to increase this size. + # Starting with 24.05 Pytorch container, the default value is 256 anyway. + # You can read more about it in the comments in Pytorch source code under path torch/_dynamo/cache_size.py. + torch._dynamo.config.accumulated_cache_size_limit = 256 + # dynamic=False means that a separate kernel is created for each shape. It incurs higher compilation costs + # at initial iterations, but can result in more specialized and efficient kernels. + # dynamic=True currently throws errors in pytorch 2.3. + self.net = torch.compile(self.net, dynamic=False, disable=not self.config.use_torch_compile) + + # ------------------------ training ------------------------ + + def training_step( + self, data_batch: dict[str, torch.Tensor], iteration: int + ) -> tuple[dict[str, torch.Tensor], torch.Tensor]: + """ + Performs a single training step for the diffusion model. + + This method is responsible for executing one iteration of the model's training. It involves: + 1. Adding noise to the input data using the SDE process. + 2. Passing the noisy data through the network to generate predictions. + 3. Computing the loss based on the difference between the predictions and the original data, \ + considering any configured loss weighting. + + Args: + data_batch (dict): raw data batch draw from the training data loader. + iteration (int): Current iteration number. + + Returns: + tuple: A tuple containing two elements: + - dict: additional data that used to debug / logging / callbacks + - Tensor: The computed loss for the training step as a PyTorch Tensor. + + Raises: + AssertionError: If the class is conditional, \ + but no number of classes is specified in the network configuration. + + Notes: + - The method handles different types of conditioning + - The method also supports Kendall's loss + """ + self._update_train_stats(data_batch) + # Get the input data to noise and denoise~(image, video) and the corresponding conditioner. + _, x0_B_C_T_H_W, condition = self.get_data_and_condition(data_batch) + + # Sample pertubation noise levels and N(0, 1) noises + epsilon_B_C_T_H_W = torch.randn(x0_B_C_T_H_W.size(), **self.flow_matching_kwargs) + batch_size = x0_B_C_T_H_W.size()[0] + t_B = self.rectified_flow.sample_train_time(batch_size).to(**self.flow_matching_kwargs) + t_B = rearrange(t_B, "b -> b 1") # add a dimension for T, all frames share the same sigma + + x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, t_B = self.broadcast_split_for_model_parallelsim( + x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, t_B + ) + timesteps = self.rectified_flow.get_discrete_timestamp(t_B, self.flow_matching_kwargs) + sigmas = self.rectified_flow.get_sigmas( + timesteps, + self.flow_matching_kwargs, + ) + timesteps = rearrange(timesteps, "b -> b 1") + sigmas = rearrange(sigmas, "b -> b 1") + xt_B_C_T_H_W, vt_B_C_T_H_W = self.rectified_flow.get_interpolation(epsilon_B_C_T_H_W, x0_B_C_T_H_W, sigmas) + + vt_pred_B_C_T_H_W = self.net( + x_B_C_T_H_W=xt_B_C_T_H_W.to(**self.tensor_kwargs), + timesteps_B_T=timesteps, + **condition.to_dict(), + ) + + time_weights_B = self.rectified_flow.train_time_weight(timesteps, self.flow_matching_kwargs) + per_instance_loss = torch.mean( + (vt_pred_B_C_T_H_W - vt_B_C_T_H_W) ** 2, dim=list(range(1, vt_pred_B_C_T_H_W.dim())) + ) + + loss = torch.mean(time_weights_B * per_instance_loss) + output_batch = {"edm_loss": loss} + + return output_batch, loss + + @staticmethod + def get_context_parallel_group(): + if parallel_state.is_initialized(): + return parallel_state.get_context_parallel_group() + return None + + def broadcast_split_for_model_parallelsim(self, x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, sigma_B_T): + """ + Broadcast and split the input data and condition for model parallelism. + Currently, we only support context parallelism. + """ + cp_group = self.get_context_parallel_group() + cp_size = 1 if cp_group is None else cp_group.size() + if condition.is_video and cp_size > 1: + x0_B_C_T_H_W = broadcast_split_tensor(x0_B_C_T_H_W, seq_dim=2, process_group=cp_group) + epsilon_B_C_T_H_W = broadcast_split_tensor(epsilon_B_C_T_H_W, seq_dim=2, process_group=cp_group) + if sigma_B_T is not None: + assert sigma_B_T.ndim == 2, "sigma_B_T should be 2D tensor" + if sigma_B_T.shape[-1] == 1: # single sigma is shared across all frames + sigma_B_T = broadcast(sigma_B_T, cp_group) + else: # different sigma for each frame + sigma_B_T = broadcast_split_tensor(sigma_B_T, seq_dim=1, process_group=cp_group) + if condition is not None: + condition = condition.broadcast(cp_group) + self.net.enable_context_parallel(cp_group) + else: + self.net.disable_context_parallel() + + return x0_B_C_T_H_W, condition, epsilon_B_C_T_H_W, sigma_B_T + + def _update_train_stats(self, data_batch: dict[str, torch.Tensor]) -> None: + is_image = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image else self.input_data_key + if isinstance(self.net, WeightTrainingStat): + if is_image: + self.net.accum_image_sample_counter += data_batch[input_key].shape[0] * self.data_parallel_size + else: + self.net.accum_video_sample_counter += data_batch[input_key].shape[0] * self.data_parallel_size + + # ------------------------ Sampling ------------------------ + + def get_x0_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `x0_fn` based on the provided data batch and guidance factor. + + This function first processes the input data batch through a conditioning workflow (`conditioner`) to obtain conditioned and unconditioned states. It then defines a nested function `x0_fn` which applies a denoising operation on an input `noise_x` at a given noise level `sigma` using both the conditioned and unconditioned states. + + Args: + - data_batch (Dict): A batch of data used for conditioning. The format and content of this dictionary should align with the expectations of the `self.conditioner` + - guidance (float, optional): A scalar value that modulates the influence of the conditioned state relative to the unconditioned state in the output. Defaults to 1.5. + - is_negative_prompt (bool): use negative prompt t5 in uncondition if true + + Returns: + - Callable: A function `x0_fn(noise_x, sigma)` that takes two arguments, `noise_x` and `sigma`, and return x0 predictoin + + The returned function is suitable for use in scenarios where a denoised state is required based on both conditioned and unconditioned inputs, with an adjustable level of guidance influence. + """ + _, x0, _ = self.get_data_and_condition(data_batch) # we need always process the data batch first. + is_image_batch = self.is_image_batch(data_batch) + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + # For inference, check if parallel_state is initialized + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def x0_fn(noise_x: torch.Tensor, timestep: torch.Tensor) -> torch.Tensor: + cond_v = self.denoise(noise_x, timestep, condition) + uncond_v = self.denoise(noise_x, timestep, uncondition) + noise_pred = uncond_v + guidance * (cond_v - uncond_v) + return noise_pred + + return x0_fn + + @torch.no_grad() + def generate_samples_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + seed: int = 1, + state_shape: Tuple | None = None, + n_sample: int | None = None, + is_negative_prompt: bool = False, + num_steps: int = 35, + shift: float = 5.0, + **kwargs, + ) -> torch.Tensor: + """ + Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples. + Args: + data_batch (dict): raw data batch draw from the training data loader. + iteration (int): Current iteration number. + guidance (float): guidance weights + seed (int): random seed + state_shape (tuple): shape of the state, default to data batch if not provided + n_sample (int): number of samples to generate + is_negative_prompt (bool): use negative prompt t5 in uncondition if true + num_steps (int): number of steps for the diffusion process + solver_option (str): differential equation solver option, default to "2ab"~(mulitstep solver) + """ + + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + input_key = self.input_image_key if is_image_batch else self.input_data_key + if n_sample is None: + n_sample = data_batch[input_key].shape[0] + if state_shape is None: + _T, _H, _W = data_batch[input_key].shape[-3:] + state_shape = [ + self.config.state_ch, + self.tokenizer.get_latent_num_frames(_T), + _H // self.tokenizer.spatial_compression_factor, + _W // self.tokenizer.spatial_compression_factor, + ] + + noise = misc.arch_invariant_rand( + (n_sample,) + tuple(state_shape), + torch.float32, + self.tensor_kwargs["device"], + seed, + ) + + seed_g = torch.Generator(device=self.tensor_kwargs["device"]) + seed_g.manual_seed(seed) + + self.sample_scheduler.set_timesteps(num_steps, device=self.tensor_kwargs["device"], shift=shift) + + timesteps = self.sample_scheduler.timesteps + + x0_fn = self.get_x0_fn_from_batch(data_batch, guidance, is_negative_prompt=is_negative_prompt) + latents = noise + + if self.net.is_context_parallel_enabled: + latents = broadcast_split_tensor(tensor=latents, seq_dim=2, process_group=self.get_context_parallel_group()) + + for _, t in enumerate(timesteps): + latent_model_input = latents + timestep = [t] + + timestep = torch.stack(timestep) + + noise_pred = x0_fn(latent_model_input, timestep.unsqueeze(0)) + temp_x0 = self.sample_scheduler.step( + noise_pred.unsqueeze(0), t, latents[0].unsqueeze(0), return_dict=False, generator=seed_g + )[0] + latents = temp_x0.squeeze(0) + + if self.net.is_context_parallel_enabled: + latents = cat_outputs_cp(latents, seq_dim=2, cp_group=self.get_context_parallel_group()) + + return latents + + @torch.no_grad() + def validation_step( + self, data: dict[str, torch.Tensor], iteration: int + ) -> tuple[dict[str, torch.Tensor], torch.Tensor]: + pass + + @torch.no_grad() + def forward(self, xt, t, condition: Text2WorldCondition): + pass + + def get_data_and_condition(self, data_batch: dict[str, torch.Tensor]) -> Tuple[Tensor, Tensor, Text2WorldCondition]: + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + + # Latent state + raw_state = data_batch[self.input_image_key if is_image_batch else self.input_data_key] + latent_state = self.encode(raw_state).contiguous().float() + + # Condition + condition = self.conditioner(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + return raw_state, latent_state, condition + + def _normalize_video_databatch_inplace(self, data_batch: dict[str, Tensor], input_key: str = None) -> None: + """ + Normalizes video data in-place on a CUDA device to reduce data loading overhead. + + This function modifies the video data tensor within the provided data_batch dictionary + in-place, scaling the uint8 data from the range [0, 255] to the normalized range [-1, 1]. + + Warning: + A warning is issued if the data has not been previously normalized. + + Args: + data_batch (dict[str, Tensor]): A dictionary containing the video data under a specific key. + This tensor is expected to be on a CUDA device and have dtype of torch.uint8. + + Side Effects: + Modifies the 'input_data_key' tensor within the 'data_batch' dictionary in-place. + + Note: + This operation is performed directly on the CUDA device to avoid the overhead associated + with moving data to/from the GPU. Ensure that the tensor is already on the appropriate device + and has the correct dtype (torch.uint8) to avoid unexpected behaviors. + """ + input_key = self.input_data_key if input_key is None else input_key + # only handle video batch + if input_key in data_batch: + # Check if the data has already been normalized and avoid re-normalizing + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert torch.is_floating_point(data_batch[input_key]), "Video data is not in float format." + assert torch.all((data_batch[input_key] >= -1.0001) & (data_batch[input_key] <= 1.0001)), ( + f"Video data is not in the range [-1, 1]. get data range [{data_batch[input_key].min()}, {data_batch[input_key].max()}]" + ) + else: + assert data_batch[input_key].dtype == torch.uint8, "Video data is not in uint8 format." + data_batch[input_key] = data_batch[input_key].to(**self.tensor_kwargs) / 127.5 - 1.0 + data_batch[IS_PREPROCESSED_KEY] = True + + def _augment_image_dim_inplace(self, data_batch: dict[str, Tensor], input_key: str = None) -> None: + input_key = self.input_image_key if input_key is None else input_key + if input_key in data_batch: + # Check if the data has already been augmented and avoid re-augmenting + if IS_PREPROCESSED_KEY in data_batch and data_batch[IS_PREPROCESSED_KEY] is True: + assert data_batch[input_key].shape[2] == 1, ( + f"Image data is claimed be augmented while its shape is {data_batch[input_key].shape}" + ) + return + else: + data_batch[input_key] = rearrange(data_batch[input_key], "b c h w -> b c 1 h w").contiguous() + data_batch[IS_PREPROCESSED_KEY] = True + + # ------------------ Checkpointing ------------------ + + def state_dict(self) -> Dict[str, Any]: + net_state_dict = self.net.state_dict(prefix="net.") + if self.config.ema.enabled: + ema_state_dict = self.net_ema.state_dict(prefix="net_ema.") + net_state_dict.update(ema_state_dict) + return net_state_dict + + def load_state_dict(self, state_dict: Mapping[str, Any], strict: bool = True, assign: bool = False): + """ + Loads a state dictionary into the model and optionally its EMA counterpart. + Different from torch strict=False mode, the method will not raise error for unmatched state shape while raise warning. + + Parameters:e + state_dict (Mapping[str, Any]): A dictionary containing separate state dictionaries for the model and + potentially for an EMA version of the model under the keys 'model' and 'ema', respectively. + strict (bool, optional): If True, the method will enforce that the keys in the state dict match exactly + those in the model and EMA model (if applicable). Defaults to True. + assign (bool, optional): If True and in strict mode, will assign the state dictionary directly rather than + matching keys one-by-one. This is typically used when loading parts of state dicts + or using customized loading procedures. Defaults to False. + """ + _reg_state_dict = collections.OrderedDict() + _ema_state_dict = collections.OrderedDict() + for k, v in state_dict.items(): + if k.startswith("net."): + _reg_state_dict[k.replace("net.", "")] = v + elif k.startswith("net_ema."): + _ema_state_dict[k.replace("net_ema.", "")] = v + + state_dict = _reg_state_dict + + if strict: + reg_results: _IncompatibleKeys = self.net.load_state_dict(_reg_state_dict, strict=strict, assign=assign) + + if self.config.ema.enabled: + ema_results: _IncompatibleKeys = self.net_ema.load_state_dict( + _ema_state_dict, strict=strict, assign=assign + ) + + return _IncompatibleKeys( + missing_keys=reg_results.missing_keys + (ema_results.missing_keys if self.config.ema.enabled else []), + unexpected_keys=reg_results.unexpected_keys + + (ema_results.unexpected_keys if self.config.ema.enabled else []), + ) + else: + log.critical("load model in non-strict mode") + log.critical(non_strict_load_model(self.net, _reg_state_dict), rank0_only=False) + if self.config.ema.enabled: + log.critical("load ema model in non-strict mode") + log.critical(non_strict_load_model(self.net_ema, _ema_state_dict), rank0_only=False) + + # ------------------ public methods ------------------ + def ema_beta(self, iteration: int) -> float: + """ + Calculate the beta value for EMA update. + weights = weights * beta + (1 - beta) * new_weights + + Args: + iteration (int): Current iteration number. + + Returns: + float: The calculated beta value. + """ + iteration = iteration + self.config.ema.iteration_shift + if iteration < 1: + return 0.0 + return (1 - 1 / (iteration + 1)) ** (self.ema_exp_coefficient + 1) + + def model_param_stats(self) -> Dict[str, int]: + return {"total_learnable_param_num": self._param_count} + + def is_image_batch(self, data_batch: dict[str, Tensor]) -> bool: + """We hanlde two types of data_batch. One comes from a joint_dataloader where "dataset_name" can be used to differenciate image_batch and video_batch. + Another comes from a dataloader which we by default assumes as video_data for video model training. + """ + is_image = self.input_image_key in data_batch + is_video = self.input_data_key in data_batch + assert is_image != is_video, ( + "Only one of the input_image_key or input_data_key should be present in the data_batch." + ) + return is_image + + def denoise( + self, xt_B_C_T_H_W: torch.Tensor, timestep: torch.Tensor, condition: Text2WorldCondition + ) -> DenoisePrediction: + """ + Performs denoising on the input noise data, noise level, and condition + + Args: + xt (torch.Tensor): The input noise data. + timestep (torch.Tensor): The timestep level. + condition (Text2WorldCondition): conditional information, generated from self.conditioner + + Returns: + DenoisePrediction: The denoised prediction, it includes clean data predicton (x0), \ + noise prediction (eps_pred). + """ + # forward pass through the network + net_output_B_C_T_H_W = self.net( + x_B_C_T_H_W=(xt_B_C_T_H_W).to(**self.tensor_kwargs), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + timesteps_B_T=timestep, # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + **condition.to_dict(), + ).float() + + return net_output_B_C_T_H_W + + @torch.no_grad() + def encode(self, state: torch.Tensor) -> torch.Tensor: + return self.tokenizer.encode(state) + + @torch.no_grad() + def decode(self, latent: torch.Tensor) -> torch.Tensor: + return self.tokenizer.decode(latent) + + def get_video_height_width(self) -> Tuple[int, int]: + return VIDEO_RES_SIZE_INFO[self.config.resolution]["9,16"] + + def get_video_latent_height_width(self) -> Tuple[int, int]: + height, width = VIDEO_RES_SIZE_INFO[self.config.resolution]["9,16"] + return height // self.tokenizer.spatial_compression_factor, width // self.tokenizer.spatial_compression_factor + + def get_num_video_latent_frames(self) -> int: + return self.config.state_t + + @contextmanager + def ema_scope(self, context=None, is_cpu=False): + if self.config.ema.enabled: + # https://github.com/pytorch/pytorch/issues/144289 + for module in self.net.modules(): + if isinstance(module, FSDPModule): + module.reshard() + self.net_ema_worker.cache(self.net.parameters(), is_cpu=is_cpu) + self.net_ema_worker.copy_to(src_model=self.net_ema, tgt_model=self.net) + if context is not None: + log.info(f"{context}: Switched to EMA weights") + try: + yield None + finally: + if self.config.ema.enabled: + for module in self.net.modules(): + if isinstance(module, FSDPModule): + module.reshard() + self.net_ema_worker.restore(self.net.parameters()) + if context is not None: + log.info(f"{context}: Restored training weights") + + def clip_grad_norm_( + self, + max_norm: float, + norm_type: float = 2.0, + error_if_nonfinite: bool = False, + foreach: Optional[bool] = None, + ): + return clip_grad_norm_( + self.net.parameters(), + max_norm, + norm_type=norm_type, + error_if_nonfinite=error_if_nonfinite, + foreach=foreach, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/utils.py b/REGEN-main/cosmos_policy/_src/predict2/models/utils.py new file mode 100644 index 0000000000000000000000000000000000000000..07d8839bdae4bcf3ef7563281351be3c771f9606 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/utils.py @@ -0,0 +1,209 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import hashlib +import os +from contextlib import contextmanager + +import torch +from safetensors.torch import load as safetensors_torch_load + +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io + + +@contextmanager +def init_weights_on_device(device=torch.device("meta"), include_buffers: bool = False): + old_register_parameter = torch.nn.Module.register_parameter + if include_buffers: + old_register_buffer = torch.nn.Module.register_buffer + + def register_empty_parameter(module, name, param): + old_register_parameter(module, name, param) + if param is not None: + param_cls = type(module._parameters[name]) + kwargs = module._parameters[name].__dict__ + kwargs["requires_grad"] = param.requires_grad + module._parameters[name] = param_cls(module._parameters[name].to(device), **kwargs) + + def register_empty_buffer(module, name, buffer, persistent=True): + old_register_buffer(module, name, buffer, persistent=persistent) + if buffer is not None: + module._buffers[name] = module._buffers[name].to(device) + + def patch_tensor_constructor(fn): + def wrapper(*args, **kwargs): + kwargs["device"] = device + return fn(*args, **kwargs) + + return wrapper + + if include_buffers: + tensor_constructors_to_patch = { + torch_function_name: getattr(torch, torch_function_name) + for torch_function_name in ["empty", "zeros", "ones", "full"] + } + else: + tensor_constructors_to_patch = {} + + try: + torch.nn.Module.register_parameter = register_empty_parameter + if include_buffers: + torch.nn.Module.register_buffer = register_empty_buffer + for torch_function_name in tensor_constructors_to_patch.keys(): + setattr(torch, torch_function_name, patch_tensor_constructor(getattr(torch, torch_function_name))) + yield + finally: + torch.nn.Module.register_parameter = old_register_parameter + if include_buffers: + torch.nn.Module.register_buffer = old_register_buffer + for torch_function_name, old_torch_function in tensor_constructors_to_patch.items(): + setattr(torch, torch_function_name, old_torch_function) + + +def load_state_dict_from_folder(file_path, torch_dtype=None): + state_dict = {} + for file_name in os.listdir(file_path): + if "." in file_name and file_name.split(".")[-1] in ["safetensors", "bin", "ckpt", "pth", "pt"]: + state_dict.update(load_state_dict(os.path.join(file_path, file_name), torch_dtype=torch_dtype)) + return state_dict + + +def load_state_dict(file_path, torch_dtype=None, s3_credential_path=None): + if file_path.endswith(".safetensors"): + return load_state_dict_from_safetensors( + file_path, torch_dtype=torch_dtype, s3_credential_path=s3_credential_path + ) + else: + return load_state_dict_from_bin(file_path, torch_dtype=torch_dtype, s3_credential_path=s3_credential_path) + + +def load_state_dict_from_safetensors(file_path, torch_dtype=None, s3_credential_path=None): + backend_args = ( + {"backend": "s3", "s3_credential_path": s3_credential_path} if file_path.startswith("s3://") else None + ) + state_dict = {} + byte_stream = easy_io.load(file_path, backend_args=backend_args, file_format="byte") + state_dict = safetensors_torch_load(byte_stream) + return state_dict + + +def load_state_dict_from_bin(file_path, torch_dtype=None, s3_credential_path=None): + backend_args = ( + {"backend": "s3", "s3_credential_path": s3_credential_path} if file_path.startswith("s3://") else None + ) + state_dict = easy_io.load( + file_path, backend_args=backend_args, file_format="pt", map_location="cpu", weights_only=False + ) + if torch_dtype is not None: + for i in state_dict: + if isinstance(state_dict[i], torch.Tensor): + state_dict[i] = state_dict[i].to(torch_dtype) + return state_dict + + +def search_for_embeddings(state_dict): + embeddings = [] + for k in state_dict: + if isinstance(state_dict[k], torch.Tensor): + embeddings.append(state_dict[k]) + elif isinstance(state_dict[k], dict): + embeddings += search_for_embeddings(state_dict[k]) + return embeddings + + +def search_parameter(param, state_dict): + for name, param_ in state_dict.items(): + if param.numel() == param_.numel(): + if param.shape == param_.shape: + if torch.dist(param, param_) < 1e-3: + return name + else: + if torch.dist(param.flatten(), param_.flatten()) < 1e-3: + return name + return None + + +def build_rename_dict(source_state_dict, target_state_dict, split_qkv=False): + matched_keys = set() + with torch.no_grad(): + for name in source_state_dict: + rename = search_parameter(source_state_dict[name], target_state_dict) + if rename is not None: + print(f'"{name}": "{rename}",') + matched_keys.add(rename) + elif split_qkv and len(source_state_dict[name].shape) >= 1 and source_state_dict[name].shape[0] % 3 == 0: + length = source_state_dict[name].shape[0] // 3 + rename = [] + for i in range(3): + rename.append( + search_parameter(source_state_dict[name][i * length : i * length + length], target_state_dict) + ) + if None not in rename: + print(f'"{name}": {rename},') + for rename_ in rename: + matched_keys.add(rename_) + for name in target_state_dict: + if name not in matched_keys: + print("Cannot find", name, target_state_dict[name].shape) + + +def search_for_files(folder, extensions): + files = [] + if os.path.isdir(folder): + for file in sorted(os.listdir(folder)): + files += search_for_files(os.path.join(folder, file), extensions) + elif os.path.isfile(folder): + for extension in extensions: + if folder.endswith(extension): + files.append(folder) + break + return files + + +def convert_state_dict_keys_to_single_str(state_dict, with_shape=True): + keys = [] + for key, value in state_dict.items(): + if isinstance(key, str): + if isinstance(value, torch.Tensor): + if with_shape: + shape = "_".join(map(str, list(value.shape))) + keys.append(key + ":" + shape) + keys.append(key) + elif isinstance(value, dict): + keys.append(key + "|" + convert_state_dict_keys_to_single_str(value, with_shape=with_shape)) + keys.sort() + keys_str = ",".join(keys) + return keys_str + + +def split_state_dict_with_prefix(state_dict): + keys = sorted([key for key in state_dict if isinstance(key, str)]) + prefix_dict = {} + for key in keys: + prefix = key if "." not in key else key.split(".")[0] + if prefix not in prefix_dict: + prefix_dict[prefix] = [] + prefix_dict[prefix].append(key) + state_dicts = [] + for prefix, keys in prefix_dict.items(): + sub_state_dict = {key: state_dict[key] for key in keys} + state_dicts.append(sub_state_dict) + return state_dicts + + +def hash_state_dict_keys(state_dict, with_shape=True): + keys_str = convert_state_dict_keys_to_single_str(state_dict, with_shape=with_shape) + keys_str = keys_str.encode(encoding="UTF-8") + return hashlib.md5(keys_str).hexdigest() diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/video2world_model.py b/REGEN-main/cosmos_policy/_src/predict2/models/video2world_model.py new file mode 100644 index 0000000000000000000000000000000000000000..b3230df9676bd48a55380b04e2ae4093f8014e87 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/video2world_model.py @@ -0,0 +1,335 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import math +from enum import Enum +from typing import Any, Callable, Dict, Optional, Tuple + +import attrs +import torch +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor + +from cosmos_policy._src.imaginaire.utils.high_sigma_strategy import HighSigmaStrategy as HighSigmaStrategy +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.configs.video2world.defaults.conditioner import Video2WorldCondition +from cosmos_policy._src.predict2.models.text2world_model import ( + DenoisePrediction, + Text2WorldCondition, + Text2WorldModelConfig, +) +from cosmos_policy._src.predict2.models.text2world_model import DiffusionModel as Text2WorldModel + +NUM_CONDITIONAL_FRAMES_KEY: str = "num_conditional_frames" + + +class ConditioningStrategy(str, Enum): + FRAME_REPLACE = "frame_replace" # First few frames of the video are replaced with the conditional frames + + def __str__(self) -> str: + return self.value + + +@attrs.define(slots=False) +class Video2WorldConfig(Text2WorldModelConfig): + min_num_conditional_frames: int = 1 # Minimum number of latent conditional frames + max_num_conditional_frames: int = 2 # Maximum number of latent conditional frames + sigma_conditional: float = 0.0001 # Noise level used for conditional frames + conditioning_strategy: str = str(ConditioningStrategy.FRAME_REPLACE) # What strategy to use for conditioning + denoise_replace_gt_frames: bool = True # Whether to denoise the ground truth frames + high_sigma_strategy: str = str(HighSigmaStrategy.UNIFORM80_2000) # What strategy to use for high sigma + high_sigma_ratio: float = 0.05 # Ratio of high sigma frames + low_sigma_ratio: float = 0.05 # Ratio of low sigma frames + conditional_frames_probs: Optional[Dict[int, float]] = None # Probability distribution for conditional frames + + def __attrs_post_init__(self): + super().__attrs_post_init__() + assert self.conditioning_strategy in [ + str(ConditioningStrategy.FRAME_REPLACE), + ] + assert self.high_sigma_strategy in [ + str(HighSigmaStrategy.NONE), + str(HighSigmaStrategy.UNIFORM80_2000), + str(HighSigmaStrategy.LOGUNIFORM200_100000), + str(HighSigmaStrategy.BALANCED_TWO_HEADS_V1), + str(HighSigmaStrategy.SHIFT24), + str(HighSigmaStrategy.HARDCODED_20steps), + ] + + +LOG_200 = math.log(200) +LOG_100000 = math.log(100000) + + +class Video2WorldModel(Text2WorldModel): + def get_data_and_condition( + self, data_batch: dict[str, torch.Tensor] + ) -> Tuple[Tensor, Tensor, Video2WorldCondition]: + # generate random number of conditional frames for training + raw_state, latent_state, condition = super().get_data_and_condition(data_batch) + condition = condition.set_video_condition( + gt_frames=latent_state.to(**self.tensor_kwargs), + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=data_batch.get(NUM_CONDITIONAL_FRAMES_KEY, None), + conditional_frames_probs=self.config.conditional_frames_probs, + ) + return raw_state, latent_state, condition + + def draw_training_sigma_and_epsilon(self, x0_size: int, condition: Any) -> torch.Tensor: + sigma_B_1, epsilon = super().draw_training_sigma_and_epsilon(x0_size, condition) + is_video_batch = condition.data_type == DataType.VIDEO + # if is_video_batch, with 5% ratio, we regenerate sigma_B_1 with uniformally from 80 to 2000 + # with remaining 95% ratio, we keep the original sigma_B_1 + if is_video_batch: + if self.config.high_sigma_strategy == str(HighSigmaStrategy.UNIFORM80_2000): + mask = torch.rand(sigma_B_1.shape, device=sigma_B_1.device) < self.config.high_sigma_ratio + new_sigma = torch.rand(sigma_B_1.shape, device=sigma_B_1.device).type_as(sigma_B_1) * 1920 + 80 + sigma_B_1 = torch.where(mask, new_sigma, sigma_B_1) + elif self.config.high_sigma_strategy == str(HighSigmaStrategy.LOGUNIFORM200_100000): + mask = torch.rand(sigma_B_1.shape, device=sigma_B_1.device) < self.config.high_sigma_ratio + log_new_sigma = ( + torch.rand(sigma_B_1.shape, device=sigma_B_1.device).type_as(sigma_B_1) * (LOG_100000 - LOG_200) + + LOG_200 + ) + sigma_B_1 = torch.where(mask, log_new_sigma.exp(), sigma_B_1) + elif self.config.high_sigma_strategy == str(HighSigmaStrategy.SHIFT24): + # sample t from uniform distribution between 0 and 1, with same shape as sigma_B_1 + _t = torch.rand(sigma_B_1.shape, device=sigma_B_1.device).double() + _t = 24 * _t / (24 * _t + 1 - _t) + sigma_B_1 = (_t / (1.0 - _t)).float() + + mask = torch.rand(sigma_B_1.shape, device=sigma_B_1.device) < self.config.high_sigma_ratio + new_sigma = torch.rand(sigma_B_1.shape, device=sigma_B_1.device).type_as(sigma_B_1) * 1920 + 80 + sigma_B_1 = torch.where(mask, new_sigma, sigma_B_1) + elif self.config.high_sigma_strategy == str(HighSigmaStrategy.BALANCED_TWO_HEADS_V1): + # replace high sigma parts + mask = torch.rand(sigma_B_1.shape, device=sigma_B_1.device) < self.config.high_sigma_ratio + log_new_sigma = ( + torch.rand(sigma_B_1.shape, device=sigma_B_1.device).type_as(sigma_B_1) * (LOG_100000 - LOG_200) + + LOG_200 + ) + sigma_B_1 = torch.where(mask, log_new_sigma.exp(), sigma_B_1) + # replace low sigma parts + mask = torch.rand(sigma_B_1.shape, device=sigma_B_1.device) < self.config.low_sigma_ratio + low_sigma_B_1 = torch.rand(sigma_B_1.shape, device=sigma_B_1.device).type_as(sigma_B_1) * 2.0 + 0.00001 + sigma_B_1 = torch.where(mask, low_sigma_B_1, sigma_B_1) + elif self.config.high_sigma_strategy == str(HighSigmaStrategy.HARDCODED_20steps): + if not hasattr(self, "hardcoded_20steps_sigma"): + from cosmos_policy._src.imaginaire.modules.res_sampler import get_rev_ts + + hardcoded_20steps_sigma = get_rev_ts( + t_min=self.sde.sigma_min, t_max=self.sde.sigma_max, num_steps=20, ts_order=7.0 + ) + # add extra 100000 to the beginning + self.hardcoded_20steps_sigma = torch.cat( + [torch.tensor([100000.0], device=hardcoded_20steps_sigma.device), hardcoded_20steps_sigma], + dim=0, + ) + sigma_B_1 = self.hardcoded_20steps_sigma[ + torch.randint(0, len(self.hardcoded_20steps_sigma), sigma_B_1.shape) + ].type_as(sigma_B_1) + elif self.config.high_sigma_strategy == str(HighSigmaStrategy.NONE): + pass + else: + raise ValueError(f"High sigma strategy {self.config.high_sigma_strategy} is not supported") + return sigma_B_1, epsilon + + def denoise_with_velocity( + self, noise_x_in_t_space: torch.Tensor, t_B_T: torch.Tensor, condition: Text2WorldCondition + ) -> torch.Tensor: + """ + This function is used when self.config.use_flowunipc_scheduler is set. + """ + if t_B_T.ndim == 1: + t_B_T = rearrange(t_B_T, "b -> b 1") + elif t_B_T.ndim == 2: + t_B_T = t_B_T + else: + raise ValueError(f"t_B_T shape {t_B_T.shape} is not supported") + # our model expects input of sigma and x_sigma, so convert t -> sigma, x_t to x_sigma + sigma_B_T = t_B_T / (1.0 - t_B_T) + x_B_C_T_H_W_in_sigma_space = noise_x_in_t_space * (1.0 + rearrange(sigma_B_T, "b t -> b 1 t 1 1")) + denoise_output_B_C_T_H_W = self.denoise(x_B_C_T_H_W_in_sigma_space, sigma_B_T, condition) + x0_pred_B_C_T_H_W = denoise_output_B_C_T_H_W.x0 + eps_pred_B_C_T_H_W = denoise_output_B_C_T_H_W.eps + return eps_pred_B_C_T_H_W - x0_pred_B_C_T_H_W + + def denoise( + self, xt_B_C_T_H_W: torch.Tensor, sigma: torch.Tensor, condition: Text2WorldCondition + ) -> DenoisePrediction: + """ + Performs denoising on the input noise data, noise level, and condition + + Args: + xt (torch.Tensor): The input noise data. + sigma (torch.Tensor): The noise level. + condition (Text2WorldCondition): conditional information, generated from self.conditioner + + Returns: + DenoisePrediction: The denoised prediction, it includes clean data predicton (x0), \ + noise prediction (eps_pred). + """ + + if sigma.ndim == 1: + sigma_B_T = rearrange(sigma, "b -> b 1") + elif sigma.ndim == 2: + sigma_B_T = sigma + else: + raise ValueError(f"sigma shape {sigma.shape} is not supported") + + sigma_B_1_T_1_1 = rearrange(sigma_B_T, "b t -> b 1 t 1 1") + # get precondition for the network + c_skip_B_1_T_1_1, c_out_B_1_T_1_1, c_in_B_1_T_1_1, c_noise_B_1_T_1_1 = self.scaling(sigma=sigma_B_1_T_1_1) + + net_state_in_B_C_T_H_W = xt_B_C_T_H_W * c_in_B_1_T_1_1 + + if condition.is_video: + condition_state_in_B_C_T_H_W = condition.gt_frames.type_as(net_state_in_B_C_T_H_W) / self.config.sigma_data + if not condition.use_video_condition: + # When using random dropout, we zero out the ground truth frames + condition_state_in_B_C_T_H_W = condition_state_in_B_C_T_H_W * 0 + + _, C, _, _, _ = xt_B_C_T_H_W.shape + condition_video_mask = condition.condition_video_input_mask_B_C_T_H_W.repeat(1, C, 1, 1, 1).type_as( + net_state_in_B_C_T_H_W + ) + + # Replace the first few frames of the video with the conditional frames + # Update the c_noise as the conditional frames are clean and have very low noise + + # Make the first few frames of x_t be the ground truth frames + net_state_in_B_C_T_H_W = condition_state_in_B_C_T_H_W * condition_video_mask + net_state_in_B_C_T_H_W * ( + 1 - condition_video_mask + ) + # Adjust c_noise for the conditional frames + sigma_cond_B_1_T_1_1 = torch.ones_like(sigma_B_1_T_1_1) * self.config.sigma_conditional + _, _, _, c_noise_cond_B_1_T_1_1 = self.scaling(sigma=sigma_cond_B_1_T_1_1) + condition_video_mask_B_1_T_1_1 = condition_video_mask.mean(dim=[1, 3, 4], keepdim=True) + c_noise_B_1_T_1_1 = c_noise_cond_B_1_T_1_1 * condition_video_mask_B_1_T_1_1 + c_noise_B_1_T_1_1 * ( + 1 - condition_video_mask_B_1_T_1_1 + ) + + # forward pass through the network + net_output_B_C_T_H_W = self.net( + x_B_C_T_H_W=net_state_in_B_C_T_H_W.to( + **self.tensor_kwargs + ), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + timesteps_B_T=c_noise_B_1_T_1_1.squeeze(dim=[1, 3, 4]).to( + **{ + **self.tensor_kwargs, + "dtype": torch.float32 if self.config.use_wan_fp32_strategy else self.tensor_kwargs["dtype"], + }, + ), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + **condition.to_dict(), + ).float() + + x0_pred_B_C_T_H_W = c_skip_B_1_T_1_1 * xt_B_C_T_H_W + c_out_B_1_T_1_1 * net_output_B_C_T_H_W + if condition.is_video and self.config.denoise_replace_gt_frames: + # Set the first few frames to the ground truth frames. This will ensure that the loss is not computed for the first few frames. + x0_pred_B_C_T_H_W = condition.gt_frames.type_as( + x0_pred_B_C_T_H_W + ) * condition_video_mask + x0_pred_B_C_T_H_W * (1 - condition_video_mask) + + # get noise prediction based on sde + eps_pred_B_C_T_H_W = (xt_B_C_T_H_W - x0_pred_B_C_T_H_W) / sigma_B_1_T_1_1 + + return DenoisePrediction(x0_pred_B_C_T_H_W, eps_pred_B_C_T_H_W, None) + + def get_x0_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `x0_fn` based on the provided data batch and guidance factor. + + This function first processes the input data batch through a conditioning workflow (`conditioner`) to obtain conditioned and unconditioned states. It then defines a nested function `x0_fn` which applies a denoising operation on an input `noise_x` at a given noise level `sigma` using both the conditioned and unconditioned states. + + Args: + - data_batch (Dict): A batch of data used for conditioning. The format and content of this dictionary should align with the expectations of the `self.conditioner` + - guidance (float, optional): A scalar value that modulates the influence of the conditioned state relative to the unconditioned state in the output. Defaults to 1.5. + - is_negative_prompt (bool): use negative prompt t5 in uncondition if true + + Returns: + - Callable: A function `x0_fn(noise_x, sigma)` that takes two arguments, `noise_x` and `sigma`, and return x0 predictoin + + The returned function is suitable for use in scenarios where a denoised state is required based on both conditioned and unconditioned inputs, with an adjustable level of guidance influence. + """ + + if NUM_CONDITIONAL_FRAMES_KEY in data_batch: + num_conditional_frames = data_batch[NUM_CONDITIONAL_FRAMES_KEY] + else: + num_conditional_frames = 1 + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + is_image_batch = self.is_image_batch(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + _, x0, _ = self.get_data_and_condition(data_batch) + # override condition with inference mode; num_conditional_frames used Here! + condition = condition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + conditional_frames_probs=self.config.conditional_frames_probs, + ) + uncondition = uncondition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + conditional_frames_probs=self.config.conditional_frames_probs, + ) + condition = condition.edit_for_inference(is_cfg_conditional=True, num_conditional_frames=num_conditional_frames) + uncondition = uncondition.edit_for_inference( + is_cfg_conditional=False, num_conditional_frames=num_conditional_frames + ) + + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def x0_fn(noise_x: torch.Tensor, sigma: torch.Tensor) -> torch.Tensor: + if self.config.use_flowunipc_scheduler: + cond_velocity = self.denoise_with_velocity(noise_x, sigma, condition) + uncond_velocity = self.denoise_with_velocity(noise_x, sigma, uncondition) + velocity = uncond_velocity + guidance * (cond_velocity - uncond_velocity) + return velocity + cond_x0 = self.denoise(noise_x, sigma, condition).x0 + uncond_x0 = self.denoise(noise_x, sigma, uncondition).x0 + raw_x0 = cond_x0 + guidance * (cond_x0 - uncond_x0) + if "guided_image" in data_batch: + # replacement trick that enables inpainting with base model + assert "guided_mask" in data_batch, "guided_mask should be in data_batch if guided_image is present" + guide_image = data_batch["guided_image"] + guide_mask = data_batch["guided_mask"] + raw_x0 = guide_mask * guide_image + (1 - guide_mask) * raw_x0 + return raw_x0 + + return x0_fn diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/video2world_model_rectified_flow.py b/REGEN-main/cosmos_policy/_src/predict2/models/video2world_model_rectified_flow.py new file mode 100644 index 0000000000000000000000000000000000000000..c8d70c82198daf9a7240e603d7dd50f16d86d209 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/video2world_model_rectified_flow.py @@ -0,0 +1,346 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from enum import Enum +from typing import Callable, Dict, Literal, Optional, Tuple + +import attrs +import torch +from einops import rearrange +from megatron.core import parallel_state +from torch import Tensor + +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.configs.video2world.defaults.conditioner import Video2WorldCondition +from cosmos_policy._src.predict2.models.denoise_prediction import DenoisePrediction +from cosmos_policy._src.predict2.models.text2world_model_rectified_flow import ( + Text2WorldCondition, + Text2WorldModelRectifiedFlow, + Text2WorldModelRectifiedFlowConfig, +) + +NUM_CONDITIONAL_FRAMES_KEY: str = "num_conditional_frames" + + +class ConditioningStrategy(str, Enum): + FRAME_REPLACE = "frame_replace" # First few frames of the video are replaced with the conditional frames + + def __str__(self) -> str: + return self.value + + +@attrs.define(slots=False) +class Video2WorldModelRectifiedFlowConfig(Text2WorldModelRectifiedFlowConfig): + min_num_conditional_frames: int = 1 # Minimum number of latent conditional frames + max_num_conditional_frames: int = 2 # Maximum number of latent conditional frames + conditional_frame_timestep: float = ( + -1.0 + ) # Noise level used for conditional frames; default is -1 which will not take effective + conditioning_strategy: str = str(ConditioningStrategy.FRAME_REPLACE) # What strategy to use for conditioning + denoise_replace_gt_frames: bool = True # Whether to denoise the ground truth frames + conditional_frames_probs: Optional[Dict[int, float]] = None # Probability distribution for conditional frames + + def __attrs_post_init__(self): + super().__attrs_post_init__() + assert self.conditioning_strategy in [ + str(ConditioningStrategy.FRAME_REPLACE), + ] + + +class Video2WorldModelRectifiedFlow(Text2WorldModelRectifiedFlow): + def get_data_and_condition( + self, data_batch: dict[str, torch.Tensor] + ) -> Tuple[Tensor, Tensor, Video2WorldCondition]: + # generate random number of conditional frames for training + raw_state, latent_state, condition = super().get_data_and_condition(data_batch) + condition = condition.set_video_condition( + gt_frames=latent_state.to(**self.tensor_kwargs), + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=data_batch.get(NUM_CONDITIONAL_FRAMES_KEY, None), + conditional_frames_probs=self.config.conditional_frames_probs, + ) + return raw_state, latent_state, condition + + def denoise( + self, + noise: torch.Tensor, + xt_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + condition: Text2WorldCondition, + ) -> DenoisePrediction: + """ + Args: + xt (torch.Tensor): The input noise data. + sigma (torch.Tensor): The noise level. + condition (Text2WorldCondition): conditional information, generated from self.conditioner + + Returns: + velocity prediction + """ + if condition.is_video: + condition_state_in_B_C_T_H_W = condition.gt_frames.type_as(xt_B_C_T_H_W) + if not condition.use_video_condition: + # When using random dropout, we zero out the ground truth frames + condition_state_in_B_C_T_H_W = condition_state_in_B_C_T_H_W * 0 + + _, C, _, _, _ = xt_B_C_T_H_W.shape + condition_video_mask = condition.condition_video_input_mask_B_C_T_H_W.repeat(1, C, 1, 1, 1).type_as( + xt_B_C_T_H_W + ) + + # Make the first few frames of x_t be the ground truth frames + xt_B_C_T_H_W = condition_state_in_B_C_T_H_W * condition_video_mask + xt_B_C_T_H_W * ( + 1 - condition_video_mask + ) + + if self.config.conditional_frame_timestep >= 0: + condition_video_mask_B_1_T_1_1 = condition_video_mask.mean(dim=[1, 3, 4], keepdim=True) + timestep_cond_B_1_T_1_1 = ( + torch.ones_like(condition_video_mask_B_1_T_1_1) * self.config.conditional_frame_timestep + ) + + timesteps_B_1_T_1_1 = timestep_cond_B_1_T_1_1 * condition_video_mask_B_1_T_1_1 + timesteps_B_T * ( + 1 - condition_video_mask_B_1_T_1_1 + ) + + timesteps_B_T = timesteps_B_1_T_1_1.squeeze() + timesteps_B_T = ( + timesteps_B_T.unsqueeze(0) if timesteps_B_T.ndim == 1 else timesteps_B_T + ) # add dimension for batch + + # forward pass through the network + net_output_B_C_T_H_W = self.net( + x_B_C_T_H_W=xt_B_C_T_H_W.to(**self.tensor_kwargs), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + timesteps_B_T=timesteps_B_T, # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + **condition.to_dict(), + ).float() + + if condition.is_video and self.config.denoise_replace_gt_frames: + gt_frames_x0 = condition.gt_frames.type_as(net_output_B_C_T_H_W) + gt_frames_velocity = noise - gt_frames_x0 + net_output_B_C_T_H_W = gt_frames_velocity * condition_video_mask + net_output_B_C_T_H_W * ( + 1 - condition_video_mask + ) + + return net_output_B_C_T_H_W + + def get_velocity_fn_from_batch( + self, + data_batch: Dict, + guidance: float = 1.5, + is_negative_prompt: bool = False, + ) -> Callable: + """ + Generates a callable function `x0_fn` based on the provided data batch and guidance factor. + + This function first processes the input data batch through a conditioning workflow (`conditioner`) to obtain conditioned and unconditioned states. It then defines a nested function `x0_fn` which applies a denoising operation on an input `noise_x` at a given noise level `sigma` using both the conditioned and unconditioned states. + + Args: + - data_batch (Dict): A batch of data used for conditioning. The format and content of this dictionary should align with the expectations of the `self.conditioner` + - guidance (float, optional): A scalar value that modulates the influence of the conditioned state relative to the unconditioned state in the output. Defaults to 1.5. + - is_negative_prompt (bool): use negative prompt t5 in uncondition if true + + Returns: + - Callable: A function `x0_fn(noise_x, sigma)` that takes two arguments, `noise_x` and `sigma`, and return velocity predictoin + + The returned function is suitable for use in scenarios where a denoised state is required based on both conditioned and unconditioned inputs, with an adjustable level of guidance influence. + """ + + if NUM_CONDITIONAL_FRAMES_KEY in data_batch: + num_conditional_frames = data_batch[NUM_CONDITIONAL_FRAMES_KEY] + else: + num_conditional_frames = 1 + + if is_negative_prompt: + condition, uncondition = self.conditioner.get_condition_with_negative_prompt(data_batch) + else: + condition, uncondition = self.conditioner.get_condition_uncondition(data_batch) + + is_image_batch = self.is_image_batch(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + uncondition = uncondition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + _, x0, _ = self.get_data_and_condition(data_batch) + # override condition with inference mode; num_conditional_frames used Here! + condition = condition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + conditional_frames_probs=self.config.conditional_frames_probs, + ) + uncondition = uncondition.set_video_condition( + gt_frames=x0, + random_min_num_conditional_frames=self.config.min_num_conditional_frames, + random_max_num_conditional_frames=self.config.max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + conditional_frames_probs=self.config.conditional_frames_probs, + ) + condition = condition.edit_for_inference(is_cfg_conditional=True, num_conditional_frames=num_conditional_frames) + uncondition = uncondition.edit_for_inference( + is_cfg_conditional=False, num_conditional_frames=num_conditional_frames + ) + + _, condition, _, _ = self.broadcast_split_for_model_parallelsim(x0, condition, None, None) + _, uncondition, _, _ = self.broadcast_split_for_model_parallelsim(x0, uncondition, None, None) + + if parallel_state.is_initialized(): + pass + else: + assert not self.net.is_context_parallel_enabled, ( + "parallel_state is not initialized, context parallel should be turned off." + ) + + def velocity_fn(noise: torch.Tensor, noise_x: torch.Tensor, timestep: torch.Tensor) -> torch.Tensor: + cond_v = self.denoise(noise, noise_x, timestep, condition) + uncond_v = self.denoise(noise, noise_x, timestep, uncondition) + velocity_pred = cond_v + guidance * (cond_v - uncond_v) + return velocity_pred + + return velocity_fn + + def denoise_edm( + self, + xt_B_C_T_H_W: torch.Tensor, + time: torch.Tensor, + condition: Video2WorldCondition, + net_type: Literal["teacher", "fake_score", "student"] = "teacher", + ) -> DenoisePrediction: + """ + Network forward to denoise the input noised data given noise level, and condition. + + Assumes EDM-scaling parameterization. + + Compared to base class denoise function, this function supports different net types: + - fake_score: the fake score net on student generator's outputs + - student: the student net (few-step generator) + + Args: + xt (torch.Tensor): The input noise data. + time (torch.Tensor): The noise level under TrigFlow parameterization. + condition (Video2WorldCondition): conditional information, generated from self.conditioner + + Returns: + DenoisePrediction: The denoised prediction, it includes clean data predicton (x0), \ + noise prediction (eps_pred). + """ + if time.ndim == 1: + time_B_T = rearrange(time, "b -> b 1") + elif time.ndim == 2: + time_B_T = time + else: + raise ValueError(f"time shape {time.shape} is not supported") + time_B_1_T_1_1 = rearrange(time_B_T, "b t -> b 1 t 1 1") + + if condition.is_video: + # replace the noise level of the cond frames to be the pre-defined conditional noise level (very low) + # the scaling coefficients computed later will inherit the setting. + _, C, _, _, _ = xt_B_C_T_H_W.shape + condition_video_mask = condition.condition_video_input_mask_B_C_T_H_W.repeat(1, C, 1, 1, 1).type_as( + xt_B_C_T_H_W + ) + condition_video_mask_B_1_T_1_1 = condition_video_mask.mean(dim=[1, 3, 4], keepdim=True).type_as( + time_B_1_T_1_1 + ) # (B,1,T,1,1) + t_cond = torch.atan(torch.ones_like(time_B_1_T_1_1) * (self.sigma_conditional / self.sigma_data)) + time_B_1_T_1_1 = t_cond * condition_video_mask_B_1_T_1_1 + time_B_1_T_1_1 * ( + 1 - condition_video_mask_B_1_T_1_1 + ) + + # convert noise level time to EDM-formulation coefficients + c_skip_B_1_T_1_1, c_out_B_1_T_1_1, c_in_B_1_T_1_1, c_noise_B_1_T_1_1 = self.scaling_from_time(time_B_1_T_1_1) + + # EDM preconditioning + net_state_in_B_C_T_H_W = xt_B_C_T_H_W * c_in_B_1_T_1_1 + + if net_type == "student" and self.change_time_embed: + # Use c_noise(t)=t to improve numerical stability + c_noise_B_1_T_1_1 = time_B_1_T_1_1 + + net = self.net + + # Apply vid2vid conditioning + if condition.is_video: + condition_state_in_B_C_T_H_W = condition.gt_frames.type_as(net_state_in_B_C_T_H_W) / self.sigma_data + # during training we temporarily concat some variables (e.g. x0 and G_x0) + # into a batch. They are from the same data sample, so their use_video_condition + # is a boolean tensor with batch dim; their bool value should be the same. + use_video_cond = condition.use_video_condition + if isinstance(use_video_cond, torch.Tensor): + assert bool((use_video_cond == use_video_cond[0]).all().item()), ( + "inconsistent use_video_condition in concatenated batch" + ) + use_video_cond = bool(use_video_cond[0].item()) + if not use_video_cond: + # When using random dropout, we zero out the ground truth frames + condition_state_in_B_C_T_H_W = condition_state_in_B_C_T_H_W * 0 + + _, C, _, _, _ = xt_B_C_T_H_W.shape + condition_video_mask = condition.condition_video_input_mask_B_C_T_H_W.repeat(1, C, 1, 1, 1).type_as( + net_state_in_B_C_T_H_W + ) + + # Replace the first few frames of the video with the conditional frames + # Update the c_noise as the conditional frames are clean and have very low noise + + # x_in = mask*GT + (1-mask)*x; tangent passes only through the (1-mask) branch + net_state_in_B_C_T_H_W = condition_state_in_B_C_T_H_W * condition_video_mask + net_state_in_B_C_T_H_W * ( + 1 - condition_video_mask + ) + + call_kwargs = dict( + x_B_C_T_H_W=net_state_in_B_C_T_H_W.to( + **self.tensor_kwargs + ), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + timesteps_B_T=c_noise_B_1_T_1_1.squeeze(dim=[1, 3, 4]).to( + **self.tensor_kwargs + ), # Eq. 7 of https://arxiv.org/pdf/2206.00364.pdf + **condition.to_dict(), + ) + if net_type == "fake_score" and getattr(self, "intermediate_feature_ids", None): + call_kwargs["intermediate_feature_ids"] = self.intermediate_feature_ids + + # forward pass through the network + net_out = net(**call_kwargs) + + if net_type == "fake_score" and getattr(self, "intermediate_feature_ids", None): + net_output_B_C_T_H_W, intermediate_features_outputs = net_out + net_output_B_C_T_H_W = net_output_B_C_T_H_W.float() + else: + net_output_B_C_T_H_W = net_out.float() + intermediate_features_outputs = [] + + net_output_B_C_T_H_W = net_output_B_C_T_H_W.to(dtype=xt_B_C_T_H_W.dtype) + + # EDM reconstruction of x0 + x0_pred_B_C_T_H_W = c_skip_B_1_T_1_1 * xt_B_C_T_H_W + c_out_B_1_T_1_1 * net_output_B_C_T_H_W + + # Replace GT on conditioned frames to avoid training on pinned frames (parity with base Video2WorldModel) + if getattr(self.config, "denoise_replace_gt_frames", False) and condition.is_video: + # Replace condition frames to be gt frames to zero out loss on these frames + gt_frames = condition.gt_frames.type_as(x0_pred_B_C_T_H_W) + x0_pred_B_C_T_H_W = gt_frames * condition_video_mask.type_as(x0_pred_B_C_T_H_W) + x0_pred_B_C_T_H_W * ( + 1 - condition_video_mask + ) + + if net_type == "fake_score": + return DenoisePrediction(x0=x0_pred_B_C_T_H_W, intermediate_features=intermediate_features_outputs) + else: # student and teacher need F + F_pred_B_C_T_H_W = (torch.cos(time_B_1_T_1_1) * xt_B_C_T_H_W - x0_pred_B_C_T_H_W) / ( + torch.sin(time_B_1_T_1_1) * self.sigma_data + ) + return DenoisePrediction( + x0=x0_pred_B_C_T_H_W, F=F_pred_B_C_T_H_W, intermediate_features=intermediate_features_outputs + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/models/video2world_wan2pt1_model.py b/REGEN-main/cosmos_policy/_src/predict2/models/video2world_wan2pt1_model.py new file mode 100644 index 0000000000000000000000000000000000000000..a1b27fdb17e74890de42c003266fdcac0c869f9e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/models/video2world_wan2pt1_model.py @@ -0,0 +1,54 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Tuple + +import torch +from torch import Tensor + +from cosmos_policy._src.predict2.models.text2world_model import Text2WorldCondition +from cosmos_policy._src.predict2.models.text2world_wan2pt1_model import ( + DataType, + Text2WorldModelWan2pt1Config, + WANDiffusionModel, +) + +WAN2PT1_I2V_COND_LATENT_KEY = "i2v_WAN2PT1_cond_latents" # gitleaks:allow + + +class I2VWan2pt1Model(WANDiffusionModel): + def __init__(self, config: Text2WorldModelWan2pt1Config): + # Note that I2V config.shift has better value {"480p": 3.0, "720p": 5.0} + super().__init__(config) + + def get_data_and_condition(self, data_batch: dict[str, torch.Tensor]) -> Tuple[Tensor, Text2WorldCondition]: + self._normalize_video_databatch_inplace(data_batch) + self._augment_image_dim_inplace(data_batch) + is_image_batch = self.is_image_batch(data_batch) + + # Latent state + raw_state = data_batch[self.input_image_key if is_image_batch else self.input_data_key] + latent_state = self.encode(raw_state).contiguous().float() + if WAN2PT1_I2V_COND_LATENT_KEY not in data_batch: + conditional_content = torch.zeros_like(raw_state).to(**self.tensor_kwargs) + if not is_image_batch: + conditional_content[:, :, 0] = raw_state[:, :, 0] + + data_batch[WAN2PT1_I2V_COND_LATENT_KEY] = self.encode(conditional_content).contiguous() + + # Condition + condition = self.conditioner(data_batch) + condition = condition.edit_data_type(DataType.IMAGE if is_image_batch else DataType.VIDEO) + return raw_state, latent_state, condition diff --git a/REGEN-main/cosmos_policy/_src/predict2/modules/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/modules/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/modules/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/modules/denoiser_scaling.py b/REGEN-main/cosmos_policy/_src/predict2/modules/denoiser_scaling.py new file mode 100644 index 0000000000000000000000000000000000000000..1ca8c3196ea045e3a0eca78ebabc745e459a0f96 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/modules/denoiser_scaling.py @@ -0,0 +1,67 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Conversion from the TrigFlow (sCM paper) parameterization trigflow_t to +the four c_xxx scaling coefficients as in EDM fomulation. +""" + +from typing import Tuple + +import torch + + +# xt (under TrigFlow) = cost*x0/sigma_d + sint*eps +# xt' = x0 + sigma*eps +class EDM_sCMWrapper: + def __init__(self, sigma_data: float = 1.0): + self.sigma_data = sigma_data + + def __call__(self, trigflow_t: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]: + dtype = trigflow_t.dtype + trigflow_t = trigflow_t.to(torch.float64) + sigma = torch.tan(trigflow_t) * self.sigma_data + # c_skip = self.sigma_data**2 / (sigma**2 + self.sigma_data**2) + # c_out = sigma * self.sigma_data / (sigma**2 + self.sigma_data**2) ** 0.5 + # c_in = 1 / (sigma**2 + self.sigma_data**2) ** 0.5 + c_skip = self.sigma_data * torch.cos(trigflow_t) + c_out = self.sigma_data * torch.sin(trigflow_t) + c_in = torch.ones_like(trigflow_t) + c_noise = 0.25 * sigma.log() + return c_skip.to(dtype), c_out.to(dtype), c_in.to(dtype), c_noise.to(dtype) + + +class RectifiedFlow_sCMWrapper: + def __init__(self, sigma_data: float = 1.0): + self.sigma_data = sigma_data + + def __call__(self, trigflow_t: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]: + dtype = trigflow_t.dtype + trigflow_t = trigflow_t.to(torch.float64) + # sigma = torch.tan(trigflow_t) * self.sigma_data + # t = sigma / (sigma + 1) + # c_skip = 1.0 - t + # c_out = -t + # c_in = 1.0 - t + # c_noise = t + c_skip = self.sigma_data / (torch.cos(trigflow_t) + self.sigma_data * torch.sin(trigflow_t)) + c_out = ( + -self.sigma_data * torch.sin(trigflow_t) / (torch.cos(trigflow_t) + self.sigma_data * torch.sin(trigflow_t)) + ) + c_in = self.sigma_data / (torch.cos(trigflow_t) + self.sigma_data * torch.sin(trigflow_t)) + c_noise = ( + self.sigma_data * torch.sin(trigflow_t) / (torch.cos(trigflow_t) + self.sigma_data * torch.sin(trigflow_t)) + ) + return c_skip.to(dtype), c_out.to(dtype), c_in.to(dtype), c_noise.to(dtype) diff --git a/REGEN-main/cosmos_policy/_src/predict2/modules/neighborhood_attn.py b/REGEN-main/cosmos_policy/_src/predict2/modules/neighborhood_attn.py new file mode 100644 index 0000000000000000000000000000000000000000..bab5afab71e5b9fde05ce5136c78952e45dac6ba --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/modules/neighborhood_attn.py @@ -0,0 +1,248 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from collections import namedtuple +from collections.abc import Mapping, Sequence +from typing import Optional + +import torch +from torch import nn + +from cosmos_policy._src.imaginaire.utils import log + +try: + import natten + from natten.functional import neighborhood_attention_generic + + natten.use_kv_parallelism_in_fused_na(True) + natten.set_memory_usage_preference("unrestricted") + + HAS_NATTEN = True + +except ImportError: + HAS_NATTEN = False + + def neighborhood_attention_generic(*args, **kwargs): + raise RuntimeError( + "You attempted to run Cosmos-Predict2 + NATTEN, but NATTEN is not installed. " + "Refer to natten.org/install for install instructions, or use " + "the cosmos-predict2 container image." + ) + + +from cosmos_policy._src.predict2.networks.attention import get_device_cc + +VideoSize = namedtuple("VideoSize", ["T", "H", "W"]) + + +# Only allowing on Hopper and Blackwell for now, since Hopper FNA and +# Blackwell FNA can deliver excellent speedup over SOL baselines. +# Other architectures will be enabled as soon as good kernels for them +# land in NATTEN. +ALLOWED_COMPUTE_CAPS = [80, 90, 100] + + +class NeighborhoodAttention(nn.Module): + def __init__(self, natten_parameters, base_attn_op): + super(NeighborhoodAttention, self).__init__() + + self.base_attn_op = base_attn_op + + self.natten_parameters = natten_parameters + if ( + not isinstance(natten_parameters, Mapping) + or "window_size" not in natten_parameters + or "layer_id" not in natten_parameters + ): + raise ValueError( + f"Expected `natten_parameters` to be a dict with at least keys `window_size` and `layer_id`, got {natten_parameters=}." + ) + + self.layer_id = natten_parameters["layer_id"] + self.window_size = natten_parameters["window_size"] + self.stride = 1 if "stride" not in natten_parameters else natten_parameters["stride"] + self.dilation = 1 if "dilation" not in natten_parameters else natten_parameters["dilation"] + self.is_causal = False if "is_causal" not in natten_parameters else natten_parameters["is_causal"] + + if not isinstance(self.window_size, Sequence) or len(self.window_size) != 3: + raise ValueError(f"Invalid window_size value. Expected an iterable of length 3, got {self.window_size}.") + + if (not isinstance(self.stride, Sequence) or len(self.stride) != 3) and not isinstance(self.stride, int): + raise ValueError(f"Invalid stride value. Expected an iterable of length 3, or integer, got {self.stride}.") + + if (not isinstance(self.dilation, Sequence) or len(self.dilation) != 3) and not isinstance(self.dilation, int): + raise ValueError( + f"Invalid dilation value. Expected an iterable of length 3, or integer, got {self.dilation}." + ) + + if (not isinstance(self.is_causal, Sequence) or len(self.is_causal) != 3) and not isinstance( + self.is_causal, bool + ): + raise ValueError( + f"Invalid is_causal value. Expected an iterable of length 3, or boolean, got {self.is_causal}." + ) + + log.info( + f"NeighborhoodAttention op registered for layer {self.layer_id}, with " + f"window_size={self.window_size}, stride={self.stride}, dilation={self.dilation}." + ) + + self.base_size = None if "base_size" not in natten_parameters else natten_parameters["base_size"] + if self.base_size is not None and (not isinstance(self.base_size, Sequence) or len(self.base_size) != 3): + raise ValueError( + f"Invalid base feature map size. Expected an iterable of length 3, or None, got {self.base_size}." + ) + + # Configurations + # Tuned for 720p and window sizes (24, 12, 24), (16, 12, 24), and stride (1, 4, 8). + # They also assume head dim = 128. + self.performance_configs = { + # Ampere (SM80). Also serves as the default option for RTX cards (Ampere RTX, Ada, Blackwell RTX.) + 80: { + "backend": "cutlass-fna", + "q_tile_shape": (4, 4, 4), + "kv_tile_shape": (4, 4, 8), + "backward_q_tile_shape": (4, 4, 8), + "backward_kv_tile_shape": (4, 4, 8), + "backward_use_pt_reduction": False, + }, + # Hopper (SM90) + 90: { + "backend": "hopper-fna", + "q_tile_shape": (4, 4, 8), + "kv_tile_shape": (4, 4, 8), + "backward_q_tile_shape": (4, 4, 4), + "backward_kv_tile_shape": (4, 4, 8), + }, + # Blackwell (SM100) + 100: { + "backend": "blackwell-fna", + "q_tile_shape": (8, 4, 8), + "kv_tile_shape": (4, 4, 8), + "backward_q_tile_shape": (4, 4, 8), + "backward_kv_tile_shape": (4, 4, 8), + "run_persistent_kernel": True, + }, + } + + def get_adaptive_parameters(self, window_size, stride, dilation, is_causal, input_shape, base_size=None): + window_size = tuple(w if w > 1 else x for x, w in zip(input_shape, window_size)) + stride = tuple(stride for _ in range(3)) if isinstance(stride, int) else tuple(x for x in stride) + dilation = tuple(dilation for _ in range(3)) if isinstance(dilation, int) else tuple(x for x in dilation) + is_causal = tuple(is_causal for _ in range(3)) if isinstance(is_causal, bool) else tuple(x for x in is_causal) + + # Scale window size and stride according to some base input size + # For example, if window size is (8, 8, 8), stride is (1, 2, 2), for a base + # input/feature map size of (16, 16, 16); then if the input feat map in this iteration + # has shape (8, 8, 8), we should use window size (4, 4, 4), and stride (1, 1, 1). + if base_size is not None: + base_shape = tuple(b if b > 0 else x for x, b in zip(input_shape, base_size)) + + scale = tuple(x / b for x, b in zip(input_shape, base_shape)) + + scaled_window_size = tuple(min(max(2, round(w * s)), x) for w, s, x in zip(window_size, scale, input_shape)) + scaled_stride = tuple(min(max(1, round(st * s)), w) for w, s, st in zip(scaled_window_size, scale, stride)) + + max_dilation = tuple(x // w for x, w in zip(input_shape, scaled_window_size)) + scaled_dilation = tuple( + min(max(1, round(d * s)), max_d) for d, s, max_d in zip(dilation, scale, max_dilation) + ) + + window_size = scaled_window_size + stride = scaled_stride + dilation = scaled_dilation + + assert all(x >= w * d for x, w, d in zip(input_shape, window_size, dilation)) + assert all(w >= s for w, s in zip(window_size, stride)) + assert all(isinstance(c, bool) for c in is_causal) + + return window_size, stride, dilation, is_causal + + def forward( + self, + q_B_L_H_D: torch.Tensor, + k_B_L_H_D: torch.Tensor, + v_B_L_H_D: torch.Tensor, + video_size: Optional[VideoSize] = None, + ): + if not (q_B_L_H_D.shape == k_B_L_H_D.shape == v_B_L_H_D.shape): + raise ValueError( + f"NATTEN requires QKV shapes to match, got {q_B_L_H_D.shape=}, {k_B_L_H_D.shape=}, {v_B_L_H_D.shape=}." + ) + + device = q_B_L_H_D.device + compute_cap = get_device_cc(device) + requires_grad = q_B_L_H_D.requires_grad or k_B_L_H_D.requires_grad or v_B_L_H_D.requires_grad + is_cuda = torch.cuda.is_available() and torch.version.cuda and device.type == "cuda" + + if not is_cuda: + raise NotImplementedError(f"Cosmos-Predict2 + NATTEN requires CUDA, tensors were on {device=}.") + + if compute_cap not in ALLOWED_COMPUTE_CAPS: + raise NotImplementedError( + "Cosmos-Predict2 + NATTEN is only allowed on devices with the following " + f"compute capabilities: {ALLOWED_COMPUTE_CAPS}, got {compute_cap}." + ) + + natten_configuration = None + assert 80 in self.performance_configs.keys() + if is_cuda and compute_cap in self.performance_configs.keys(): + natten_configuration = self.performance_configs[compute_cap] + elif is_cuda and compute_cap >= 80: + natten_configuration = self.performance_configs[80] + else: + raise ValueError(f"No NATTEN config found for this use case: {requires_grad=}, {is_cuda=}, {compute_cap=}.") + + batch, seqlen, heads, head_dim = q_B_L_H_D.shape + T, H, W = video_size + + if seqlen != T * H * W: + raise ValueError(f"Mismatch between seqlen and video_size dimensions; got {video_size=}, {seqlen=}.") + + if T > 1: + input_shape = (T, H, W) + + window_size, stride, dilation, is_causal = self.get_adaptive_parameters( + window_size=self.window_size, + stride=self.stride, + dilation=self.dilation, + is_causal=self.is_causal, + input_shape=input_shape, + base_size=self.base_size, + ) + + elif T == 1: + # Do self attention for image model; skip natten + return self.base_attn_op(q_B_L_H_D, k_B_L_H_D, v_B_L_H_D) + + else: + raise ValueError(f"Invalid dimension {T=}.") + + q = q_B_L_H_D.view(batch, *input_shape, heads, head_dim) + k = k_B_L_H_D.view(batch, *input_shape, heads, head_dim) + v = v_B_L_H_D.view(batch, *input_shape, heads, head_dim) + + out = neighborhood_attention_generic( + query=q, + key=k, + value=v, + kernel_size=window_size, + stride=stride, + dilation=dilation, + is_causal=is_causal, + **natten_configuration, + ) + + return out.view(batch, seqlen, heads, head_dim) diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/networks/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/a2a_cp.py b/REGEN-main/cosmos_policy/_src/predict2/networks/a2a_cp.py new file mode 100644 index 0000000000000000000000000000000000000000..273f5443445afb8a35da96de967ce3c9306cd7f0 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/a2a_cp.py @@ -0,0 +1,226 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# from neophilia; Author: Qsh (qsh.zh27@gmail.com) +# MIT License + +from typing import Any, Callable, List, Tuple, Union + +import torch +import torch.distributed as dist +from einops import rearrange +from torch import Tensor +from torch.distributed import ProcessGroup +from torch.nn import Module + +from cosmos_policy._src.predict2.modules.neighborhood_attn import NeighborhoodAttention +from cosmos_policy._src.predict2.networks.attention import attention + + +def post_all2all(local_seq_2_local_head, seq_world_size): + def post_func(input): + # b, s, n, h + if local_seq_2_local_head: + output = rearrange(input, "w bs seq h d -> bs (w seq) h d") + else: + output = rearrange(input, "w bs s h d -> bs s (w h) d", w=seq_world_size) + + return output + + return post_func + + +def single_all_to_all(input, local_seq_2_local_head, group, async_op=False): + seq_world_size = dist.get_world_size(group) + + # b, s, n, h + if local_seq_2_local_head: + bs, local_seq_len, num_total_head, head_dim = input.shape + assert num_total_head % seq_world_size == 0, ( + f"Number of heads ({num_total_head}) must be divisible by the sequence parallel size ({seq_world_size})!" + ) + input_t = rearrange( + input, "bs seq_len (w h) d -> w bs seq_len h d", w=seq_world_size, h=num_total_head // seq_world_size + ).contiguous() + post_all2all_fun = post_all2all(local_seq_2_local_head, seq_world_size) + else: + bs, global_seq_len, num_local_head, head_dim = input.shape + input_t = rearrange( + input, "bs (w s) h d -> w bs s h d", w=seq_world_size, s=global_seq_len // seq_world_size + ).contiguous() + post_all2all_fun = post_all2all(local_seq_2_local_head, seq_world_size) + + output = torch.empty_like(input_t) + dist.all_to_all_single(output, input_t, group=group, async_op=async_op) + + res = post_all2all_fun(output) + return res + + +def async_a2a_communicate( + a2a_inputs: Union[torch.Tensor, List[torch.Tensor]], + cp_size: int, + cp_group: ProcessGroup, + cp_stream: torch.cuda.Stream, + local_seq_2_local_head: bool, +) -> Union[torch.Tensor, List[torch.Tensor]]: + """ + A2A communication for context parallelism. best used in communicate qkv + Modified from Nvidia Transformer Engine. + """ + a2a_inputs = [a2a_inputs] if not isinstance(a2a_inputs, list) else a2a_inputs + a2a_outputs, a2a_reqs = [None] * len(a2a_inputs), [None] * len(a2a_inputs) + a2a_post_fns = [None] * len(a2a_inputs) + if local_seq_2_local_head: + for i in range(len(a2a_inputs) + 2): + if 0 < i < len(a2a_inputs) + 1: + a2a_outputs[i - 1] = torch.empty_like(a2a_inputs[i - 1]) + a2a_reqs[i - 1] = torch.distributed.all_to_all_single( + a2a_outputs[i - 1], a2a_inputs[i - 1], group=cp_group, async_op=True + ) + a2a_post_fns[i - 1] = post_all2all(local_seq_2_local_head, cp_size) + if i > 1: + with torch.cuda.stream(cp_stream): + a2a_reqs[i - 2].wait() + a2a_outputs[i - 2] = a2a_post_fns[i - 2](a2a_outputs[i - 2]) + if i < len(a2a_inputs): + a2a_inputs[i] = rearrange( + a2a_inputs[i], "bs seq_len (w h) d -> w bs seq_len h d", w=cp_size + ).contiguous() + else: + for i in range(len(a2a_inputs) + 2): + if 0 < i < len(a2a_inputs) + 1: + a2a_outputs[i - 1] = torch.empty_like(a2a_inputs[i - 1]) + a2a_reqs[i - 1] = torch.distributed.all_to_all_single( + a2a_outputs[i - 1], a2a_inputs[i - 1], group=cp_group, async_op=True + ) + a2a_post_fns[i - 1] = post_all2all(local_seq_2_local_head, cp_size) + if i < len(a2a_inputs): + a2a_inputs[i] = rearrange(a2a_inputs[i], "bs (w s) h d -> w bs s h d", w=cp_size).contiguous() + if i > 1: + with torch.cuda.stream(cp_stream): + a2a_reqs[i - 2].wait() + a2a_outputs[i - 2] = a2a_post_fns[i - 2](a2a_outputs[i - 2]) + torch.cuda.current_stream().wait_stream(cp_stream) + return a2a_outputs[0] if len(a2a_inputs) == 1 else a2a_outputs + + +class _SeqAllToAll(torch.autograd.Function): + @staticmethod + def forward(ctx: Any, group: dist.ProcessGroup, input: Tensor, local_seq_2_local_head: bool) -> Tensor: + ctx.group = group + res = single_all_to_all(input, local_seq_2_local_head, group, False) + ctx.local_seq_2_local_head = local_seq_2_local_head + return res + + @staticmethod + def backward(ctx: Any, *grad_output: Tensor) -> Tuple[None, Tensor, None]: + return (None, _SeqAllToAll.apply(ctx.group, *grad_output, not ctx.local_seq_2_local_head), None) + + +class _SeqAllToAllQKV(torch.autograd.Function): + @staticmethod + def forward( + ctx: Any, + group: dist.ProcessGroup, + q: Tensor, + k: Tensor, + v: Tensor, + cp_size: int, + cp_stream: torch.cuda.Stream, + local_seq_2_local_head: bool, + ) -> Tuple[Tensor, Tensor, Tensor]: + ctx.group = group + ctx.cp_size = cp_size + ctx.cp_stream = cp_stream + ctx.local_seq_2_local_head = local_seq_2_local_head + q, k, v = async_a2a_communicate([q, k, v], cp_size, group, cp_stream, local_seq_2_local_head) + return q, k, v + + @staticmethod + def backward(ctx: Any, *grad_output: Tensor) -> Tuple[None, Tensor, Tensor, Tensor, None, None, None]: + q_grad, k_grad, v_grad = _SeqAllToAllQKV.apply( + ctx.group, *grad_output, ctx.cp_size, ctx.cp_stream, not ctx.local_seq_2_local_head + ) + return (None, q_grad, k_grad, v_grad, None, None, None) + + +class DistributedAttention(torch.nn.Module): + """Initialization. + + Arguments: + local_attention (Module): local attention with q,k,v + sequence_process_group (ProcessGroup): sequence parallel process group + """ + + def __init__( + self, + local_attention: Union[Module, Callable], + ) -> None: + super(DistributedAttention, self).__init__() + self.local_attn = local_attention + self.pg = None + self.stream = None + + def forward(self, query: Tensor, key: Tensor, value: Tensor, *args: Any, **kwargs) -> Tensor: + """forward + + Arguments: + query (Tensor): query input to the layer + key (Tensor): key input to the layer + value (Tensor): value input to the layer + args: other args + + Returns: + * output (Tensor): context output + """ + if self.pg is None: + return self.local_attn(query, key, value, *args, **kwargs) + pg_size = dist.get_world_size(self.pg) + if pg_size < 2: + return self.local_attn(query, key, value, *args, **kwargs) + + query_layer, key_layer, value_layer = _SeqAllToAllQKV.apply( + self.pg, query, key, value, pg_size, self.stream, True + ) + context_layer = self.local_attn(query_layer, key_layer, value_layer, *args, **kwargs) + + output = _SeqAllToAll.apply(self.pg, context_layer, False) + return output + + def set_context_parallel_group(self, group, stream): + self.pg = group + self.stream = stream + + +class MinimalA2AAttnOp(DistributedAttention): + def __init__(self, *args, **kwargs): + del args, kwargs + super(MinimalA2AAttnOp, self).__init__(attention) + + def set_context_parallel_group(self, process_group, ranks, stream, cp_comm_type: str = "p2p"): + del ranks + super().set_context_parallel_group(process_group, stream) + + def forward(self, query: Tensor, key: Tensor, value: Tensor, *args: Any, **kwargs) -> Tensor: + results = super().forward(query, key, value, *args, **kwargs) + return rearrange(results, "b ... h l -> b ... (h l)") + + +class NattenA2AAttnOp(MinimalA2AAttnOp): + def __init__(self, *args, **kwargs): + super(NattenA2AAttnOp, self).__init__(None) + self.natten_op = NeighborhoodAttention(*args, **kwargs, base_attn_op=attention) + self.local_attn = self.natten_op diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/attention.py b/REGEN-main/cosmos_policy/_src/predict2/networks/attention.py new file mode 100644 index 0000000000000000000000000000000000000000..a8cfbdf66a50ac8b228505badd80014de7e3da31 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/attention.py @@ -0,0 +1,181 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# From Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved. + +# Description: +# Single point of entry for all generic attention ops (self and cross attention), that tries to +# deliver the best performance possible given any use case (GPU and environment). +# +# On Hopper GPUs (i.e. H100, H20, H200), Flash Attention 3 is the best-performing choice, but it +# needs to be installed. When it is not available, the second best choice is cuDNN attention, which +# we get using PyTorch's SDPA API. +# +# For all other use cases, we will just use PyTorch's SDPA, but we need to specify backends and +# priorities. +# Flash Attention 2, which is one of the backends, is the best choice for Ampere GPUs (both RTX and +# datacenter-class). +# +# For anything pre-Ampere, the only choice is "memory-efficient" (xformers) FMHA. +# +# For Ada and Blackwell RTX, it is unclear at the moment, so we defer to Flash Attention 2, and +# fallbacks are cuDNN and xformers. +# +# For Blackwell datacenter-class (B200, GB200), cuDNN is the best choice. +# +# +# Dispatching to the desired backends/paths are done by checking the compute capability (really SM +# number, which is just compute capability * 10) of the GPU device the input tensors are on. +# +# Here's a breakdown of relevant compute capabilities: +# +# | GPU / category | Arch | +# |================|=======| +# | A100 | SM80 | +# | A40 | SM80 | +# | Ampere RTX | SM86 | +# |----------------|-------| +# | Ada Lovelace | SM89 | +# |----------------|-------| +# | H20 | SM90 | +# | H100 | SM90 | +# | H200 | SM90 | +# |----------------|-------| +# | B200 | SM100 | +# | Blackwell RTX | SM103 | +# |----------------|-------| +# + +from functools import partial + +import torch +from torch.nn.attention import SDPBackend, sdpa_kernel + +try: + from flash_attn_3.flash_attn_interface import flash_attn_func + + FLASH_ATTN_3_AVAILABLE = True +except ModuleNotFoundError: + FLASH_ATTN_3_AVAILABLE = False + + +def get_device_cc(device) -> int: + """ + Returns the compute capability of a given torch device if it's a CUDA device, otherwise returns 0. + + Args: + device: torch device. + + Returns: + device_cc (int): compute capability in the SmXXX format (i.e. 90 for Hopper). + """ + if torch.cuda.is_available() and torch.version.cuda and device.type == "cuda": + major, minor = torch.cuda.get_device_capability(device) + return major * 10 + minor + return 0 + + +def attention( + q, + k, + v, + q_lens=None, + k_lens=None, + dropout_p=0.0, + softmax_scale=None, + q_scale=None, + causal=False, + deterministic=False, + dtype=torch.bfloat16, +): + supported_dtypes = [torch.bfloat16, torch.float16, torch.float32] + is_half = dtype in [torch.bfloat16, torch.float16] + compute_cap = get_device_cc(q.device) + + if dtype not in supported_dtypes: + raise NotImplementedError(f"{dtype=} is not supported.") + + q = q.to(dtype) + k = k.to(dtype) + v = v.to(dtype) + + if q_scale is not None: + q = q * q_scale + + # If Flash Attention 3 is installed, and the user's running on a Hopper GPU (compute capability + # 9.0, or SM90), use Flash Attention 3. + if compute_cap == 90 and FLASH_ATTN_3_AVAILABLE and is_half: + return flash_attn_func( + q=q, + k=k, + v=v, + softmax_scale=softmax_scale, + causal=causal, + deterministic=deterministic, + )[0] + else: + # If Blackwell or Hopper (SM100 or SM90), cuDNN has native FMHA kernels. The Hopper one is + # not always as fast as Flash Attention 3, but when Flash Attention is unavailable, it's + # still a far better choice than Flash Attention 2 (Ampere). + if compute_cap in [90, 100] and is_half: + SDPA_BACKENDS = [ + SDPBackend.CUDNN_ATTENTION, + SDPBackend.FLASH_ATTENTION, + SDPBackend.EFFICIENT_ATTENTION, + ] + BEST_SDPA_BACKEND = SDPBackend.CUDNN_ATTENTION + elif is_half: + SDPA_BACKENDS = [ + SDPBackend.FLASH_ATTENTION, + SDPBackend.CUDNN_ATTENTION, + SDPBackend.EFFICIENT_ATTENTION, + ] + BEST_SDPA_BACKEND = SDPBackend.FLASH_ATTENTION if compute_cap >= 80 else SDPBackend.EFFICIENT_ATTENTION + else: + assert dtype == torch.float32, f"Unrecognized {dtype=}." + SDPA_BACKENDS = [SDPBackend.EFFICIENT_ATTENTION] + BEST_SDPA_BACKEND = SDPBackend.EFFICIENT_ATTENTION + + if deterministic: + raise NotImplementedError( + "Deterministic mode in attention is only supported when Flash Attention 3 is available." + ) + + # Torch 2.6 and later allows priorities for backends, but for older versions + # we can only run with a specific backend. As long as we pick ones we're certain + # will work on that device, it should be fine. + try: + sdpa_kernel(backends=SDPA_BACKENDS, set_priority_order=True) + sdpa_kernel_ = partial(sdpa_kernel, set_priority_order=True) + except TypeError: + sdpa_kernel_ = sdpa_kernel + SDPA_BACKENDS = [BEST_SDPA_BACKEND] + + q = q.transpose(1, 2) + k = k.transpose(1, 2) + v = v.transpose(1, 2) + + with sdpa_kernel_(backends=SDPA_BACKENDS): + out = torch.nn.functional.scaled_dot_product_attention( + q, + k, + v, + is_causal=causal, + dropout_p=dropout_p, + scale=softmax_scale, + ) + + out = out.transpose(1, 2).contiguous() + return out diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/clip.py b/REGEN-main/cosmos_policy/_src/predict2/networks/clip.py new file mode 100644 index 0000000000000000000000000000000000000000..4d2f9901b61d45c5db261e653a90b251301cd159 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/clip.py @@ -0,0 +1,592 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Modified from ``https://github.com/openai/CLIP'' and ``https://github.com/mlfoundations/open_clip'' +# Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved. + +import math +from typing import Dict, List, Optional + +import torch +import torch.nn as nn +import torch.nn.functional as F +import torchvision.transforms as T + +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.predict2.conditioner import AbstractEmbModel +from cosmos_policy._src.predict2.inference.get_umt5_emb import HuggingfaceTokenizer +from cosmos_policy._src.predict2.networks.attention import attention +from cosmos_policy._src.predict2.networks.xlm_roberta import XLMRoberta + +__all__ = [ + "CLIPModel", +] + + +class QuickGELU(nn.Module): + def forward(self, x): + return x * torch.sigmoid(1.702 * x) + + +class LayerNorm(nn.LayerNorm): + def forward(self, x): + return super().forward(x.float()).type_as(x) + + +class SelfAttention(nn.Module): + def __init__(self, dim, num_heads, causal=False, attn_dropout=0.0, proj_dropout=0.0): + assert dim % num_heads == 0 + super().__init__() + self.dim = dim + self.num_heads = num_heads + self.head_dim = dim // num_heads + self.causal = causal + self.attn_dropout = attn_dropout + self.proj_dropout = proj_dropout + + # layers + self.to_qkv = nn.Linear(dim, dim * 3) + self.proj = nn.Linear(dim, dim) + + def forward(self, x): + """ + x: [B, L, C]. + """ + b, s, c, n, d = *x.size(), self.num_heads, self.head_dim + + # compute query, key, value + q, k, v = self.to_qkv(x).view(b, s, 3, n, d).unbind(2) + + # compute attention + p = self.attn_dropout if self.training else 0.0 + x = attention(q, k, v, dropout_p=p, causal=self.causal) + x = x.reshape(b, s, c) + + # output + x = self.proj(x) + x = F.dropout(x, self.proj_dropout, self.training) + return x + + +class SwiGLU(nn.Module): + def __init__(self, dim, mid_dim): + super().__init__() + self.dim = dim + self.mid_dim = mid_dim + + # layers + self.fc1 = nn.Linear(dim, mid_dim) + self.fc2 = nn.Linear(dim, mid_dim) + self.fc3 = nn.Linear(mid_dim, dim) + + def forward(self, x): + x = F.silu(self.fc1(x)) * self.fc2(x) + x = self.fc3(x) + return x + + +class AttentionBlock(nn.Module): + def __init__( + self, + dim, + mlp_ratio, + num_heads, + post_norm=False, + causal=False, + activation="quick_gelu", + attn_dropout=0.0, + proj_dropout=0.0, + norm_eps=1e-5, + ): + assert activation in ["quick_gelu", "gelu", "swi_glu"] + super().__init__() + self.dim = dim + self.mlp_ratio = mlp_ratio + self.num_heads = num_heads + self.post_norm = post_norm + self.causal = causal + self.norm_eps = norm_eps + + # layers + self.norm1 = LayerNorm(dim, eps=norm_eps) + self.attn = SelfAttention(dim, num_heads, causal, attn_dropout, proj_dropout) + self.norm2 = LayerNorm(dim, eps=norm_eps) + if activation == "swi_glu": + self.mlp = SwiGLU(dim, int(dim * mlp_ratio)) + else: + self.mlp = nn.Sequential( + nn.Linear(dim, int(dim * mlp_ratio)), + QuickGELU() if activation == "quick_gelu" else nn.GELU(), + nn.Linear(int(dim * mlp_ratio), dim), + nn.Dropout(proj_dropout), + ) + + def forward(self, x): + if self.post_norm: + x = x + self.norm1(self.attn(x)) + x = x + self.norm2(self.mlp(x)) + else: + x = x + self.attn(self.norm1(x)) + x = x + self.mlp(self.norm2(x)) + return x + + +class AttentionPool(nn.Module): + def __init__(self, dim, mlp_ratio, num_heads, activation="gelu", proj_dropout=0.0, norm_eps=1e-5): + assert dim % num_heads == 0 + super().__init__() + self.dim = dim + self.mlp_ratio = mlp_ratio + self.num_heads = num_heads + self.head_dim = dim // num_heads + self.proj_dropout = proj_dropout + self.norm_eps = norm_eps + + # layers + gain = 1.0 / math.sqrt(dim) + self.cls_embedding = nn.Parameter(gain * torch.randn(1, 1, dim)) + self.to_q = nn.Linear(dim, dim) + self.to_kv = nn.Linear(dim, dim * 2) + self.proj = nn.Linear(dim, dim) + self.norm = LayerNorm(dim, eps=norm_eps) + self.mlp = nn.Sequential( + nn.Linear(dim, int(dim * mlp_ratio)), + QuickGELU() if activation == "quick_gelu" else nn.GELU(), + nn.Linear(int(dim * mlp_ratio), dim), + nn.Dropout(proj_dropout), + ) + + def forward(self, x): + """ + x: [B, L, C]. + """ + b, s, c, n, d = *x.size(), self.num_heads, self.head_dim + + # compute query, key, value + q = self.to_q(self.cls_embedding).view(1, 1, n, d).expand(b, -1, -1, -1) + k, v = self.to_kv(x).view(b, s, 2, n, d).unbind(2) + + # compute attention + x = attention(q, k, v) + x = x.reshape(b, 1, c) + + # output + x = self.proj(x) + x = F.dropout(x, self.proj_dropout, self.training) + + # mlp + x = x + self.mlp(self.norm(x)) + return x[:, 0] + + +class VisionTransformer(nn.Module): + def __init__( + self, + image_size=224, + patch_size=16, + dim=768, + mlp_ratio=4, + out_dim=512, + num_heads=12, + num_layers=12, + pool_type="token", + pre_norm=True, + post_norm=False, + activation="quick_gelu", + attn_dropout=0.0, + proj_dropout=0.0, + embedding_dropout=0.0, + norm_eps=1e-5, + ): + if image_size % patch_size != 0: + print("[WARNING] image_size is not divisible by patch_size", flush=True) + assert pool_type in ("token", "token_fc", "attn_pool") + out_dim = out_dim or dim + super().__init__() + self.image_size = image_size + self.patch_size = patch_size + self.num_patches = (image_size // patch_size) ** 2 + self.dim = dim + self.mlp_ratio = mlp_ratio + self.out_dim = out_dim + self.num_heads = num_heads + self.num_layers = num_layers + self.pool_type = pool_type + self.post_norm = post_norm + self.norm_eps = norm_eps + + # embeddings + gain = 1.0 / math.sqrt(dim) + self.patch_embedding = nn.Conv2d(3, dim, kernel_size=patch_size, stride=patch_size, bias=not pre_norm) + if pool_type in ("token", "token_fc"): + self.cls_embedding = nn.Parameter(gain * torch.randn(1, 1, dim)) + self.pos_embedding = nn.Parameter( + gain * torch.randn(1, self.num_patches + (1 if pool_type in ("token", "token_fc") else 0), dim) + ) + self.dropout = nn.Dropout(embedding_dropout) + + # transformer + self.pre_norm = LayerNorm(dim, eps=norm_eps) if pre_norm else None + self.transformer = nn.Sequential( + *[ + AttentionBlock( + dim, mlp_ratio, num_heads, post_norm, False, activation, attn_dropout, proj_dropout, norm_eps + ) + for _ in range(num_layers) + ] + ) + self.post_norm = LayerNorm(dim, eps=norm_eps) + + # head + if pool_type == "token": + self.head = nn.Parameter(gain * torch.randn(dim, out_dim)) + elif pool_type == "token_fc": + self.head = nn.Linear(dim, out_dim) + elif pool_type == "attn_pool": + self.head = AttentionPool(dim, mlp_ratio, num_heads, activation, proj_dropout, norm_eps) + + def forward(self, x, interpolation=False, use_31_block=False): + b = x.size(0) + + # embeddings + x = self.patch_embedding(x).flatten(2).permute(0, 2, 1) + if self.pool_type in ("token", "token_fc"): + x = torch.cat([self.cls_embedding.expand(b, -1, -1), x], dim=1) + if interpolation: + e = pos_interpolate(self.pos_embedding, x.size(1)) # noqa: F821 + else: + e = self.pos_embedding + x = self.dropout(x + e) + if self.pre_norm is not None: + x = self.pre_norm(x) + + # transformer + if use_31_block: + x = self.transformer[:-1](x) + return x + else: + x = self.transformer(x) + return x + + +class XLMRobertaWithHead(XLMRoberta): + def __init__(self, **kwargs): + self.out_dim = kwargs.pop("out_dim") + super().__init__(**kwargs) + + # head + mid_dim = (self.dim + self.out_dim) // 2 + self.head = nn.Sequential( + nn.Linear(self.dim, mid_dim, bias=False), nn.GELU(), nn.Linear(mid_dim, self.out_dim, bias=False) + ) + + def forward(self, ids): + # xlm-roberta + x = super().forward(ids) + + # average pooling + mask = ids.ne(self.pad_id).unsqueeze(-1).to(x) + x = (x * mask).sum(dim=1) / mask.sum(dim=1) + + # head + x = self.head(x) + return x + + +class XLMRobertaCLIP(nn.Module): + def __init__( + self, + embed_dim=1024, + image_size=224, + patch_size=14, + vision_dim=1280, + vision_mlp_ratio=4, + vision_heads=16, + vision_layers=32, + vision_pool="token", + vision_pre_norm=True, + vision_post_norm=False, + activation="gelu", + vocab_size=250002, + max_text_len=514, + type_size=1, + pad_id=1, + text_dim=1024, + text_heads=16, + text_layers=24, + text_post_norm=True, + text_dropout=0.1, + attn_dropout=0.0, + proj_dropout=0.0, + embedding_dropout=0.0, + norm_eps=1e-5, + ): + super().__init__() + self.embed_dim = embed_dim + self.image_size = image_size + self.patch_size = patch_size + self.vision_dim = vision_dim + self.vision_mlp_ratio = vision_mlp_ratio + self.vision_heads = vision_heads + self.vision_layers = vision_layers + self.vision_pre_norm = vision_pre_norm + self.vision_post_norm = vision_post_norm + self.activation = activation + self.vocab_size = vocab_size + self.max_text_len = max_text_len + self.type_size = type_size + self.pad_id = pad_id + self.text_dim = text_dim + self.text_heads = text_heads + self.text_layers = text_layers + self.text_post_norm = text_post_norm + self.norm_eps = norm_eps + + # models + self.visual = VisionTransformer( + image_size=image_size, + patch_size=patch_size, + dim=vision_dim, + mlp_ratio=vision_mlp_ratio, + out_dim=embed_dim, + num_heads=vision_heads, + num_layers=vision_layers, + pool_type=vision_pool, + pre_norm=vision_pre_norm, + post_norm=vision_post_norm, + activation=activation, + attn_dropout=attn_dropout, + proj_dropout=proj_dropout, + embedding_dropout=embedding_dropout, + norm_eps=norm_eps, + ) + self.textual = XLMRobertaWithHead( + vocab_size=vocab_size, + max_seq_len=max_text_len, + type_size=type_size, + pad_id=pad_id, + dim=text_dim, + out_dim=embed_dim, + num_heads=text_heads, + num_layers=text_layers, + post_norm=text_post_norm, + dropout=text_dropout, + ) + self.log_scale = nn.Parameter(math.log(1 / 0.07) * torch.ones([])) + + def forward(self, imgs, txt_ids): + """ + imgs: [B, 3, H, W] of torch.float32. + - mean: [0.48145466, 0.4578275, 0.40821073] + - std: [0.26862954, 0.26130258, 0.27577711] + txt_ids: [B, L] of torch.long. + Encoded by data.CLIPTokenizer. + """ + xi = self.visual(imgs) + xt = self.textual(txt_ids) + return xi, xt + + def param_groups(self): + groups = [ + { + "params": [p for n, p in self.named_parameters() if "norm" in n or n.endswith("bias")], + "weight_decay": 0.0, + }, + {"params": [p for n, p in self.named_parameters() if not ("norm" in n or n.endswith("bias"))]}, + ] + return groups + + +def _clip( + pretrained=False, + pretrained_name=None, + model_cls=XLMRobertaCLIP, + return_transforms=False, + return_tokenizer=False, + tokenizer_padding="eos", + dtype=torch.float32, + device="cpu", + **kwargs, +): + # init a model on device + with torch.device(device): + model = model_cls(**kwargs) + + # set device + model = model.to(dtype=dtype, device=device) + output = (model,) + + # init transforms + if return_transforms: + # mean and std + if "siglip" in pretrained_name.lower(): + mean, std = [0.5, 0.5, 0.5], [0.5, 0.5, 0.5] + else: + mean = [0.48145466, 0.4578275, 0.40821073] + std = [0.26862954, 0.26130258, 0.27577711] + + # transforms + transforms = T.Compose( + [ + T.Resize((model.image_size, model.image_size), interpolation=T.InterpolationMode.BICUBIC), + T.ToTensor(), + T.Normalize(mean=mean, std=std), + ] + ) + output += (transforms,) + return output[0] if len(output) == 1 else output + + +def clip_xlm_roberta_vit_h_14(pretrained=False, pretrained_name="open-clip-xlm-roberta-large-vit-huge-14", **kwargs): + cfg = dict( + embed_dim=1024, + image_size=224, + patch_size=14, + vision_dim=1280, + vision_mlp_ratio=4, + vision_heads=16, + vision_layers=32, + vision_pool="token", + activation="gelu", + vocab_size=250002, + max_text_len=514, + type_size=1, + pad_id=1, + text_dim=1024, + text_heads=16, + text_layers=24, + text_post_norm=True, + text_dropout=0.1, + attn_dropout=0.0, + proj_dropout=0.0, + embedding_dropout=0.0, + ) + cfg.update(**kwargs) + return _clip(pretrained, pretrained_name, XLMRobertaCLIP, **cfg) + + +def load_model_torch(model, ckpt_path, credential_path: Optional[str] = None): + log.info(f"loading weights from {ckpt_path}") + if distributed.is_rank0(): + if ckpt_path.startswith("s3://"): + backend_key = "_clip_model" + easy_io.set_s3_backend( + key=backend_key, + backend_args={ + "backend": "s3", + "s3_credential_path": credential_path, + }, + ) + else: + backend_key = None + + ckpt = easy_io.load(ckpt_path, backend_key=backend_key, map_location="cpu") + model.load_state_dict(ckpt) + + distributed.sync_model_states(model, src=0) + return model + + +class CLIPModel: + def __init__( + self, + dtype=torch.float16, + device="cuda", + checkpoint_path="s3://bucket/cosmos_diffusion_v2/pretrain_weights/models_clip_open-clip-xlm-roberta-large-vit-huge-14_fp16.pth", + tokenizer_path="xlm-roberta-large", + credential_path: Optional[str] = "credentials/s3_training.secret", + ): + self.dtype = dtype + self.device = device + self.checkpoint_path = checkpoint_path + self.tokenizer_path = tokenizer_path + + # init model + self.model, self.transforms = clip_xlm_roberta_vit_h_14( + pretrained=False, return_transforms=True, return_tokenizer=False, dtype=dtype, device=device + ) + self.model = self.model.cuda().eval().requires_grad_(False) + self.model = load_model_torch(self.model, checkpoint_path, credential_path=credential_path) + + # init tokenizer + self.tokenizer = HuggingfaceTokenizer( + name=tokenizer_path, seq_len=self.model.max_text_len - 2, clean="whitespace" + ) + + def visual(self, videos_B_C_H_W_n1_p1): + # preprocess + size = (self.model.image_size,) * 2 + videos = F.interpolate(videos_B_C_H_W_n1_p1, size=size, mode="bicubic", align_corners=False) + videos = self.transforms.transforms[-1](videos.mul_(0.5).add_(0.5)) + + # forward + with torch.amp.autocast("cuda", dtype=self.dtype): + out = self.model.visual(videos, use_31_block=True) + return out + + +class Wan2pt1CLIPEmb(AbstractEmbModel): + def __init__( + self, + input_key: List[str], + dropout_rate: Optional[float] = 0.0, + num_token: int = 257, + dtype: str = "bfloat16", + ): + super().__init__() + self.num_token = num_token + self.model_dim = 1280 + self.clip_model = CLIPModel() + + self._input_key = input_key + self._output_key = None + self._dropout_rate = dropout_rate + self.dtype = { + "bfloat16": torch.bfloat16, + "float16": torch.float16, + "float32": torch.float32, + }[dtype] + + def random_dropout_input( + self, in_tensor: Optional[torch.Tensor] = None, dropout_rate: Optional[float] = None, key: Optional[str] = None + ) -> Optional[torch.Tensor]: + if in_tensor is None: + return None + return super().random_dropout_input(in_tensor, dropout_rate, key) + + def forward( + self, + image_tensor: Optional[torch.Tensor] = None, + video_tensor: Optional[torch.Tensor] = None, + media_latents: Optional[torch.Tensor] = None, + ) -> Dict[str, torch.Tensor]: + b, _, latent_f, latent_h, latent_w = media_latents.shape + mask = torch.zeros(b, 4, latent_f, latent_h, latent_w).type_as(media_latents).to(self.dtype) + if image_tensor is not None: # image case + context_B_L_D = torch.zeros(b, self.num_token, self.model_dim).type_as(media_latents).to(self.dtype) + else: + first_frame_B_C_H_W = video_tensor[:, :, 0, :, :] + with torch.no_grad(): + context_B_L_D = self.clip_model.visual(first_frame_B_C_H_W).to(self.dtype) + + mask[:, :, :1] = 1.0 + y = torch.concat([mask, media_latents.to(self.dtype)], dim=1) + + return {"frame_cond_crossattn_emb_B_L_D": context_B_L_D, "y_B_C_T_H_W": y} + + def details(self) -> str: + output_key = ["frame_cond_crossattn_emb_B_L_D", "y_B_C_T_H_W"] + return f"Input key: {self.input_key} \n\tOutput key: {output_key}" diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/minimal_v1_lvg_dit.py b/REGEN-main/cosmos_policy/_src/predict2/networks/minimal_v1_lvg_dit.py new file mode 100644 index 0000000000000000000000000000000000000000..ff5669be05ccacd74f2cf982ffd5950b3ba1ff52 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/minimal_v1_lvg_dit.py @@ -0,0 +1,62 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import List, Optional, Tuple + +import torch + +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.networks.minimal_v4_dit import MiniTrainDIT + + +class MinimalV1LVGDiT(MiniTrainDIT): + def __init__(self, *args, timestep_scale: float = 1.0, **kwargs): + assert "in_channels" in kwargs, "in_channels must be provided" + kwargs["in_channels"] += 1 # Add 1 for the condition mask + self.timestep_scale = timestep_scale + super().__init__(*args, **kwargs) + + def forward( + self, + x_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + crossattn_emb: torch.Tensor, + condition_video_input_mask_B_C_T_H_W: Optional[torch.Tensor] = None, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + data_type: Optional[DataType] = DataType.VIDEO, + intermediate_feature_ids: Optional[List[int]] = None, + img_context_emb: Optional[torch.Tensor] = None, + **kwargs, + ) -> torch.Tensor | List[torch.Tensor] | Tuple[torch.Tensor, List[torch.Tensor]]: + del kwargs + + if data_type == DataType.VIDEO: + x_B_C_T_H_W = torch.cat([x_B_C_T_H_W, condition_video_input_mask_B_C_T_H_W.type_as(x_B_C_T_H_W)], dim=1) + else: + B, _, T, H, W = x_B_C_T_H_W.shape + x_B_C_T_H_W = torch.cat( + [x_B_C_T_H_W, torch.zeros((B, 1, T, H, W), dtype=x_B_C_T_H_W.dtype, device=x_B_C_T_H_W.device)], dim=1 + ) + return super().forward( + x_B_C_T_H_W=x_B_C_T_H_W, + timesteps_B_T=timesteps_B_T * self.timestep_scale, + crossattn_emb=crossattn_emb, + fps=fps, + padding_mask=padding_mask, + data_type=data_type, + intermediate_feature_ids=intermediate_feature_ids, + img_context_emb=img_context_emb, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/minimal_v4_dit.py b/REGEN-main/cosmos_policy/_src/predict2/networks/minimal_v4_dit.py new file mode 100644 index 0000000000000000000000000000000000000000..aa530afe337f8272be3d5dbb53b6a5fa6cf6d532 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/minimal_v4_dit.py @@ -0,0 +1,1949 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import collections +import math +from collections import namedtuple +from collections.abc import Sequence +from dataclasses import dataclass +from enum import Enum +from typing import List, Optional, Tuple, Union + +from cosmos_policy._src.predict2.utils.kv_cache import AttentionOpWithKVCache, KVCacheConfig + +try: + import megatron.core.parallel_state as parallel_state + + USE_MEGATRON = True +except ImportError: + USE_MEGATRON = False + +import numpy as np +import torch +import torch.amp as amp +import transformer_engine as te +from einops import rearrange, repeat +from einops.layers.torch import Rearrange +from packaging.version import Version +from torch import nn +from torch.distributed import ProcessGroup, get_process_group_ranks +from torch.distributed._composable.fsdp import fully_shard +from torch.distributed.algorithms._checkpoint.checkpoint_wrapper import checkpoint_wrapper as ptd_checkpoint_wrapper + +try: + from torch.utils.checkpoint import CheckpointPolicy, create_selective_checkpoint_contexts +except ImportError: + CheckpointPolicy = None + +from torchvision import transforms + +if Version(te.__version__) >= Version("2.8.0"): + from transformer_engine.pytorch.attention.rope import apply_rotary_pos_emb +else: + from transformer_engine.pytorch.attention import apply_rotary_pos_emb +from torch.nn.attention.flex_attention import BlockMask, create_block_mask, flex_attention + +from cosmos_policy._src.imaginaire.attention import attention +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.context_parallel import split_inputs_cp +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.modules.neighborhood_attn import NeighborhoodAttention +from cosmos_policy._src.predict2.networks.a2a_cp import MinimalA2AAttnOp, NattenA2AAttnOp +from cosmos_policy._src.predict2.networks.model_weights_stats import WeightTrainingStat +from cosmos_policy._src.predict2.networks.selective_activation_checkpoint import SACConfig as _SACConfig + + +# selective activation checkpoint; only apply to the minimal v4 model. if there are change in the networks, some policy will not work as we expect. +def predict2_2B_720_context_fn(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + op_count_key = f"{mode}_mm_count" + # from cosmos_policy._src.imaginaire.utils import log + # log.info(f"op_count_key: {op_count_key}, op_count[op_count_key]: {op_count[op_count_key]}, {args[0].shape}, {args[1].shape}") + # there are totally 6 + 4 + 4 + 2 = 16 block + op_count[op_count_key] = (op_count[op_count_key] + 1) % 16 + if op_count[op_count_key] > 8: # recompute self attn first 3 linear layers + return CheckpointPolicy.MUST_SAVE + if "flash_attn" in str(func): + op_count_key = f"{mode}_flash_attn_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 2 + if op_count[op_count_key]: + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_2B_720_context_fn_aggressive(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + # The default policy is to recompute everything. This is the most memory-efficient + # starting point. We then selectively choose what to save. + default_policy = CheckpointPolicy.PREFER_RECOMPUTE + + # Save the output of Flash Attention. This is the most computationally + # expensive part of a transformer block. Saving its output provides a + # good balance between memory savings and computational overhead. + if "flash_attn" in str(func): + return CheckpointPolicy.MUST_SAVE + + # All other operations (e.g., torch.ops.aten.mm.default, layer norms, additions) + # will fall through to the default policy and be recomputed. + return default_policy + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_2B_720_context_fn_aggressive_v2(): + """ + The most memory-aggressive checkpointing policy. Recomputes ALL operations. + """ + + def policy_fn(ctx, func, *args, **kwargs): + # The policy is to always recompute everything. + # This saves the maximum amount of memory but incurs the highest + # computational cost during the backward pass. + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_14B_720_context_fn(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + op_count_key = f"{mode}_mm_count" + # from cosmos_policy._src.imaginaire.utils import log + # log.info(f"op_count_key: {op_count_key}, op_count[op_count_key]: {op_count[op_count_key]}, {args[0].shape}, {args[1].shape}") + # there are totally 6 + 4 + 4 + 2 = 16 block + op_count[op_count_key] = (op_count[op_count_key] + 1) % 16 + if op_count[op_count_key] > 8: # recompute self attn first 1 linear layers + return CheckpointPolicy.MUST_SAVE + if "flash_attn" in str(func): + op_count_key = f"{mode}_flash_attn_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 2 + if op_count[op_count_key]: + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def predict2_14B_720_context_fn_aggressive(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + op_count_key = f"{mode}_mm_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 16 + if op_count[op_count_key] > 12: # recompute self attn first 1 linear layers + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +def linear_selfattn_context_fn(): + op_count = collections.defaultdict(int) + + def policy_fn(ctx, func, *args, **kwargs): + mode = "recompute" if ctx.is_recompute else "forward" + if func == torch.ops.aten.mm.default: + return CheckpointPolicy.MUST_SAVE + if "flash_attn" in str(func): + op_count_key = f"{mode}_flash_attn_count" + op_count[op_count_key] = (op_count[op_count_key] + 1) % 2 + if op_count[op_count_key]: + return CheckpointPolicy.MUST_SAVE + return CheckpointPolicy.PREFER_RECOMPUTE + + return create_selective_checkpoint_contexts(policy_fn) + + +class CheckpointMode(str, Enum): + NONE = "none" + MM_ONLY = "mm_only" + BLOCK_WISE = "block_wise" + LINEAR_SELFATTN = "linear_selfattn" + PREDICT2_2B_720 = "predict2_2b_720" + PREDICT2_14B_720 = "predict2_14b_720" + PREDICT2_2B_720_AGGRESSIVE = "predict2_2b_720_aggressive" + PREDICT2_2B_720_AGGRESSIVE_V2 = "predict2_2b_720_aggressive_v2" + PREDICT2_14B_720_AGGRESSIVE = "predict2_14b_720_aggressive" + + def __str__(self) -> str: + return self.value + + +@dataclass +class SACConfig(_SACConfig): + def get_context_fn(self): + if self.mode == CheckpointMode.LINEAR_SELFATTN: + return linear_selfattn_context_fn + elif self.mode == CheckpointMode.PREDICT2_2B_720: + return predict2_2B_720_context_fn + elif self.mode == CheckpointMode.PREDICT2_2B_720_AGGRESSIVE: + return predict2_2B_720_context_fn_aggressive + elif self.mode == CheckpointMode.PREDICT2_2B_720_AGGRESSIVE_V2: + return predict2_2B_720_context_fn_aggressive_v2 + elif self.mode == CheckpointMode.PREDICT2_14B_720: + return predict2_14B_720_context_fn + elif self.mode == CheckpointMode.PREDICT2_14B_720_AGGRESSIVE: + return predict2_14B_720_context_fn_aggressive + else: + # Reuse parent class implementation for other modes + return super().get_context_fn() + + +VideoSize = namedtuple("VideoSize", ["T", "H", "W"]) + + +class RMSNorm(torch.nn.Module): + def __init__(self, dim: int, eps: float = 1e-5): + super().__init__() + self.eps = eps + self.weight = nn.Parameter(torch.ones(dim)) + + def reset_parameters(self): + torch.nn.init.ones_(self.weight) + + def _norm(self, x): + return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps) + + def forward(self, x: torch.Tensor) -> torch.Tensor: + output = self._norm(x.float()).type_as(x) + return output * self.weight + + +# ---------------------- Feed Forward Network ----------------------- +class GPT2FeedForward(nn.Module): + def __init__(self, d_model: int, d_ff: int): + super().__init__() + self.activation = nn.GELU() + self.layer1 = nn.Linear(d_model, d_ff, bias=False) + self.layer2 = nn.Linear(d_ff, d_model, bias=False) + + self._layer_id = None + self._dim = d_model + self._hidden_dim = d_ff + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self._dim) + torch.nn.init.trunc_normal_(self.layer1.weight, std=std, a=-3 * std, b=3 * std) + + # scale init by depth as in https://arxiv.org/abs/1908.11365 -- worked slightly better. + std = 1.0 / math.sqrt(self._hidden_dim) + if self._layer_id is not None: + std = std / math.sqrt(2 * (self._layer_id + 1)) + torch.nn.init.trunc_normal_(self.layer2.weight, std=std, a=-3 * std, b=3 * std) + + def forward(self, x: torch.Tensor): + x = self.layer1(x) + + x = self.activation(x) + x = self.layer2(x) + return x + + +def torch_attention_op( + q_B_S_H_D: torch.Tensor, + k_B_S_H_D: torch.Tensor, + v_B_S_H_D: torch.Tensor, + attn_mask: Optional[torch.Tensor] = None, + flatten_heads: bool = True, +) -> torch.Tensor: + """Scaled dot-product attention with optional mask. + + Inputs are shaped [B, S, H, D]. If flatten_heads=True, flattens heads to return [B, S, H*D]. + Otherwise returns [B, S, H, D]. + """ + q_B_H_S_D = rearrange(q_B_S_H_D, "b s h d -> b h s d") + k_B_H_S_D = rearrange(k_B_S_H_D, "b s h d -> b h s d") + v_B_H_S_D = rearrange(v_B_S_H_D, "b s h d -> b h s d") + result_B_H_S_D = torch.nn.functional.scaled_dot_product_attention( + q_B_H_S_D, k_B_H_S_D, v_B_H_S_D, attn_mask=attn_mask + ) + if flatten_heads: + return rearrange(result_B_H_S_D, "b h s d -> b s (h d)") + else: + return rearrange(result_B_H_S_D, "b h s d -> b s h d") + + +def flex_attention_op( + q_B_S_H_D: torch.Tensor, + k_B_S_H_D: torch.Tensor, + v_B_S_H_D: torch.Tensor, + attn_mask: Optional[BlockMask] = None, + flatten_heads: bool = True, +) -> torch.Tensor: + # Rearrange to [B, H, S, D] + q_B_H_Sq_D = rearrange(q_B_S_H_D, "b s h d -> b h s d") + k_B_H_Sk_D = rearrange(k_B_S_H_D, "b s h d -> b h s d") + v_B_H_Sk_D = rearrange(v_B_S_H_D, "b s h d -> b h s d") + + S_q = q_B_H_Sq_D.shape[2] + S_kv = k_B_H_Sk_D.shape[2] + # Right-pad to multiples of 128 for optimal FlexAttention kernels + pad_q = ((S_q + 127) // 128) * 128 - S_q + pad_kv = ((S_kv + 127) // 128) * 128 - S_kv + + if pad_q > 0: + q_pad_tensor = torch.zeros( + (q_B_H_Sq_D.shape[0], q_B_H_Sq_D.shape[1], pad_q, q_B_H_Sq_D.shape[3]), + device=q_B_H_Sq_D.device, + dtype=q_B_H_Sq_D.dtype, + ) + q_cat = torch.cat([q_B_H_Sq_D, q_pad_tensor], dim=2) + else: + q_cat = q_B_H_Sq_D + + if pad_kv > 0: + kv_pad_tensor = torch.zeros( + (k_B_H_Sk_D.shape[0], k_B_H_Sk_D.shape[1], pad_kv, k_B_H_Sk_D.shape[3]), + device=k_B_H_Sk_D.device, + dtype=k_B_H_Sk_D.dtype, + ) + k_cat = torch.cat([k_B_H_Sk_D, kv_pad_tensor], dim=2) + v_cat = torch.cat([v_B_H_Sk_D, kv_pad_tensor], dim=2) + else: + k_cat, v_cat = k_B_H_Sk_D, v_B_H_Sk_D + + block_mask = None + if attn_mask is not None and isinstance(attn_mask, BlockMask): + block_mask = attn_mask + else: + # When padding is introduced without an explicit mask, build a validity mask + if pad_q > 0 or pad_kv > 0: + + def allow_valid(b, h, q_idx, kv_idx): + return (q_idx < S_q) & (kv_idx < S_kv) + + block_mask = create_block_mask( + allow_valid, + B=None, + H=None, + Q_LEN=q_cat.shape[2], + KV_LEN=k_cat.shape[2], + _compile=True, + device=q_cat.device, + ) + + if block_mask is not None: + out_B_H_Sqp_D = torch.compile(flex_attention)(query=q_cat, key=k_cat, value=v_cat, block_mask=block_mask) + else: + out_B_H_Sqp_D = torch.compile(flex_attention)(query=q_cat, key=k_cat, value=v_cat) + + out_B_H_Sq_D = out_B_H_Sqp_D[:, :, :S_q] if pad_q > 0 else out_B_H_Sqp_D + if flatten_heads: + return rearrange(out_B_H_Sq_D, "b h s d -> b s (h d)") + else: + return rearrange(out_B_H_Sq_D, "b h s d -> b s h d") + + +def i4_attention_op( + q_B_S_H_D: torch.Tensor, + k_B_S_H_D: torch.Tensor, + v_B_S_H_D: torch.Tensor, + flatten_heads: bool = True, + **kwargs: dict, +) -> torch.Tensor: + """ + I4 regular (bidirectional) attention. + Matches torch_attention_op's signature but omits attn_mask (full attention assumed). + Ignores any additional kwargs (e.g., video_size). + """ + out_B_S_H_D = attention( + query=q_B_S_H_D, + key=k_B_S_H_D, + value=v_B_S_H_D, + is_causal=False, + ) + if isinstance(out_B_S_H_D, tuple): + out_B_S_H_D = out_B_S_H_D[0] + if flatten_heads: + return rearrange(out_B_S_H_D, "b s h d -> b s (h d)") + else: + return out_B_S_H_D + + +class Attention(nn.Module): + """ + A flexible attention module supporting both self-attention and cross-attention mechanisms. + + This module implements a multi-head attention layer that can operate in either self-attention + or cross-attention mode. The mode is determined by whether a context dimension is provided. + The implementation uses scaled dot-product attention and supports optional bias terms and + dropout regularization. + + Args: + query_dim (int): The dimensionality of the query vectors. + context_dim (int, optional): The dimensionality of the context (key/value) vectors. + If None, the module operates in self-attention mode using query_dim. Default: None + n_heads (int, optional): Number of attention heads for multi-head attention. Default: 8 + head_dim (int, optional): The dimension of each attention head. Default: 64 + dropout (float, optional): Dropout probability applied to the output. Default: 0.0 + qkv_format (str, optional): Format specification for QKV tensors. Default: "bshd" + backend (str, optional): Backend to use for the attention operation. Default: "transformer_engine" + + Examples: + >>> # Self-attention with 512 dimensions and 8 heads + >>> self_attn = Attention(query_dim=512) + >>> x = torch.randn(32, 16, 512) # (batch_size, seq_len, dim) + >>> out = self_attn(x) # (32, 16, 512) + + >>> # Cross-attention + >>> cross_attn = Attention(query_dim=512, context_dim=256) + >>> query = torch.randn(32, 16, 512) + >>> context = torch.randn(32, 8, 256) + >>> out = cross_attn(query, context) # (32, 16, 512) + """ + + def __init__( + self, + query_dim: int, + context_dim=None, + n_heads=8, + head_dim=64, + dropout=0.0, + qkv_format: str = "bshd", + backend: str = "transformer_engine", + use_wan_fp32_strategy: bool = False, + ) -> None: + super().__init__() + log.debug( + f"Setting up {self.__class__.__name__}. Query dim is {query_dim}, context_dim is {context_dim} and using " + f"{n_heads} heads with a dimension of {head_dim}." + ) + self.is_selfattn = context_dim is None # self attention + + assert backend in ["transformer_engine", "torch", "torch-flex", "minimal_a2a", "i4"], ( + f"Invalid backend: {backend}" + ) + self.backend = backend + + context_dim = query_dim if context_dim is None else context_dim + inner_dim = head_dim * n_heads + + self.n_heads = n_heads + self.head_dim = head_dim + self.qkv_format = qkv_format + self.query_dim = query_dim + self.context_dim = context_dim + self.use_wan_fp32_strategy = use_wan_fp32_strategy + + self.q_proj = nn.Linear(query_dim, inner_dim, bias=False) + self.q_norm = te.pytorch.RMSNorm(self.head_dim, eps=1e-6) + + self.k_proj = nn.Linear(context_dim, inner_dim, bias=False) + self.k_norm = te.pytorch.RMSNorm(self.head_dim, eps=1e-6) + + self.v_proj = nn.Linear(context_dim, inner_dim, bias=False) + self.v_norm = nn.Identity() + + self.output_proj = nn.Linear(inner_dim, query_dim, bias=False) + self.output_dropout = nn.Dropout(dropout) if dropout > 1e-4 else nn.Identity() + + if self.backend == "transformer_engine": + from transformer_engine.pytorch.attention import DotProductAttention + + self.attn_op = DotProductAttention( + self.n_heads, + self.head_dim, + num_gqa_groups=self.n_heads, + attention_dropout=0, + qkv_format=qkv_format, + attn_mask_type="no_mask", + ) + elif self.backend == "minimal_a2a": + self.attn_op = MinimalA2AAttnOp() + elif self.backend == "torch": + self.attn_op = torch_attention_op + elif self.backend == "torch-flex": + # FlexAttention backend; returns [B, S, H*D] + self.attn_op = flex_attention_op + elif self.backend == "i4": + # I4 spatio-temporal attention; returns [B, S, H*D] + self.attn_op = i4_attention_op + + if not hasattr(self.attn_op, "set_context_parallel_group"): + + def set_context_parallel_group(*args, **kwargs) -> None: + return None + + self.attn_op.set_context_parallel_group = set_context_parallel_group + + self._query_dim = query_dim + self._context_dim = context_dim + self._inner_dim = inner_dim + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self._query_dim) + torch.nn.init.trunc_normal_(self.q_proj.weight, std=std, a=-3 * std, b=3 * std) + std = 1.0 / math.sqrt(self._context_dim) + torch.nn.init.trunc_normal_(self.k_proj.weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.v_proj.weight, std=std, a=-3 * std, b=3 * std) + + std = 1.0 / math.sqrt(self._inner_dim) + torch.nn.init.trunc_normal_(self.output_proj.weight, std=std, a=-3 * std, b=3 * std) + + for layer in self.q_norm, self.k_norm, self.v_norm: + if hasattr(layer, "reset_parameters"): + layer.reset_parameters() + + def compute_qkv(self, x, context=None, rope_emb=None) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor]: + q = self.q_proj(x) + context = x if context is None else context + k = self.k_proj(context) + v = self.v_proj(context) + q, k, v = map( + lambda t: rearrange(t, "b ... (h d) -> b ... h d", h=self.n_heads, d=self.head_dim), + (q, k, v), + ) + + def apply_norm_and_rotary_pos_emb(q, k, v, rope_emb): + q = self.q_norm(q) + k = self.k_norm(k) + v = self.v_norm(v) + original_dtype = q.dtype + if self.is_selfattn and rope_emb is not None: # only apply to self-attention! + if self.use_wan_fp32_strategy: # wan will force q and k to fp32 before rotary pos emb + q = q.to(torch.float32) + k = k.to(torch.float32) + q = apply_rotary_pos_emb(q, rope_emb, tensor_format=self.qkv_format, fused=True) + k = apply_rotary_pos_emb(k, rope_emb, tensor_format=self.qkv_format, fused=True) + if self.use_wan_fp32_strategy: + q = q.to(original_dtype) + k = k.to(original_dtype) + return q, k, v + + q, k, v = apply_norm_and_rotary_pos_emb(q, k, v, rope_emb) + + return q, k, v + + def compute_attention( + self, + q, + k, + v, + video_size: Optional[VideoSize] = None, + kv_cache_cfg: Optional[KVCacheConfig] = None, + ): + additional_args = {} + if isinstance(self.attn_op, (NattenA2AAttnOp, NeighborhoodAttention)) or self.backend == "i4": + additional_args["video_size"] = video_size + if isinstance(self.attn_op, AttentionOpWithKVCache): + additional_args["kv_cache_cfg"] = kv_cache_cfg + + result = self.attn_op(q, k, v, **additional_args) # [B, S, H, D] + return self.output_dropout(self.output_proj(result)) + + def forward( + self, + x, + context: Optional[torch.Tensor] = None, + rope_emb: Optional[torch.Tensor] = None, + video_size: Optional[VideoSize] = None, + kv_cache_cfg: Optional[KVCacheConfig] = None, + ): + """ + Args: + x (Tensor): The query tensor of shape [B, Mq, K] + context (Optional[Tensor]): The key tensor of shape [B, Mk, K] or use x as context [self attention] if None + rope_emb (Optional[Tensor]): RoPE embedding tensor, or no RoPE embeddings (i.e. in cross attention) + video_size(VideoSize): Shape [T, H, W] + """ + q, k, v = self.compute_qkv(x, context, rope_emb=rope_emb) + return self.compute_attention(q, k, v, video_size=video_size, kv_cache_cfg=kv_cache_cfg) + + def set_context_parallel_group(self, process_group, ranks, stream, cp_comm_type: str = "p2p"): + # self.attn_op.set_context_parallel_group(process_group, ranks, stream, cp_comm_type="a2a") + self.attn_op.set_context_parallel_group(process_group, ranks, stream, cp_comm_type=cp_comm_type) + + +class I2VCrossAttention(Attention): + def __init__(self, *args, img_latent_dim: int = 1024, **kwargs): + super().__init__(*args, **kwargs) + inner_dim = self.head_dim * self.n_heads + self.k_img = nn.Linear(img_latent_dim, inner_dim, bias=False) + self.v_img = nn.Linear(img_latent_dim, inner_dim, bias=False) + self.k_img_norm = te.pytorch.RMSNorm(self.head_dim, eps=1e-6) + + def init_weights(self) -> None: + super().init_weights() + torch.nn.init.trunc_normal_(self.k_img.weight, std=1.0 / math.sqrt(self._inner_dim)) + torch.nn.init.trunc_normal_(self.v_img.weight, std=1.0 / math.sqrt(self._inner_dim)) + self.k_img_norm.reset_parameters() + + def compute_qkv( + self, x, context, rope_emb=None + ) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]: + text_context, img_context = context + q, k, v = super().compute_qkv(x, text_context, rope_emb) + k_img = self.k_img(img_context) + v_img = self.v_img(img_context) + # Rearrange k_img, v_img + k_img, v_img = map( + lambda t: rearrange(t, "b ... (h d) -> b ... h d", h=self.n_heads, d=self.head_dim), + (k_img, v_img), + ) + + return q, k, v, self.k_img_norm(k_img), v_img + + def compute_attention(self, q, k, v, k_img, v_img): + result = self.attn_op(q, k, v) # [B, S, H, D] + result_img = self.attn_op(q, k_img, v_img) + return self.output_dropout(self.output_proj(result + result_img)) + + def forward( + self, + x, + context=None, + rope_emb=None, + ): + q, k, v, k_img, v_img = self.compute_qkv(x, context, rope_emb) + return self.compute_attention(q, k, v, k_img, v_img) + + +class VideoPositionEmb(nn.Module): + def __init__(self): + super().__init__() + self._cp_group = None + + def enable_context_parallel(self, process_group: ProcessGroup): + self._cp_group = process_group + + def disable_context_parallel(self): + self._cp_group = None + + @property + def seq_dim(self): + return 1 + + def forward(self, x_B_T_H_W_C: torch.Tensor, fps=Optional[torch.Tensor]) -> torch.Tensor: + """ + With CP, the function assume that the input tensor is already split. + It delegates the embedding generation to generate_embeddings function. + """ + B_T_H_W_C = x_B_T_H_W_C.shape + if self._cp_group is not None: + cp_ranks = get_process_group_ranks(self._cp_group) + cp_size = len(cp_ranks) + cp_size_t = cp_size + if USE_MEGATRON and hasattr(parallel_state, "cp_size_t"): + # We saved cp_size_t in find_split function for combined temporal and spatial splitting. + # We need cp_size_t to find out the split values for T and H dimensions for correct embedding calculations. + cp_size_t = parallel_state.cp_size_t + cp_size_h = max(1, cp_size // cp_size_t) + B, T, H, W, C = B_T_H_W_C + B_T_H_W_C = (B, T * cp_size_t, H * cp_size_h, W, C) + embeddings = self.generate_embeddings(B_T_H_W_C, fps=fps) + + return self._split_for_context_parallel(embeddings) + + def generate_embeddings(self, B_T_H_W_C: torch.Size, fps=Optional[torch.Tensor]): + raise NotImplementedError + + def _split_for_context_parallel(self, embeddings): + if self._cp_group is not None: + embeddings = split_inputs_cp(x=embeddings, seq_dim=self.seq_dim, cp_group=self._cp_group) + return embeddings + + +class VideoRopePosition3DEmb(VideoPositionEmb): + def __init__( + self, + *, # enforce keyword arguments + head_dim: int, + len_h: int, + len_w: int, + len_t: int, + base_fps: int = 24, + h_extrapolation_ratio: float = 1.0, + w_extrapolation_ratio: float = 1.0, + t_extrapolation_ratio: float = 1.0, + enable_fps_modulation: bool = True, + **kwargs, # used for compatibility with other positional embeddings; unused in this class + ): + del kwargs + super().__init__() + self.register_buffer("seq", torch.arange(max(len_h, len_w, len_t), dtype=torch.float)) + self.base_fps = base_fps + self.max_h = len_h + self.max_w = len_w + self.max_t = len_t + self.enable_fps_modulation = enable_fps_modulation + dim = head_dim + dim_h = dim // 6 * 2 + dim_w = dim_h + dim_t = dim - 2 * dim_h + assert dim == dim_h + dim_w + dim_t, f"bad dim: {dim} != {dim_h} + {dim_w} + {dim_t}" + + self.register_buffer( + "dim_spatial_range", + torch.arange(0, dim_h, 2)[: (dim_h // 2)].float() / dim_h, + persistent=True, + ) + self.register_buffer( + "dim_temporal_range", + torch.arange(0, dim_t, 2)[: (dim_t // 2)].float() / dim_t, + persistent=True, + ) + self._dim_h = dim_h + self._dim_t = dim_t + + self.h_ntk_factor = h_extrapolation_ratio ** (dim_h / (dim_h - 2)) + self.w_ntk_factor = w_extrapolation_ratio ** (dim_w / (dim_w - 2)) + self.t_ntk_factor = t_extrapolation_ratio ** (dim_t / (dim_t - 2)) + self.reset_parameters() + + def reset_parameters(self) -> None: + dim_h = self._dim_h + dim_t = self._dim_t + + self.seq = torch.arange(max(self.max_h, self.max_w, self.max_t)).float().to(self.dim_spatial_range.device) + self.dim_spatial_range = ( + torch.arange(0, dim_h, 2)[: (dim_h // 2)].float().to(self.dim_spatial_range.device) / dim_h + ) + self.dim_temporal_range = ( + torch.arange(0, dim_t, 2)[: (dim_t // 2)].float().to(self.dim_spatial_range.device) / dim_t + ) + + def generate_embeddings( + self, + B_T_H_W_C: torch.Size, + fps: Optional[torch.Tensor] = None, + h_ntk_factor: Optional[float] = None, + w_ntk_factor: Optional[float] = None, + t_ntk_factor: Optional[float] = None, + ): + """ + Generate embeddings for the given input size. + + Args: + B_T_H_W_C (torch.Size): Input tensor size (Batch, Time, Height, Width, Channels). + fps (Optional[torch.Tensor], optional): Frames per second. Defaults to None. + h_ntk_factor (Optional[float], optional): Height NTK factor. If None, uses self.h_ntk_factor. + w_ntk_factor (Optional[float], optional): Width NTK factor. If None, uses self.w_ntk_factor. + t_ntk_factor (Optional[float], optional): Time NTK factor. If None, uses self.t_ntk_factor. + + Returns: + Not specified in the original code snippet. + """ + h_ntk_factor = h_ntk_factor if h_ntk_factor is not None else self.h_ntk_factor + w_ntk_factor = w_ntk_factor if w_ntk_factor is not None else self.w_ntk_factor + t_ntk_factor = t_ntk_factor if t_ntk_factor is not None else self.t_ntk_factor + + h_theta = 10000.0 * h_ntk_factor + w_theta = 10000.0 * w_ntk_factor + t_theta = 10000.0 * t_ntk_factor + + h_spatial_freqs = 1.0 / (h_theta ** self.dim_spatial_range.float()) + w_spatial_freqs = 1.0 / (w_theta ** self.dim_spatial_range.float()) + temporal_freqs = 1.0 / (t_theta ** self.dim_temporal_range.float()) + + B, T, H, W, _ = B_T_H_W_C + assert H <= self.max_h and W <= self.max_w, ( + f"Input dimensions (H={H}, W={W}) exceed the maximum dimensions (max_h={self.max_h}, max_w={self.max_w})" + ) + half_emb_h = torch.outer(self.seq[:H], h_spatial_freqs) + half_emb_w = torch.outer(self.seq[:W], w_spatial_freqs) + + if self.enable_fps_modulation: + uniform_fps = (fps is None) or (fps.min() == fps.max()) + assert uniform_fps or B == 1 or T == 1, ( + "For video batch, batch size should be 1 for non-uniform fps. For image batch, T should be 1" + ) + + # apply sequence scaling in temporal dimension + if fps is None: # image case + assert T == 1, "T should be 1 for image batch." + half_emb_t = torch.outer(self.seq[:T], temporal_freqs) + else: + half_emb_t = torch.outer(self.seq[:T] / fps[:1] * self.base_fps, temporal_freqs) + else: + half_emb_t = torch.outer(self.seq[:T], temporal_freqs) + + em_T_H_W_D = torch.cat( + [ + repeat(half_emb_t, "t d -> t h w d", h=H, w=W), + repeat(half_emb_h, "h d -> t h w d", t=T, w=W), + repeat(half_emb_w, "w d -> t h w d", t=T, h=H), + ] + * 2, + dim=-1, + ) + + return rearrange(em_T_H_W_D, "t h w d -> (t h w) 1 1 d").float() + + @property + def seq_dim(self): + return 0 + + +class LearnablePosEmbAxis(VideoPositionEmb): + def __init__( + self, + *, # enforce keyword arguments + interpolation: str, + model_channels: int, + len_h: int, + len_w: int, + len_t: int, + **kwargs, + ): + """ + Args: + interpolation (str): we curretly only support "crop", ideally when we need extrapolation capacity, we should adjust frequency or other more advanced methods. they are not implemented yet. + """ + del kwargs # unused + super().__init__() + self.interpolation = interpolation + assert self.interpolation in ["crop"], f"Unknown interpolation method {self.interpolation}" + self.model_channels = model_channels + + self.pos_emb_h = nn.Parameter(torch.zeros(len_h, model_channels)) + self.pos_emb_w = nn.Parameter(torch.zeros(len_w, model_channels)) + self.pos_emb_t = nn.Parameter(torch.zeros(len_t, model_channels)) + + self.reset_parameters() + + def reset_parameters(self): + std = 1.0 / math.sqrt(self.model_channels) + torch.nn.init.trunc_normal_(self.pos_emb_h, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.pos_emb_w, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.pos_emb_t, std=std, a=-3 * std, b=3 * std) + + def generate_embeddings(self, B_T_H_W_C: torch.Size, fps=Optional[torch.Tensor]) -> torch.Tensor: + B, T, H, W, _ = B_T_H_W_C + if self.interpolation == "crop": + emb_h_H = self.pos_emb_h[:H] + emb_w_W = self.pos_emb_w[:W] + emb_t_T = self.pos_emb_t[:T] + emb = ( + repeat(emb_t_T, "t d-> b t h w d", b=B, h=H, w=W) + + repeat(emb_h_H, "h d-> b t h w d", b=B, t=T, w=W) + + repeat(emb_w_W, "w d-> b t h w d", b=B, t=T, h=H) + ) + assert list(emb.shape)[:4] == [B, T, H, W], f"bad shape: {list(emb.shape)[:4]} != {B, T, H, W}" + else: + raise ValueError(f"Unknown interpolation method {self.interpolation}") + + norm = torch.linalg.vector_norm(emb, dim=-1, keepdim=True, dtype=torch.float32) + norm = torch.add(1e-6, norm, alpha=np.sqrt(norm.numel() / emb.numel())) + return emb / norm.to(emb.dtype) + + +def modulate(x, shift, scale): + return x * (1 + scale) + shift + + +class Timesteps(nn.Module): + def __init__(self, num_channels): + super().__init__() + self.num_channels = num_channels + + def forward(self, timesteps_B_T): + assert timesteps_B_T.ndim == 2, f"Expected 2D input, got {timesteps_B_T.ndim}" + # wan need emb to be in fp32 + in_dype = timesteps_B_T.dtype + timesteps = timesteps_B_T.flatten().float() + half_dim = self.num_channels // 2 + exponent = -math.log(10000) * torch.arange(half_dim, dtype=torch.float32, device=timesteps.device) + exponent = exponent / (half_dim - 0.0) + + emb = torch.exp(exponent) + emb = timesteps[:, None].float() * emb[None, :] + + sin_emb = torch.sin(emb) + cos_emb = torch.cos(emb) + emb = torch.cat([cos_emb, sin_emb], dim=-1) + + return rearrange(emb.to(dtype=in_dype), "(b t) d -> b t d", b=timesteps_B_T.shape[0], t=timesteps_B_T.shape[1]) + + +class TimestepEmbedding(nn.Module): + def __init__(self, in_features: int, out_features: int, use_adaln_lora: bool = False): + super().__init__() + log.debug( + f"Using AdaLN LoRA Flag: {use_adaln_lora}. We enable bias if no AdaLN LoRA for backward compatibility." + ) + self.in_dim = in_features + self.out_dim = out_features + self.linear_1 = nn.Linear(in_features, out_features, bias=not use_adaln_lora) + self.activation = nn.SiLU() + self.use_adaln_lora = use_adaln_lora + if use_adaln_lora: + self.linear_2 = nn.Linear(out_features, 3 * out_features, bias=False) + else: + self.linear_2 = nn.Linear(out_features, out_features, bias=False) + + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self.in_dim) + torch.nn.init.trunc_normal_(self.linear_1.weight, std=std, a=-3 * std, b=3 * std) + + std = 1.0 / math.sqrt(self.out_dim) + torch.nn.init.trunc_normal_(self.linear_2.weight, std=std, a=-3 * std, b=3 * std) + + def forward(self, sample: torch.Tensor) -> torch.Tensor: + emb = self.linear_1(sample) + emb = self.activation(emb) + emb = self.linear_2(emb) + + if self.use_adaln_lora: + adaln_lora_B_T_3D = emb + emb_B_T_D = sample + else: + emb_B_T_D = emb + adaln_lora_B_T_3D = None + + return emb_B_T_D, adaln_lora_B_T_3D + + +class FourierFeatures(nn.Module): + """ + Implements a layer that generates Fourier features from input tensors, based on randomly sampled + frequencies and phases. This can help in learning high-frequency functions in low-dimensional problems. + + [B] -> [B, D] + + Parameters: + num_channels (int): The number of Fourier features to generate. + bandwidth (float, optional): The scaling factor for the frequency of the Fourier features. Defaults to 1. + normalize (bool, optional): If set to True, the outputs are scaled by sqrt(2), usually to normalize + the variance of the features. Defaults to False. + + Example: + >>> layer = FourierFeatures(num_channels=256, bandwidth=0.5, normalize=True) + >>> x = torch.randn(10, 256) # Example input tensor + >>> output = layer(x) + >>> print(output.shape) # Expected shape: (10, 256) + """ + + def __init__(self, num_channels, bandwidth=1, normalize=False): + super().__init__() + self.register_buffer("freqs", 2 * np.pi * bandwidth * torch.randn(num_channels), persistent=True) + self.register_buffer("phases", 2 * np.pi * torch.rand(num_channels), persistent=True) + self.gain = np.sqrt(2) if normalize else 1 + self.bandwidth = bandwidth + self.num_channels = num_channels + + self.reset_parameters() + + def reset_parameters(self) -> None: + generator = torch.Generator() + generator.manual_seed(0) + self.freqs = ( + 2 * np.pi * self.bandwidth * torch.randn(self.num_channels, generator=generator).to(self.freqs.device) + ) + self.phases = 2 * np.pi * torch.rand(self.num_channels, generator=generator).to(self.freqs.device) + + def forward(self, x, gain: float = 1.0): + """ + Apply the Fourier feature transformation to the input tensor. + + Args: + x (torch.Tensor): The input tensor. + gain (float, optional): An additional gain factor applied during the forward pass. Defaults to 1. + + Returns: + torch.Tensor: The transformed tensor, with Fourier features applied. + """ + in_dtype = x.dtype + x = x.to(torch.float32).ger(self.freqs.to(torch.float32)).add(self.phases.to(torch.float32)) + x = x.cos().mul(self.gain * gain).to(in_dtype) + return x + + +class PatchEmbed(nn.Module): + """ + PatchEmbed is a module for embedding patches from an input tensor by applying either 3D or 2D convolutional layers, + depending on the . This module can process inputs with temporal (video) and spatial (image) dimensions, + making it suitable for video and image processing tasks. It supports dividing the input into patches + and embedding each patch into a vector of size `out_channels`. + + Parameters: + - spatial_patch_size (int): The size of each spatial patch. + - temporal_patch_size (int): The size of each temporal patch. + - in_channels (int): Number of input channels. Default: 3. + - out_channels (int): The dimension of the embedding vector for each patch. Default: 768. + - bias (bool): If True, adds a learnable bias to the output of the convolutional layers. Default: True. + """ + + def __init__( + self, + spatial_patch_size, + temporal_patch_size, + in_channels=3, + out_channels=768, + ): + super().__init__() + self.spatial_patch_size = spatial_patch_size + self.temporal_patch_size = temporal_patch_size + + self.proj = nn.Sequential( + Rearrange( + "b c (t r) (h m) (w n) -> b t h w (c r m n)", + r=temporal_patch_size, + m=spatial_patch_size, + n=spatial_patch_size, + ), + nn.Linear( + in_channels * spatial_patch_size * spatial_patch_size * temporal_patch_size, out_channels, bias=False + ), + ) + self.dim = in_channels * spatial_patch_size * spatial_patch_size * temporal_patch_size + + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self.dim) + torch.nn.init.trunc_normal_(self.proj[1].weight, std=std, a=-3 * std, b=3 * std) + + def forward(self, x): + """ + Forward pass of the PatchEmbed module. + + Parameters: + - x (torch.Tensor): The input tensor of shape (B, C, T, H, W) where + B is the batch size, + C is the number of channels, + T is the temporal dimension, + H is the height, and + W is the width of the input. + + Returns: + - torch.Tensor: The embedded patches as a tensor, with shape b t h w c. + """ + assert x.dim() == 5 + _, _, T, H, W = x.shape + assert H % self.spatial_patch_size == 0 and W % self.spatial_patch_size == 0, ( + f"H,W {(H, W)} should be divisible by spatial_patch_size {self.spatial_patch_size}" + ) + assert T % self.temporal_patch_size == 0 + x = self.proj(x) + return x + + +class FinalLayer(nn.Module): + """ + The final layer of video DiT. + """ + + def __init__( + self, + hidden_size, + spatial_patch_size, + temporal_patch_size, + out_channels, + use_adaln_lora: bool = False, + adaln_lora_dim: int = 256, + use_wan_fp32_strategy: bool = False, + ): + super().__init__() + self.use_wan_fp32_strategy = use_wan_fp32_strategy + self.layer_norm = nn.LayerNorm(hidden_size, elementwise_affine=False, eps=1e-6) + self.linear = nn.Linear( + hidden_size, spatial_patch_size * spatial_patch_size * temporal_patch_size * out_channels, bias=False + ) + self.hidden_size = hidden_size + self.n_adaln_chunks = 2 + self.use_adaln_lora = use_adaln_lora + self.adaln_lora_dim = adaln_lora_dim + if use_adaln_lora: + self.adaln_modulation = nn.Sequential( + nn.SiLU(), + nn.Linear(hidden_size, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, self.n_adaln_chunks * hidden_size, bias=False), + ) + else: + self.adaln_modulation = nn.Sequential( + nn.SiLU(), nn.Linear(hidden_size, self.n_adaln_chunks * hidden_size, bias=False) + ) + + self.init_weights() + + def init_weights(self) -> None: + std = 1.0 / math.sqrt(self.hidden_size) + torch.nn.init.trunc_normal_(self.linear.weight, std=std, a=-3 * std, b=3 * std) + if self.use_adaln_lora: + torch.nn.init.trunc_normal_(self.adaln_modulation[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.zeros_(self.adaln_modulation[2].weight) + else: + torch.nn.init.zeros_(self.adaln_modulation[1].weight) + + self.layer_norm.reset_parameters() + + def forward( + self, + # x_BT_HW_D, + x_B_T_H_W_D, + emb_B_T_D, + adaln_lora_B_T_3D: Optional[torch.Tensor] = None, + ): + if self.use_wan_fp32_strategy: + assert emb_B_T_D.dtype == torch.float32 + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if self.use_adaln_lora: + assert adaln_lora_B_T_3D is not None + shift_B_T_D, scale_B_T_D = ( + self.adaln_modulation(emb_B_T_D) + adaln_lora_B_T_3D[:, :, : 2 * self.hidden_size] + ).chunk(2, dim=-1) + else: + shift_B_T_D, scale_B_T_D = self.adaln_modulation(emb_B_T_D).chunk(2, dim=-1) + + shift_B_T_1_1_D, scale_B_T_1_1_D = ( + rearrange(shift_B_T_D, "b t d -> b t 1 1 d"), + rearrange(scale_B_T_D, "b t d -> b t 1 1 d"), + ) + + def _fn(_x_B_T_H_W_D, _norm_layer, _scale_B_T_1_1_D, _shift_B_T_1_1_D): + return _norm_layer(_x_B_T_H_W_D) * (1 + _scale_B_T_1_1_D) + _shift_B_T_1_1_D + + x_B_T_H_W_D = _fn(x_B_T_H_W_D, self.layer_norm, scale_B_T_1_1_D, shift_B_T_1_1_D) + x_B_T_H_W_O = self.linear( + x_B_T_H_W_D + ) # O = spatial_patch_size * spatial_patch_size * temporal_patch_size * out_channels + return x_B_T_H_W_O + + +class Block(nn.Module): + """ + A transformer block that combines self-attention, cross-attention and MLP layers with AdaLN modulation. + Each component (self-attention, cross-attention, MLP) has its own layer normalization and AdaLN modulation. + + Parameters: + x_dim (int): Dimension of input features + context_dim (int): Dimension of context features for cross-attention + num_heads (int): Number of attention heads + mlp_ratio (float): Multiplier for MLP hidden dimension. Default: 4.0 + use_adaln_lora (bool): Whether to use AdaLN-LoRA modulation. Default: False + adaln_lora_dim (int): Hidden dimension for AdaLN-LoRA layers. Default: 256 + use_wan_fp32_strategy (bool): Whether to use Wan's FP32 strategy. Default: False + If True, in Attention layer, if do self-attention, q and k will be forced to fp32 before rotary pos emb + also, in modulation computation, force entire computation in fp32 + + The block applies the following sequence: + 1. Self-attention with AdaLN modulation + 2. Cross-attention with AdaLN modulation + 3. MLP with AdaLN modulation + + Each component uses skip connections and layer normalization. + """ + + def __init__( + self, + x_dim: int, + context_dim: int, + num_heads: int, + mlp_ratio: float = 4.0, + use_adaln_lora: bool = False, + adaln_lora_dim: int = 256, + backend: str = "transformer_engine", + image_context_dim: Optional[int] = None, + use_wan_fp32_strategy: bool = False, + ): + super().__init__() + self.x_dim = x_dim + self.layer_norm_self_attn = nn.LayerNorm(x_dim, elementwise_affine=False, eps=1e-6) + self.self_attn = Attention( + x_dim, + None, + num_heads, + x_dim // num_heads, + qkv_format="bshd", + backend=backend, + use_wan_fp32_strategy=use_wan_fp32_strategy, + ) + + self.layer_norm_cross_attn = nn.LayerNorm(x_dim, elementwise_affine=False, eps=1e-6) + + if image_context_dim is None: + self.cross_attn = Attention( + x_dim, context_dim, num_heads, x_dim // num_heads, qkv_format="bshd", backend=backend + ) + else: + self.cross_attn = I2VCrossAttention( + x_dim, + context_dim, + num_heads, + x_dim // num_heads, + img_latent_dim=image_context_dim, + qkv_format="bshd", + backend=backend, + ) + + self.layer_norm_mlp = nn.LayerNorm(x_dim, elementwise_affine=False, eps=1e-6) + self.mlp = GPT2FeedForward(x_dim, int(x_dim * mlp_ratio)) + + self.use_adaln_lora = use_adaln_lora + if self.use_adaln_lora: + self.adaln_modulation_self_attn = nn.Sequential( + nn.SiLU(), + nn.Linear(x_dim, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, 3 * x_dim, bias=False), + ) + self.adaln_modulation_cross_attn = nn.Sequential( + nn.SiLU(), + nn.Linear(x_dim, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, 3 * x_dim, bias=False), + ) + self.adaln_modulation_mlp = nn.Sequential( + nn.SiLU(), + nn.Linear(x_dim, adaln_lora_dim, bias=False), + nn.Linear(adaln_lora_dim, 3 * x_dim, bias=False), + ) + else: + self.adaln_modulation_self_attn = nn.Sequential(nn.SiLU(), nn.Linear(x_dim, 3 * x_dim, bias=False)) + self.adaln_modulation_cross_attn = nn.Sequential(nn.SiLU(), nn.Linear(x_dim, 3 * x_dim, bias=False)) + self.adaln_modulation_mlp = nn.Sequential(nn.SiLU(), nn.Linear(x_dim, 3 * x_dim, bias=False)) + + self.cp_size = None + self.use_wan_fp32_strategy = use_wan_fp32_strategy + + def set_context_parallel_group(self, process_group, ranks, stream, cp_comm_type: str = "p2p"): + self.cp_size = None if ranks is None else len(ranks) + self.self_attn.set_context_parallel_group( + process_group=process_group, + ranks=ranks, + stream=stream, + cp_comm_type=cp_comm_type, + ) + + def reset_parameters(self) -> None: + self.layer_norm_self_attn.reset_parameters() + self.layer_norm_cross_attn.reset_parameters() + self.layer_norm_mlp.reset_parameters() + + if self.use_adaln_lora: + std = 1.0 / math.sqrt(self.x_dim) + torch.nn.init.trunc_normal_(self.adaln_modulation_self_attn[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.adaln_modulation_cross_attn[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.trunc_normal_(self.adaln_modulation_mlp[1].weight, std=std, a=-3 * std, b=3 * std) + torch.nn.init.zeros_(self.adaln_modulation_self_attn[2].weight) + torch.nn.init.zeros_(self.adaln_modulation_cross_attn[2].weight) + torch.nn.init.zeros_(self.adaln_modulation_mlp[2].weight) + else: + torch.nn.init.zeros_(self.adaln_modulation_self_attn[1].weight) + torch.nn.init.zeros_(self.adaln_modulation_cross_attn[1].weight) + torch.nn.init.zeros_(self.adaln_modulation_mlp[1].weight) + + def init_weights(self) -> None: + self.reset_parameters() + self.self_attn.init_weights() + self.cross_attn.init_weights() + self.mlp.init_weights() + + def forward( + self, + x_B_T_H_W_D: torch.Tensor, + emb_B_T_D: torch.Tensor, + crossattn_emb: torch.Tensor, + rope_emb_L_1_1_D: Optional[torch.Tensor] = None, + adaln_lora_B_T_3D: Optional[torch.Tensor] = None, + extra_per_block_pos_emb: Optional[torch.Tensor] = None, + kv_cache_cfg: Optional[KVCacheConfig] = None, + ) -> torch.Tensor: + if extra_per_block_pos_emb is not None: + x_B_T_H_W_D = x_B_T_H_W_D + extra_per_block_pos_emb + + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if self.use_adaln_lora: + shift_self_attn_B_T_D, scale_self_attn_B_T_D, gate_self_attn_B_T_D = ( + self.adaln_modulation_self_attn(emb_B_T_D) + adaln_lora_B_T_3D + ).chunk(3, dim=-1) + shift_cross_attn_B_T_D, scale_cross_attn_B_T_D, gate_cross_attn_B_T_D = ( + self.adaln_modulation_cross_attn(emb_B_T_D) + adaln_lora_B_T_3D + ).chunk(3, dim=-1) + shift_mlp_B_T_D, scale_mlp_B_T_D, gate_mlp_B_T_D = ( + self.adaln_modulation_mlp(emb_B_T_D) + adaln_lora_B_T_3D + ).chunk(3, dim=-1) + else: + shift_self_attn_B_T_D, scale_self_attn_B_T_D, gate_self_attn_B_T_D = self.adaln_modulation_self_attn( + emb_B_T_D + ).chunk(3, dim=-1) + shift_cross_attn_B_T_D, scale_cross_attn_B_T_D, gate_cross_attn_B_T_D = ( + self.adaln_modulation_cross_attn(emb_B_T_D).chunk(3, dim=-1) + ) + shift_mlp_B_T_D, scale_mlp_B_T_D, gate_mlp_B_T_D = self.adaln_modulation_mlp(emb_B_T_D).chunk(3, dim=-1) + + # Reshape tensors from (B, T, D) to (B, T, 1, 1, D) for broadcasting + shift_self_attn_B_T_1_1_D = rearrange(shift_self_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + scale_self_attn_B_T_1_1_D = rearrange(scale_self_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + gate_self_attn_B_T_1_1_D = rearrange(gate_self_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + + shift_cross_attn_B_T_1_1_D = rearrange(shift_cross_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + scale_cross_attn_B_T_1_1_D = rearrange(scale_cross_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + gate_cross_attn_B_T_1_1_D = rearrange(gate_cross_attn_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + + shift_mlp_B_T_1_1_D = rearrange(shift_mlp_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + scale_mlp_B_T_1_1_D = rearrange(scale_mlp_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + gate_mlp_B_T_1_1_D = rearrange(gate_mlp_B_T_D, "b t d -> b t 1 1 d").type_as(x_B_T_H_W_D) + + B, T, H, W, D = x_B_T_H_W_D.shape + + def _fn(_x_B_T_H_W_D, _norm_layer, _scale_B_T_1_1_D, _shift_B_T_1_1_D): + return _norm_layer(_x_B_T_H_W_D) * (1 + _scale_B_T_1_1_D) + _shift_B_T_1_1_D + + normalized_x_B_T_H_W_D = _fn( + x_B_T_H_W_D, + self.layer_norm_self_attn, + scale_self_attn_B_T_1_1_D, + shift_self_attn_B_T_1_1_D, + ) + + video_size = VideoSize(T=T, H=H, W=W) + + # (ahassani): Hack to correct `video_size` when CP is enabled. + # I really don't like this, but there doesn't seem to be any central + # piece of code that's responsible for handling CP/TP that also defines the + # layout of shardings. Other parts of the code (i.e. RoPE) seem to make this + # assumption that CP sharding is always done along T. + if self.cp_size is not None and self.cp_size > 1: + video_size = VideoSize(T=T * self.cp_size, H=H, W=W) + + result_B_T_H_W_D = rearrange( + self.self_attn( + # normalized_x_B_T_HW_D, + rearrange(normalized_x_B_T_H_W_D, "b t h w d -> b (t h w) d"), + None, + rope_emb=rope_emb_L_1_1_D, + video_size=video_size, + kv_cache_cfg=kv_cache_cfg, + ), + "b (t h w) d -> b t h w d", + t=T, + h=H, + w=W, + ) + x_B_T_H_W_D = x_B_T_H_W_D + gate_self_attn_B_T_1_1_D * result_B_T_H_W_D + + def _x_fn( + _x_B_T_H_W_D, + layer_norm_cross_attn, + _scale_cross_attn_B_T_1_1_D, + _shift_cross_attn_B_T_1_1_D, + _gate_cross_attn_B_T_1_1_D, + ): + _normalized_x_B_T_H_W_D = _fn( + _x_B_T_H_W_D, layer_norm_cross_attn, _scale_cross_attn_B_T_1_1_D, _shift_cross_attn_B_T_1_1_D + ) + _result_B_T_H_W_D = rearrange( + self.cross_attn( + rearrange(_normalized_x_B_T_H_W_D, "b t h w d -> b (t h w) d"), + crossattn_emb, + rope_emb=rope_emb_L_1_1_D, + ), + "b (t h w) d -> b t h w d", + t=T, + h=H, + w=W, + ) + # _x_B_T_H_W_D = _x_B_T_H_W_D + _gate_cross_attn_B_T_1_1_D * _result_B_T_H_W_D + return _result_B_T_H_W_D + + result_B_T_H_W_D = _x_fn( + x_B_T_H_W_D, + self.layer_norm_cross_attn, + scale_cross_attn_B_T_1_1_D, + shift_cross_attn_B_T_1_1_D, + gate_cross_attn_B_T_1_1_D, + ) + x_B_T_H_W_D = result_B_T_H_W_D * gate_cross_attn_B_T_1_1_D + x_B_T_H_W_D + + normalized_x_B_T_H_W_D = _fn( + x_B_T_H_W_D, + self.layer_norm_mlp, + scale_mlp_B_T_1_1_D, + shift_mlp_B_T_1_1_D, + ) + result_B_T_H_W_D = self.mlp(normalized_x_B_T_H_W_D) + x_B_T_H_W_D = x_B_T_H_W_D + gate_mlp_B_T_1_1_D * result_B_T_H_W_D + return x_B_T_H_W_D + + +class MiniTrainDIT(WeightTrainingStat): + """ + A clean impl of DIT that can load and reproduce the training results of the original DIT model in edify_video/v4~(cosmos 1) + A general implementation of adaln-modulated VIT-like~(DiT) transformer for video processing. + + Args: + max_img_h (int): Maximum height of the input images. + max_img_w (int): Maximum width of the input images. + max_frames (int): Maximum number of frames in the video sequence. + in_channels (int): Number of input channels (e.g., RGB channels for color images). + out_channels (int): Number of output channels. + patch_spatial (int): Spatial resolution of patches for input processing. + patch_temporal (int): Temporal resolution of patches for input processing. + concat_padding_mask (bool): If True, includes a mask channel in the input to handle padding. + model_channels (int): Base number of channels used throughout the model. + num_blocks (int): Number of transformer blocks. + num_heads (int): Number of heads in the multi-head attention layers. + mlp_ratio (float): Expansion ratio for MLP blocks. + crossattn_emb_channels (int): Number of embedding channels for cross-attention. + extra_image_context_dim (int): Number of embedding channels for extra image context. + pos_emb_cls (str): Type of positional embeddings. + pos_emb_learnable (bool): Whether positional embeddings are learnable. + pos_emb_interpolation (str): Method for interpolating positional embeddings. + min_fps (int): Minimum frames per second. + max_fps (int): Maximum frames per second. + use_adaln_lora (bool): Whether to use AdaLN-LoRA. + adaln_lora_dim (int): Dimension for AdaLN-LoRA. + rope_h_extrapolation_ratio (float): Height extrapolation ratio for RoPE. + rope_w_extrapolation_ratio (float): Width extrapolation ratio for RoPE. + rope_t_extrapolation_ratio (float): Temporal extrapolation ratio for RoPE. + extra_per_block_abs_pos_emb (bool): Whether to use extra per-block absolute positional embeddings. + extra_h_extrapolation_ratio (float): Height extrapolation ratio for extra embeddings. + extra_w_extrapolation_ratio (float): Width extrapolation ratio for extra embeddings. + extra_t_extrapolation_ratio (float): Temporal extrapolation ratio for extra embeddings. + n_dense_blocks (`int`, *optional*, defaults to -1): + Number of blocks that will remain dense (not replaced with sparse attention) + If -1, no blocks are replaced with sparse attention + If 0, all blocks use sparse attention + Otherwise, n_dense_blocks blocks will remain dense, distributed evenly across the network + natten_parameters (`dict`, *optional*, defaults to None): + NATTEN (Sparse attention) parameter list. + The list length must be the same as the number of layers, with each list element + indicating NATTEN parameters for that layer. If None, NATTEN will not be used in that + layer and it would remain a full dense self attention. If not None, it must be a + dictionary/mapping with at least the following key: + - window_size: `tuple` of size 3 indicating neighborhood attention window size. + window size of -1 along any dimension means self attention. + Other optional parameters and their keys: + - stride: `tuple` of size 3 indicating neighborhood attention stride value. + stride = 1 is standard neighborhood attention, stride = window size means + blocked/window self attention (WSA) along that dimension. Any other values are + strided neighborhood attention. Refer to the GNA paper for more information. + + - dilation: `tuple` of size 3 indicating neighborhood attention dilation value. + dilation = 1 is standard neighborhood attention. Refer to the DiNAT paper for more + information. + + - is_causal: `tuple` of 3 booleans indicating whether causal masking is enabled for + any of the T, H, W dimensions. + """ + + def __init__( + self, + max_img_h: int, + max_img_w: int, + max_frames: int, + in_channels: int, + out_channels: int, + patch_spatial: int, + patch_temporal: int, + concat_padding_mask: bool = True, + # attention settings + model_channels: int = 768, + num_blocks: int = 10, + num_heads: int = 16, + mlp_ratio: float = 4.0, + atten_backend: str = "transformer_engine", + # cross attention settings + crossattn_emb_channels: int = 1024, + use_crossattn_projection: bool = False, + crossattn_proj_in_channels: int = 1024, + extra_image_context_dim: Optional[int] = None, + # positional embedding settings + pos_emb_cls: str = "sincos", + pos_emb_learnable: bool = False, + pos_emb_interpolation: str = "crop", + min_fps: int = 1, + max_fps: int = 30, + use_adaln_lora: bool = False, + adaln_lora_dim: int = 256, + rope_h_extrapolation_ratio: float = 1.0, + rope_w_extrapolation_ratio: float = 1.0, + rope_t_extrapolation_ratio: float = 1.0, + extra_per_block_abs_pos_emb: bool = False, + extra_h_extrapolation_ratio: float = 1.0, + extra_w_extrapolation_ratio: float = 1.0, + extra_t_extrapolation_ratio: float = 1.0, + rope_enable_fps_modulation: bool = True, + sac_config: SACConfig = SACConfig(), + n_dense_blocks: int = -1, + natten_parameters: Union[dict, list] = None, + # if True, will closely match wan's strategy to use fp32 in certain layers/operations + use_wan_fp32_strategy: bool = False, + ) -> None: + super().__init__() + self.max_img_h = max_img_h + self.max_img_w = max_img_w + self.max_frames = max_frames + self.in_channels = in_channels + self.out_channels = out_channels + self.patch_spatial = patch_spatial + self.patch_temporal = patch_temporal + self.num_heads = num_heads + self.num_blocks = num_blocks + self.model_channels = model_channels + self.concat_padding_mask = concat_padding_mask + self.atten_backend = atten_backend + # positional embedding settings + self.pos_emb_cls = pos_emb_cls + self.pos_emb_learnable = pos_emb_learnable + self.pos_emb_interpolation = pos_emb_interpolation + self.min_fps = min_fps + self.max_fps = max_fps + self.rope_h_extrapolation_ratio = rope_h_extrapolation_ratio + self.rope_w_extrapolation_ratio = rope_w_extrapolation_ratio + self.rope_t_extrapolation_ratio = rope_t_extrapolation_ratio + self.extra_per_block_abs_pos_emb = extra_per_block_abs_pos_emb + self.extra_h_extrapolation_ratio = extra_h_extrapolation_ratio + self.extra_w_extrapolation_ratio = extra_w_extrapolation_ratio + self.extra_t_extrapolation_ratio = extra_t_extrapolation_ratio + self.rope_enable_fps_modulation = rope_enable_fps_modulation + self.extra_image_context_dim = extra_image_context_dim + self.build_patch_embed() + self.build_pos_embed() + self.use_adaln_lora = use_adaln_lora + self.adaln_lora_dim = adaln_lora_dim + self.t_embedder = nn.Sequential( + Timesteps(model_channels), + TimestepEmbedding(model_channels, model_channels, use_adaln_lora=use_adaln_lora), + ) + self.use_crossattn_projection = use_crossattn_projection + self.crossattn_proj_in_channels = crossattn_proj_in_channels + self.use_wan_fp32_strategy = use_wan_fp32_strategy + + self.blocks = nn.ModuleList( + [ + Block( + x_dim=model_channels, + context_dim=crossattn_emb_channels, + num_heads=num_heads, + mlp_ratio=mlp_ratio, + use_adaln_lora=use_adaln_lora, + adaln_lora_dim=adaln_lora_dim, + backend=atten_backend, + image_context_dim=None if extra_image_context_dim is None else model_channels, + use_wan_fp32_strategy=use_wan_fp32_strategy, + ) + for _ in range(num_blocks) + ] + ) + + self.final_layer = FinalLayer( + hidden_size=self.model_channels, + spatial_patch_size=self.patch_spatial, + temporal_patch_size=self.patch_temporal, + out_channels=self.out_channels, + use_adaln_lora=self.use_adaln_lora, + adaln_lora_dim=self.adaln_lora_dim, + use_wan_fp32_strategy=self.use_wan_fp32_strategy, + ) + + self.t_embedding_norm = te.pytorch.RMSNorm(model_channels, eps=1e-6) + if extra_image_context_dim is not None: + self.img_context_proj = nn.Sequential( + nn.Linear( + extra_image_context_dim, model_channels, bias=True + ), # help distinguish between image and video context + nn.GELU(), + ) + + if use_crossattn_projection: + self.crossattn_proj = nn.Sequential( + nn.Linear(crossattn_proj_in_channels, crossattn_emb_channels, bias=True), + nn.GELU(), + ) + + self.init_weights() + self.enable_selective_checkpoint(sac_config, self.blocks) + + # Replace self-attention with sparse attention if specified + if n_dense_blocks != -1: + self = replace_selfattn_op_with_sparse_attn_op(self, n_dense_blocks, natten_parameters=natten_parameters) + + self._is_context_parallel_enabled = False + + def init_weights(self): + self.x_embedder.init_weights() + self.pos_embedder.reset_parameters() + if self.extra_per_block_abs_pos_emb: + self.extra_pos_embedder.reset_parameters() + + self.t_embedder[1].init_weights() + for block in self.blocks: + block.init_weights() + + self.final_layer.init_weights() + self.t_embedding_norm.reset_parameters() + + if self.extra_image_context_dim is not None: + self.img_context_proj[0].reset_parameters() + + def build_patch_embed(self): + ( + concat_padding_mask, + in_channels, + patch_spatial, + patch_temporal, + model_channels, + ) = ( + self.concat_padding_mask, + self.in_channels, + self.patch_spatial, + self.patch_temporal, + self.model_channels, + ) + in_channels = in_channels + 1 if concat_padding_mask else in_channels + self.x_embedder = PatchEmbed( + spatial_patch_size=patch_spatial, + temporal_patch_size=patch_temporal, + in_channels=in_channels, + out_channels=model_channels, + ) + + def build_pos_embed(self): + if self.pos_emb_cls == "rope3d": + cls_type = VideoRopePosition3DEmb + else: + raise ValueError(f"Unknown pos_emb_cls {self.pos_emb_cls}") + + log.debug(f"Building positional embedding with {self.pos_emb_cls} class, impl {cls_type}") + kwargs = dict( + model_channels=self.model_channels, + len_h=self.max_img_h // self.patch_spatial, + len_w=self.max_img_w // self.patch_spatial, + len_t=self.max_frames // self.patch_temporal, + max_fps=self.max_fps, + min_fps=self.min_fps, + is_learnable=self.pos_emb_learnable, + interpolation=self.pos_emb_interpolation, + head_dim=self.model_channels // self.num_heads, + h_extrapolation_ratio=self.rope_h_extrapolation_ratio, + w_extrapolation_ratio=self.rope_w_extrapolation_ratio, + t_extrapolation_ratio=self.rope_t_extrapolation_ratio, + enable_fps_modulation=self.rope_enable_fps_modulation, + ) + self.pos_embedder = cls_type( + **kwargs, + ) + + if self.extra_per_block_abs_pos_emb: + kwargs["h_extrapolation_ratio"] = self.extra_h_extrapolation_ratio + kwargs["w_extrapolation_ratio"] = self.extra_w_extrapolation_ratio + kwargs["t_extrapolation_ratio"] = self.extra_t_extrapolation_ratio + self.extra_pos_embedder = LearnablePosEmbAxis( + **kwargs, + ) + + def prepare_embedded_sequence( + self, + x_B_C_T_H_W: torch.Tensor, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[torch.Tensor]]: + """ + Prepares an embedded sequence tensor by applying positional embeddings and handling padding masks. + + Args: + x_B_C_T_H_W (torch.Tensor): video + fps (Optional[torch.Tensor]): Frames per second tensor to be used for positional embedding when required. + If None, a default value (`self.base_fps`) will be used. + padding_mask (Optional[torch.Tensor]): current it is not used + + Returns: + Tuple[torch.Tensor, Optional[torch.Tensor]]: + - A tensor of shape (B, T, H, W, D) with the embedded sequence. + - An optional positional embedding tensor, returned only if the positional embedding class + (`self.pos_emb_cls`) includes 'rope'. Otherwise, None. + + Notes: + - If `self.concat_padding_mask` is True, a padding mask channel is concatenated to the input tensor. + - The method of applying positional embeddings depends on the value of `self.pos_emb_cls`. + - If 'rope' is in `self.pos_emb_cls` (case insensitive), the positional embeddings are generated using + the `self.pos_embedder` with the shape [T, H, W]. + - If "fps_aware" is in `self.pos_emb_cls`, the positional embeddings are generated using the + `self.pos_embedder` with the fps tensor. + - Otherwise, the positional embeddings are generated without considering fps. + """ + if self.concat_padding_mask: + padding_mask = transforms.functional.resize( + padding_mask, list(x_B_C_T_H_W.shape[-2:]), interpolation=transforms.InterpolationMode.NEAREST + ) + x_B_C_T_H_W = torch.cat( + [x_B_C_T_H_W, padding_mask.unsqueeze(1).repeat(1, 1, x_B_C_T_H_W.shape[2], 1, 1)], dim=1 + ) + x_B_T_H_W_D = self.x_embedder(x_B_C_T_H_W) + + if self.extra_per_block_abs_pos_emb: + extra_pos_emb = self.extra_pos_embedder(x_B_T_H_W_D, fps=fps) + else: + extra_pos_emb = None + + if "rope" in self.pos_emb_cls.lower(): + return x_B_T_H_W_D, self.pos_embedder(x_B_T_H_W_D, fps=fps), extra_pos_emb + x_B_T_H_W_D = x_B_T_H_W_D + self.pos_embedder(x_B_T_H_W_D) # [B, T, H, W, D] + + return x_B_T_H_W_D, None, extra_pos_emb + + def unpatchify(self, x_B_T_H_W_M): + x_B_C_Tt_Hp_Wp = rearrange( + x_B_T_H_W_M, + "B T H W (p1 p2 t C) -> B C (T t) (H p1) (W p2)", + p1=self.patch_spatial, + p2=self.patch_spatial, + t=self.patch_temporal, + ) + return x_B_C_Tt_Hp_Wp + + def forward( + self, + x_B_C_T_H_W: torch.Tensor, + timesteps_B_T: torch.Tensor, + crossattn_emb: torch.Tensor, + fps: Optional[torch.Tensor] = None, + padding_mask: Optional[torch.Tensor] = None, + data_type: Optional[DataType] = DataType.VIDEO, + intermediate_feature_ids: Optional[List[int]] = None, + img_context_emb: Optional[torch.Tensor] = None, + ) -> torch.Tensor | List[torch.Tensor] | Tuple[torch.Tensor, List[torch.Tensor]]: + """ + Args: + x: (B, C, T, H, W) tensor of spatial-temp inputs + timesteps: (B, ) tensor of timesteps + crossattn_emb: (B, N, D) tensor of cross-attention embeddings + """ + assert isinstance(data_type, DataType), ( + f"Expected DataType, got {type(data_type)}. We need discuss this flag later." + ) + x_B_T_H_W_D, rope_emb_L_1_1_D, extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D = self.prepare_embedded_sequence( + x_B_C_T_H_W, + fps=fps, + padding_mask=padding_mask, + ) + + if self.use_crossattn_projection: + crossattn_emb = self.crossattn_proj(crossattn_emb) + + if img_context_emb is not None: + assert self.extra_image_context_dim is not None, ( + "extra_image_context_dim must be set if img_context_emb is provided" + ) + img_context_emb = self.img_context_proj(img_context_emb) + context_input = (crossattn_emb, img_context_emb) + else: + context_input = crossattn_emb + + with amp.autocast("cuda", enabled=self.use_wan_fp32_strategy, dtype=torch.float32): + if timesteps_B_T.ndim == 1: + timesteps_B_T = timesteps_B_T.unsqueeze(1) + t_embedding_B_T_D, adaln_lora_B_T_3D = self.t_embedder(timesteps_B_T) + t_embedding_B_T_D = self.t_embedding_norm(t_embedding_B_T_D) + + # for logging purpose + affline_scale_log_info = {} + affline_scale_log_info["t_embedding_B_T_D"] = t_embedding_B_T_D.detach() + self.affline_scale_log_info = affline_scale_log_info + self.affline_emb = t_embedding_B_T_D + self.crossattn_emb = crossattn_emb + + if extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D is not None: + assert x_B_T_H_W_D.shape == extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D.shape, ( + f"{x_B_T_H_W_D.shape} != {extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D.shape}" + ) + + B, T, H, W, D = x_B_T_H_W_D.shape + # x_B_THW_D = rearrange(x_B_T_H_W_D, "b t h w d -> b (t h w) d") + + intermediate_features_outputs = [] + for i, block in enumerate(self.blocks): + x_B_T_H_W_D = block( + x_B_T_H_W_D, + t_embedding_B_T_D, + context_input, + rope_emb_L_1_1_D=rope_emb_L_1_1_D, + adaln_lora_B_T_3D=adaln_lora_B_T_3D, + extra_per_block_pos_emb=extra_pos_emb_B_T_H_W_D_or_T_H_W_B_D, + ) + if intermediate_feature_ids and i in intermediate_feature_ids: + x_reshaped_for_disc = rearrange(x_B_T_H_W_D, "b tp hp wp d -> b (tp hp wp) d") + intermediate_features_outputs.append(x_reshaped_for_disc) + + # x_B_T_H_W_D = rearrange(x_B_THW_D, "b (t h w) d -> b t h w d", t=T, h=H, w=W) + # O = out_channels * spatial_patch_size * spatial_patch_size * temporal_patch_size + x_B_T_H_W_O = self.final_layer(x_B_T_H_W_D, t_embedding_B_T_D, adaln_lora_B_T_3D=adaln_lora_B_T_3D) + x_B_C_Tt_Hp_Wp = self.unpatchify(x_B_T_H_W_O) + if intermediate_feature_ids: + if len(intermediate_features_outputs) != len(intermediate_feature_ids): + log.warning( + f"Collected {len(intermediate_features_outputs)} intermediate features, " + f"but expected {len(intermediate_feature_ids)}. " + f"Requested IDs: {intermediate_feature_ids}" + ) + return x_B_C_Tt_Hp_Wp, intermediate_features_outputs + + return x_B_C_Tt_Hp_Wp + + def enable_selective_checkpoint(self, sac_config: SACConfig, blocks: nn.ModuleList): + if sac_config.mode == CheckpointMode.NONE: + return self + + log.info( + f"Enable selective checkpoint with {sac_config.mode}, for every {sac_config.every_n_blocks} blocks. Total blocks: {len(blocks)}" + ) + _context_fn = sac_config.get_context_fn() + for block_id, block in blocks.named_children(): + if int(block_id) % sac_config.every_n_blocks == 0: + log.info(f"Enable selective checkpoint for block {block_id}") + block = ptd_checkpoint_wrapper( + block, + context_fn=_context_fn, + preserve_rng_state=False, + ) + blocks.register_module(block_id, block) + self.register_module( + "final_layer", + ptd_checkpoint_wrapper( + self.final_layer, + context_fn=_context_fn, + preserve_rng_state=False, + ), + ) + + return self + + def fully_shard(self, mesh): + for i, block in enumerate(self.blocks): + reshard_after_forward = i < len(self.blocks) - 1 + fully_shard(block, mesh=mesh, reshard_after_forward=reshard_after_forward) + + fully_shard(self.final_layer, mesh=mesh, reshard_after_forward=True) + if self.extra_per_block_abs_pos_emb: + fully_shard(self.extra_pos_embedder, mesh=mesh, reshard_after_forward=True) + fully_shard(self.t_embedder, mesh=mesh, reshard_after_forward=False) + if self.extra_image_context_dim is not None: + fully_shard(self.img_context_proj, mesh=mesh, reshard_after_forward=False) + + def disable_context_parallel(self): + # pos_embedder + self.pos_embedder.disable_context_parallel() + if self.extra_per_block_abs_pos_emb: + self.extra_pos_embedder.disable_context_parallel() + + # attention + for block in self.blocks: + block.set_context_parallel_group( + process_group=None, + ranks=None, + stream=torch.cuda.Stream(), + ) + + self._is_context_parallel_enabled = False + + def enable_context_parallel(self, process_group: Optional[ProcessGroup] = None): + # pos_embedder + self.pos_embedder.enable_context_parallel(process_group=process_group) + if self.extra_per_block_abs_pos_emb: + self.extra_pos_embedder.enable_context_parallel(process_group=process_group) + + # attention + cp_ranks = get_process_group_ranks(process_group) + for block in self.blocks: + block.set_context_parallel_group( + process_group=process_group, + ranks=cp_ranks, + stream=torch.cuda.Stream(), + ) + + self._is_context_parallel_enabled = True + + @property + def is_context_parallel_enabled(self): + return self._is_context_parallel_enabled + + +def replace_selfattn_op_with_sparse_attn_op( + model: MiniTrainDIT, n_dense_blocks: int = 0, natten_parameters: Union[dict, list] = None +) -> MiniTrainDIT: + """ + Replace the self-attention operator with a sparse self-attention operator. + + Args: + model: MiniTrainDIT instance + n_dense_blocks: Number of blocks that will remain dense (not replaced with NeighborhoodAttention) + If 0, all blocks use NeighborhoodAttention. + If -1, return model directly without any modifications. + Otherwise, n_dense_blocks blocks will remain dense, distributed evenly across the network. + + Returns: + Modified instance + """ + # Special case: return model directly without modifications + if n_dense_blocks == -1: + return model + + num_blocks = len(model.blocks) + + if natten_parameters is None: + raise ValueError("Please specify natten_parameters when n_dense_blocks > -1.") + + if isinstance(natten_parameters, Sequence) and len(natten_parameters) != num_blocks: + raise ValueError( + "List of NATTEN parameters must be the same length as the number of blocks, " + f"got {len(natten_parameters)=} != {num_blocks=}." + ) + + if isinstance(natten_parameters, Sequence) and n_dense_blocks > 0: + log.warning(f"NATTEN parameters was a list; ignoring {n_dense_blocks=}.") + + if isinstance(natten_parameters, Sequence): + natten_parameters_list = natten_parameters + else: + if n_dense_blocks >= num_blocks: + raise ValueError(f"n_dense_blocks ({n_dense_blocks}) must be less than the number of blocks ({num_blocks})") + + # Determine which blocks should remain dense + dense_indices = set() + + if n_dense_blocks > 0: + # General rule: distribute n_dense_blocks blocks evenly across the network + if n_dense_blocks == 1: + # Special case: just the middle block + dense_indices.add(num_blocks // 2) + else: + # For multiple blocks, distribute them evenly from start to end + indices = np.linspace(0, num_blocks - 1, n_dense_blocks, dtype=int) + dense_indices.update(indices.tolist()) + + natten_parameters_list = [None if i in dense_indices else natten_parameters for i in range(num_blocks)] + + # Replace self-attention with NeighborhoodAttention for non-dense blocks + for i, block in enumerate(model.blocks): + natten_params = natten_parameters_list[i] + if natten_params is not None: + natten_parameters_layer = {k: v for k, v in natten_params.items()} + natten_parameters_layer["layer_id"] = i + if block.self_attn.backend == "minimal_a2a": + sparse_attn_op = NattenA2AAttnOp(natten_parameters=natten_parameters_layer) + else: + raise NotImplementedError( + f"Using sparsity with attention backend {block.self_attn.backend} is not supported." + ) + + block.self_attn.register_module("attn_op", sparse_attn_op) + + return model diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/minimal_v4_dit_test_sparse_attn_e2e_speedup.py b/REGEN-main/cosmos_policy/_src/predict2/networks/minimal_v4_dit_test_sparse_attn_e2e_speedup.py new file mode 100644 index 0000000000000000000000000000000000000000..47dc1f6ce0b9bea6dce61164ea91197c5901f411 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/minimal_v4_dit_test_sparse_attn_e2e_speedup.py @@ -0,0 +1,375 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import time + +import torch + +from cosmos_policy._src.imaginaire.lazy_config import instantiate +from cosmos_policy._src.predict2.conditioner import DataType +from cosmos_policy._src.predict2.configs.text2world.defaults.net import ( + COSMOS_V1_2B_NET_MININET, + COSMOS_V1_14B_NET_MININET, +) +from cosmos_policy._src.predict2.datasets.utils import VIDEO_RES_SIZE_INFO +from cosmos_policy._src.predict2.networks.minimal_v4_dit import MiniTrainDIT, replace_selfattn_op_with_sparse_attn_op + +natten_parameters_90pct = {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)} + +natten_parameters_2b_comb01 = [ + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 0, 90% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 1, 50% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 2, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 3, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 4, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 5, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 6, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 7, 90% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 8, 90% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 9, 50% + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 10, 90% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 11, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 12, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 13, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 14, 50% + None, # blk 15, SA + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 16, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 17, 50% + None, # blk 18, SA + None, # blk 19, SA + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 20, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 21, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 22, 50% + None, # blk 23, SA + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 24, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 25, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 26, 50% + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 27, 50% +] + +natten_parameters_2b_comb02 = [ + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 0 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 1 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 2 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 3 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 4 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 5 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 6 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 7 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 8 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 9 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 10 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 11 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 12 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 13 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 14 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 15 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 16 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 17 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 18 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 19 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 20 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 21 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 22 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 23 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 24 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 25 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 26 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 27 +] + +natten_parameters_2b_comb03 = [ + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 0 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 1 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 2 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 3 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 4 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 5 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 6 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 7 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 8 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 9 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 10 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 11 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 12 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 13 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 14 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 15 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 16 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 17 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 18 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 19 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 20 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 21 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 22 + None, # blk 23 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 24 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 25 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 26 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 27 +] + +natten_parameters_2b_comb04 = [ + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 0 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 1 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 2 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 3 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 4 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 5 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 6 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 7 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 8 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 9 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 10 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 11 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 12 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 13 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 14 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 15 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 16 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 17 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 18 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 19 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 20 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 21 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 22 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 23 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 24 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 25 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 26 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 27 +] + +natten_parameters_2b_comb05 = [ + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 0 + {"window_size": (-1, 4, 24), "stride": (1, 1, 8), "dilation": (1, 11, 1), "base_size": (-1, 44, 80)}, # blk 1 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 2 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 3 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 4 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 5 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 6 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 7 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 8 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 9 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 10 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 11 + None, # blk 12 + None, # blk 13 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 14 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 15 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 16 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 17 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 18 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 19 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 20 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 21 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 22 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 23 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 24 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 25 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 26 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 27 +] + +natten_parameters_14b_comb01 = [ + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 0 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 1 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 2 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 3 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 4 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 5 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 6 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 7 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 8 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 9 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 10 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 11 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 12 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 13 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 14 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 15 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 16 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 17 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 18 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 19 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 20 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 21 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 22 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 23 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 24 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 25 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 26 + None, # blk 27 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 28 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 29 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 30 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 31 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 32 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 33 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 34 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 35 +] + +natten_parameters_14b_comb02 = [ + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 0 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 1 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 2 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 3 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 4 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 5 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 6 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 7 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 8 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 9 + {"window_size": (-1, 4, 16), "stride": (1, 1, 1), "dilation": (1, 11, 5), "base_size": (-1, 44, 80)}, # blk 10 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 11 + {"window_size": (-1, 12, 16), "stride": (1, 4, 1), "dilation": (1, 1, 5), "base_size": (-1, 44, 80)}, # blk 12 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 13 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 14 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 15 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 16 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 17 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 18 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 19 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 20 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 21 + {"window_size": (-1, 12, 24), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 22 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 23 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 24 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 25 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 26 + None, # blk 27 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 28 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 29 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 30 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 31 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 32 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 33 + {"window_size": (-1, 28, 56), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 34 + {"window_size": (-1, 20, 40), "stride": (1, 4, 8), "base_size": (-1, 44, 80)}, # blk 35 +] + +""" +Forward pass only +""" + + +def measure_e2e_perf(model, inputs, warmup_iters, iters): + torch.cuda.synchronize() + for _ in range(warmup_iters): + output = model(**inputs) + torch.cuda.synchronize() + + start_time = time.perf_counter() + + for _ in range(iters): + output = model(**inputs) + torch.cuda.synchronize() + + end_time = time.perf_counter() + + e2e_time_total = end_time - start_time + + avg_e2e_time = e2e_time_total / iters + + print(f"E2E time avg: {avg_e2e_time:.2f} s") + + +@torch.no_grad() +def test_inference_e2e(): + try: + import natten # noqa: F401 + except ImportError: + print("NATTEN is not installed, skipping...") + return + + batch_size = 1 + warmup_iters = 5 + iters = 5 + + res = "720" + video_size_options = VIDEO_RES_SIZE_INFO[res] + H, W = 704, 1280 + T_list = [ + 24, + ] + dH, dW = 8, 8 + pH, pW = 2, 2 + + COSMOS_V1_2B_NET_MININET.atten_backend = "minimal_a2a" + COSMOS_V1_14B_NET_MININET.atten_backend = "minimal_a2a" + + for net_str, net_cfg, n_dense_blocks, natten_parameters in [ + # ("2B", COSMOS_V1_2B_NET_MININET, -1, {}), + # ("2B", COSMOS_V1_2B_NET_MININET, 4, natten_parameters_90pct), + # ("2B", COSMOS_V1_2B_NET_MININET, 7, natten_parameters_90pct), + # ("2B", COSMOS_V1_2B_NET_MININET, 9, natten_parameters_90pct), + # ("2B", COSMOS_V1_2B_NET_MININET, 0, natten_parameters_2b_comb01), + # ("2B", COSMOS_V1_2B_NET_MININET, 0, natten_parameters_2b_comb02), + # ("2B", COSMOS_V1_2B_NET_MININET, 0, natten_parameters_2b_comb03), + # ("2B", COSMOS_V1_2B_NET_MININET, 0, natten_parameters_2b_comb04), + # ("2B", COSMOS_V1_2B_NET_MININET, 0, natten_parameters_2b_comb05), + ("14B", COSMOS_V1_14B_NET_MININET, -1, {}), + # ("14B", COSMOS_V1_14B_NET_MININET, 5, natten_parameters_90pct), + # ("14B", COSMOS_V1_14B_NET_MININET, 7, natten_parameters_90pct), + # ("14B", COSMOS_V1_14B_NET_MININET, 9, natten_parameters_90pct), + # ("14B", COSMOS_V1_14B_NET_MININET, 12, natten_parameters_90pct), + # ("14B", COSMOS_V1_14B_NET_MININET, 0, natten_parameters_14b_comb01), + ("14B", COSMOS_V1_14B_NET_MININET, 0, natten_parameters_14b_comb02), + ]: + torch.cuda.empty_cache() + print() + print() + print(f"Model: Minimal v4 DiT - {net_str}") + + replace_self_attn = n_dense_blocks >= 0 + model: MiniTrainDIT = instantiate(net_cfg).cuda().bfloat16() + model.eval() + if replace_self_attn: + model = replace_selfattn_op_with_sparse_attn_op( + model, n_dense_blocks=n_dense_blocks, natten_parameters=natten_parameters + ) + print() + print(f"Replaced Self Attention with NATTEN: {n_dense_blocks=}, {natten_parameters=}") + else: + print() + print("Self Attention case (unmodified)") + + for T in T_list: + T_, H_, W_ = T, H // dH, W // dW + print( + f"Res: {T=}, {H=}, {W=}; DiT input: ({T_}, {H_}, {W_}); feature map size: ({T}, {H_ // pH}, {W_ // pW})" + ) + video_example_input = { + "x_B_C_T_H_W": torch.randn(batch_size, 16, T_, H_, W_, dtype=torch.bfloat16, device="cuda"), + "timesteps_B_T": torch.randn(batch_size, dtype=torch.bfloat16, device="cuda"), + "crossattn_emb": torch.randn(batch_size, 512, 1024, dtype=torch.bfloat16, device="cuda"), + "fps": torch.randint(size=(batch_size,), low=2, high=30, device="cuda"), + "padding_mask": torch.randn(batch_size, 1, H_, W_, dtype=torch.bfloat16, device="cuda"), + "data_type": DataType.VIDEO, + } + + output = model(**video_example_input) + assert output.shape == (batch_size, 16, T_, H_, W_) + + print("Video model") + measure_e2e_perf(model, video_example_input, warmup_iters=warmup_iters, iters=iters) + print() + + del model + + +if __name__ == "__main__": + test_inference_e2e() diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/model_weights_stats.py b/REGEN-main/cosmos_policy/_src/predict2/networks/model_weights_stats.py new file mode 100644 index 0000000000000000000000000000000000000000..4b5c266907c44b9759ab79a8f67f1ffcfa6d8db9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/model_weights_stats.py @@ -0,0 +1,64 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from abc import ABC, abstractmethod +from dataclasses import dataclass +from typing import Any + +import torch +from torch import nn + + +@dataclass +class TrainingStats: + """Data class to hold training statistics.""" + + video_samples: int = 0 + image_samples: int = 0 + iterations: int = 0 + training_hours: float = 0.0 + + +class WeightTrainingStat(nn.Module, ABC): + """Abstract base class for tracking training statistics.""" + + def __init__(self) -> None: + super().__init__() + self._initialize_tracking_buffers() + + def _initialize_tracking_buffers(self) -> None: + """Initialize tracking buffers with default values.""" + tracking_buffers = { + "accum_video_sample_counter": torch.tensor(0, dtype=torch.int64), + "accum_image_sample_counter": torch.tensor(0, dtype=torch.int64), + "accum_iteration": torch.tensor(0, dtype=torch.int64), + "accum_train_in_hours": torch.tensor(0.0, dtype=torch.float32), + } + + for name, tensor in tracking_buffers.items(): + self.register_buffer(name, tensor) + + def get_training_stats(self) -> TrainingStats: + """Return current training statistics.""" + return TrainingStats( + video_samples=self.accum_video_sample_counter.item(), + image_samples=self.accum_image_sample_counter.item(), + iterations=self.accum_iteration.item(), + training_hours=self.accum_train_in_hours.item(), + ) + + @abstractmethod + def forward(self, *args, **kwargs) -> Any: + pass diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/selective_activation_checkpoint.py b/REGEN-main/cosmos_policy/_src/predict2/networks/selective_activation_checkpoint.py new file mode 100644 index 0000000000000000000000000000000000000000..1bcc7b5466cf5c2b70feff12f64ea6b141d0b25e --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/selective_activation_checkpoint.py @@ -0,0 +1,73 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from dataclasses import dataclass +from enum import Enum + +import torch + +try: + from torch.utils.checkpoint import CheckpointPolicy, create_selective_checkpoint_contexts, noop_context_fn +except ImportError: + CheckpointPolicy = None + +mm_only_save_list = { + torch.ops.aten.mm.default, + torch.ops.aten._scaled_dot_product_efficient_attention.default, + torch.ops.aten._scaled_dot_product_flash_attention.default, + torch.ops.aten.addmm.default, +} + + +class CheckpointMode(str, Enum): + """ + Enum for the different checkpoint modes. + """ + + NONE = "none" + MM_ONLY = "mm_only" + BLOCK_WISE = "block_wise" + + def __str__(self) -> str: + # Optional: makes print() show just the value + return self.value + + +def mm_only_policy(ctx, func, *args, **kwargs): + """ + In newer flash-attn and TE versions, FA2 shows up in the list of ops with the name of 'flash_attn._flash_attn_forward'. + However, FA2 is much slower (2-3x) than FA3 or cuDNN kernel. Registering cuDNN kernel would require heavy changes in TE code. + That's why the best option is to use FA3 with small modifications to flash_attn_interface.py to register FA3 as PyTorch op. + """ + to_save = func in mm_only_save_list or "flash_attn" in str(func) + return CheckpointPolicy.MUST_SAVE if to_save else CheckpointPolicy.PREFER_RECOMPUTE + + +def mm_only_context_fn(): + return create_selective_checkpoint_contexts(mm_only_policy) + + +@dataclass +class SACConfig: + mode: str = "mm_only" + every_n_blocks: int = 1 + + def get_context_fn(self): + if self.mode == CheckpointMode.MM_ONLY: + return mm_only_context_fn + elif self.mode == CheckpointMode.BLOCK_WISE: + return noop_context_fn + else: + raise ValueError(f"Invalid mode: {self.mode}") diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/wan2pt1.py b/REGEN-main/cosmos_policy/_src/predict2/networks/wan2pt1.py new file mode 100644 index 0000000000000000000000000000000000000000..290681302353ec7a1349bdcf89aa3961f75bbbe1 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/wan2pt1.py @@ -0,0 +1,1029 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# from Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved. + +import math +from typing import Optional + +import torch +import torch.amp as amp +import torch.nn as nn +from einops import rearrange, repeat + +from cosmos_policy._src.predict2.networks.a2a_cp import MinimalA2AAttnOp +from cosmos_policy._src.predict2.networks.attention import attention + +try: + from flash_attn.layers.rotary import apply_rotary_emb as flash_apply_rotary_emb +except ImportError: + flash_apply_rotary_emb = None + print("flash_attn is not installed.") + +from torch.distributed import ProcessGroup, get_process_group_ranks +from torch.distributed._composable.fsdp import fully_shard +from torch.distributed.algorithms._checkpoint.checkpoint_wrapper import checkpoint_wrapper as ptd_checkpoint_wrapper +from torch.nn.modules.module import _IncompatibleKeys +from torchvision import transforms +from transformer_engine.pytorch.attention import DotProductAttention + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.context_parallel import split_inputs_cp +from cosmos_policy._src.predict2.networks.model_weights_stats import WeightTrainingStat +from cosmos_policy._src.predict2.networks.selective_activation_checkpoint import ( + CheckpointMode, +) +from cosmos_policy._src.predict2.networks.selective_activation_checkpoint import SACConfig as SACConfig + +T5_CONTEXT_TOKEN_NUMBER = 512 +FIRST_LAST_FRAME_CONTEXT_TOKEN_NUMBER = 257 * 2 + +""" +TODO: (qsh 2025-05-07) + - [ ] add cp + - [ ] add init method + - [x] Clean up code, remove auto cast, memory saving improvements + - [ ] It is not per head qk norm. +""" +from collections import namedtuple + +VideoSize = namedtuple("VideoSize", ["T", "H", "W"]) + + +class VideoPositionEmb(nn.Module): + def __init__(self): + super().__init__() + self._cp_group = None + + def enable_context_parallel(self, process_group: ProcessGroup): + self._cp_group = process_group + + def disable_context_parallel(self): + self._cp_group = None + + @property + def seq_dim(self): + return 1 + + def forward(self, x_B_T_H_W_C: torch.Tensor) -> torch.Tensor: + """ + With CP, the function assume that the input tensor is already split. + It delegates the embedding generation to generate_embeddings function. + """ + B_T_H_W_C = x_B_T_H_W_C.shape + if self._cp_group is not None: + cp_ranks = get_process_group_ranks(self._cp_group) + cp_size = len(cp_ranks) + B, T, H, W, C = B_T_H_W_C + B_T_H_W_C = (B, T * cp_size, H, W, C) + embeddings = self.generate_embeddings(B_T_H_W_C) + + return self._split_for_context_parallel(embeddings) + + def generate_embeddings(self, B_T_H_W_C: torch.Size): + raise NotImplementedError + + def _split_for_context_parallel(self, embeddings): + if self._cp_group is not None: + embeddings = split_inputs_cp(x=embeddings, seq_dim=self.seq_dim, cp_group=self._cp_group) + return embeddings + + +class VideoRopePosition3DEmb(VideoPositionEmb): + def __init__( + self, + head_dim: int, + len_h: int, + len_w: int, + len_t: int, + h_extrapolation_ratio: float = 1.0, + w_extrapolation_ratio: float = 1.0, + t_extrapolation_ratio: float = 1.0, + ): + super().__init__() + self.max_h = len_h + self.max_w = len_w + self.max_t = len_t + dim = head_dim + dim_h = dim // 6 * 2 + dim_w = dim_h + dim_t = dim - 2 * dim_h + assert dim == dim_h + dim_w + dim_t, f"bad dim: {dim} != {dim_h} + {dim_w} + {dim_t}" + self._dim_h = dim_h + self._dim_t = dim_t + + self.h_ntk_factor = h_extrapolation_ratio ** (dim_h / (dim_h - 2)) + self.w_ntk_factor = w_extrapolation_ratio ** (dim_w / (dim_w - 2)) + self.t_ntk_factor = t_extrapolation_ratio ** (dim_t / (dim_t - 2)) + + self._is_initialized = False + + def cache_parameters(self) -> None: + if self._is_initialized: + return + + dim_h = self._dim_h + dim_t = self._dim_t + + self.seq = torch.arange(max(self.max_h, self.max_w, self.max_t)).float().cuda() + self.dim_spatial_range = torch.arange(0, dim_h, 2)[: (dim_h // 2)].float().cuda() / dim_h + self.dim_temporal_range = torch.arange(0, dim_t, 2)[: (dim_t // 2)].float().cuda() / dim_t + self._is_initialized = True + + def generate_embeddings( + self, + B_T_H_W_C: torch.Size, + h_ntk_factor: Optional[float] = None, + w_ntk_factor: Optional[float] = None, + t_ntk_factor: Optional[float] = None, + ): + """ + Generate embeddings for the given input size. + + Args: + B_T_H_W_C (torch.Size): Input tensor size (Batch, Time, Height, Width, Channels). + fps (Optional[torch.Tensor], optional): Frames per second. Defaults to None. + h_ntk_factor (Optional[float], optional): Height NTK factor. If None, uses self.h_ntk_factor. + w_ntk_factor (Optional[float], optional): Width NTK factor. If None, uses self.w_ntk_factor. + t_ntk_factor (Optional[float], optional): Time NTK factor. If None, uses self.t_ntk_factor. + + Returns: + Not specified in the original code snippet. + """ + self.cache_parameters() + + h_ntk_factor = h_ntk_factor if h_ntk_factor is not None else self.h_ntk_factor + w_ntk_factor = w_ntk_factor if w_ntk_factor is not None else self.w_ntk_factor + t_ntk_factor = t_ntk_factor if t_ntk_factor is not None else self.t_ntk_factor + + h_theta = 10000.0 * h_ntk_factor + w_theta = 10000.0 * w_ntk_factor + t_theta = 10000.0 * t_ntk_factor + + h_spatial_freqs = 1.0 / (h_theta**self.dim_spatial_range) + w_spatial_freqs = 1.0 / (w_theta**self.dim_spatial_range) + temporal_freqs = 1.0 / (t_theta**self.dim_temporal_range) + + B, T, H, W, _ = B_T_H_W_C + assert H <= self.max_h and W <= self.max_w, ( + f"Input dimensions (H={H}, W={W}) exceed the maximum dimensions (max_h={self.max_h}, max_w={self.max_w})" + ) + freqs_h = torch.outer(self.seq[:H], h_spatial_freqs) + freqs_w = torch.outer(self.seq[:W], w_spatial_freqs) + freqs_t = torch.outer(self.seq[:T], temporal_freqs) + freqs_T_H_W_D = torch.cat( + [ + repeat(freqs_t, "t d -> t h w d", h=H, w=W), + repeat(freqs_h, "h d -> t h w d", t=T, w=W), + repeat(freqs_w, "w d -> t h w d", t=T, h=H), + ], + dim=-1, + ) + + return rearrange(freqs_T_H_W_D, "t h w d -> (t h w) 1 1 d").float() + + @property + def seq_dim(self): + return 0 + + +def sinusoidal_embedding_1d(dim, position): + # preprocess + assert dim % 2 == 0 + half = dim // 2 + position = position.type(torch.float64) + + # calculation + sinusoid = torch.outer(position, torch.pow(10000, -torch.arange(half).to(position).div(half))) + x = torch.cat([torch.cos(sinusoid), torch.sin(sinusoid)], dim=1) + return x + + +def rope_apply(x, video_size: VideoSize, freqs): + """ + Optimized version of rope_apply using flash_attention's rotary embedding implementation. + This version processes the entire batch at once for efficiency. + + Args: + x (Tensor): Input tensor with shape [batch_size, seq_len, n_heads, head_dim] + video_size (VideoSize): Video dimensions with shape [T, H, W] + freqs (Tensor): Complex frequencies with shape [max_seq_len, head_dim // 2] + + Returns: + Tensor: Rotary-embedded tensor with same shape as input + """ + batch_size, seq_len, n_heads, head_dim = x.shape + + # Since all items in the batch share the same grid dimensions, we can use the first item + T, H, W = video_size + curr_seq_len = T * H * W + + # Make sure the sequence length matches the grid size + assert seq_len == curr_seq_len, "Sequence length must be equal to T*H*W" + + freqs = freqs.view(seq_len, head_dim // 2) + cos = torch.cos(freqs).to(torch.float32) + sin = torch.sin(freqs).to(torch.float32) + + # Apply the rotation + rotated = flash_apply_rotary_emb(x.to(torch.float32), cos, sin, interleaved=True, inplace=False) + + return rotated.to(x.dtype) + + +class WanRMSNorm(nn.Module): + def __init__(self, dim, eps=1e-5): + super().__init__() + self.dim = dim + self.eps = eps + self.weight = nn.Parameter(torch.ones(dim)) + + def reset_parameters(self): + self.weight.data.fill_(1.0) + + def forward(self, x): + r""" + Args: + x(Tensor): Shape [B, L, C] + """ + return self._norm(x.float()).type_as(x) * self.weight + + def _norm(self, x): + return x * torch.rsqrt(x.pow(2).mean(dim=-1, keepdim=True) + self.eps) + + +class WanLayerNorm(nn.LayerNorm): + def __init__(self, dim, eps=1e-6, elementwise_affine=False): + super().__init__(dim, elementwise_affine=elementwise_affine, eps=eps) + + def forward(self, x): + r""" + Args: + x(Tensor): Shape [B, L, C] + """ + # return super().forward(x.float()).type_as(x) + return super().forward(x) + + +class SelfAttnOp(DotProductAttention): + def forward( + self, + q_B_L_H_D, + k_B_L_H_D, + v_B_L_H_D, + video_size: Optional[VideoSize] = None, + ): + return super().forward(q_B_L_H_D, k_B_L_H_D, v_B_L_H_D) + + +class WanSelfAttention(nn.Module): + def __init__( + self, + dim, + num_heads, + window_size=(-1, -1), + qk_norm=True, + eps=1e-6, + cp_comm_type="p2p", + attention_backend="transformer_engine", + ): + assert dim % num_heads == 0 + super().__init__() + self.dim = dim + self.num_heads = num_heads + self.head_dim = dim // num_heads + self.window_size = window_size + self.qk_norm = qk_norm + self.eps = eps + self.qk_norm = qk_norm + self.cp_comm_type = cp_comm_type + self.attention_backend = attention_backend + + # layers + self.q = nn.Linear(dim, dim) + self.k = nn.Linear(dim, dim) + self.v = nn.Linear(dim, dim) + self.o = nn.Linear(dim, dim) + self.norm_q = WanRMSNorm(dim, eps=eps) if qk_norm else nn.Identity() + self.norm_k = WanRMSNorm(dim, eps=eps) if qk_norm else nn.Identity() + + if self.attention_backend == "transformer_engine": + self.attn_op = SelfAttnOp( + self.num_heads, + self.head_dim, + num_gqa_groups=self.num_heads, + attention_dropout=0, + qkv_format="bshd", + attn_mask_type="no_mask", + ) + elif self.attention_backend == "minimal_a2a": + self.attn_op = MinimalA2AAttnOp() + else: + assert False, f"Unreckognized attention backend: {self.attention_backend}" + + def init_weights(self): + std = 1.0 / math.sqrt(self.dim) + torch.nn.init.trunc_normal_(self.q.weight, std=std) + torch.nn.init.trunc_normal_(self.k.weight, std=std) + torch.nn.init.trunc_normal_(self.v.weight, std=std) + torch.nn.init.trunc_normal_(self.o.weight, std=std) + # zero out bias + self.q.bias.data.zero_() + self.k.bias.data.zero_() + self.v.bias.data.zero_() + self.o.bias.data.zero_() + # reset norm weights + if self.qk_norm: + self.norm_q.reset_parameters() + self.norm_k.reset_parameters() + + def forward(self, x, seq_lens, video_size: VideoSize, freqs): + r""" + Args: + x(Tensor): Shape [B, L, num_heads, C / num_heads] + seq_lens(Tensor): Shape [B] + video_size(VideoSize): Shape [T, H, W] + freqs(Tensor): Rope freqs, shape [1024, C / num_heads / 2] + """ + b, s, n, d = *x.shape[:2], self.num_heads, self.head_dim + + # query, key, value function + def qkv_fn(x): + q = self.norm_q(self.q(x)).view(b, s, n, d) + k = self.norm_k(self.k(x)).view(b, s, n, d) + v = self.v(x).view(b, s, n, d) + return q, k, v + + q, k, v = qkv_fn(x) + + x = self.attn_op(rope_apply(q, video_size, freqs), rope_apply(k, video_size, freqs), v, video_size) + + # output + x = x.flatten(2) + x = self.o(x) + return x + + def set_context_parallel_group(self, process_group, ranks, stream): + if self.attention_backend == "transformer_engine": + self.attn_op.set_context_parallel_group(process_group, ranks, stream, cp_comm_type=self.cp_comm_type) + elif self.attention_backend == "minimal_a2a": + self.attn_op.set_context_parallel_group(process_group, ranks, stream) + else: + assert False, f"Unreckognized attention backend: {self.attention_backend}" + + +class WanT2VCrossAttention(WanSelfAttention): + def forward(self, x, context, context_lens): + r""" + Args: + x(Tensor): Shape [B, L1, C] + context(Tensor): Shape [B, L2, C] + context_lens(Tensor): Shape [B] + """ + b, n, d = x.size(0), self.num_heads, self.head_dim + + # compute query, key, value + q = self.norm_q(self.q(x)).view(b, -1, n, d) + k = self.norm_k(self.k(context)).view(b, -1, n, d) + v = self.v(context).view(b, -1, n, d) + + # compute attention + x = self.attn_op(q, k, v, None) + # output + x = x.flatten(2) + x = self.o(x) + return x + + +class WanI2VCrossAttention(WanSelfAttention): + def __init__( + self, + dim, + num_heads, + window_size=(-1, -1), + qk_norm=True, + eps=1e-6, + cp_comm_type="p2p", + attention_backend="transformer_engine", + ): + super().__init__(dim, num_heads, window_size, qk_norm, eps, cp_comm_type, attention_backend) + + self.k_img = nn.Linear(dim, dim) + self.v_img = nn.Linear(dim, dim) + # self.alpha = nn.Parameter(torch.zeros((1, ))) + self.norm_k_img = WanRMSNorm(dim, eps=eps) if qk_norm else nn.Identity() + + if self.attention_backend == "transformer_engine": + self.attn_op_image = DotProductAttention( + self.num_heads, + self.head_dim, + num_gqa_groups=self.num_heads, + attention_dropout=0, + qkv_format="bshd", + attn_mask_type="no_mask", + ) + elif self.attention_backend == "minimal_a2a": + self.attn_op_image = attention + else: + assert False, f"Unreckognized attention backend: {self.attention_backend}" + + def init_weights(self): + super().init_weights() + std = 1.0 / math.sqrt(self.dim) + torch.nn.init.trunc_normal_(self.k_img.weight, std=std) + torch.nn.init.trunc_normal_(self.v_img.weight, std=std) + # zero out bias + self.k_img.bias.data.zero_() + self.v_img.bias.data.zero_() + # reset norm weights + if self.qk_norm: + self.norm_k_img.reset_parameters() + + def forward(self, x, context, context_lens): + r""" + Args: + x(Tensor): Shape [B, L1, C] + context(Tensor): Shape [B, L2, C] + context_lens(Tensor): Shape [B] + """ + image_context_length = context.shape[1] - T5_CONTEXT_TOKEN_NUMBER + context_img = context[:, :image_context_length] + context = context[:, image_context_length:] + b, n, d = x.size(0), self.num_heads, self.head_dim + + # compute query, key, value + q = self.norm_q(self.q(x)).view(b, -1, n, d) + k = self.norm_k(self.k(context)).view(b, -1, n, d) + v = self.v(context).view(b, -1, n, d) + k_img = self.norm_k_img(self.k_img(context_img)).view(b, -1, n, d) + v_img = self.v_img(context_img).view(b, -1, n, d) + img_x = self.attn_op_image(q, k_img, v_img) + # compute attention + x = self.attn_op(q, k, v) + + # output + x = x.flatten(2) + img_x = img_x.flatten(2) + x = x + img_x + x = self.o(x) + return x + + +WAN_CROSSATTENTION_CLASSES = { + "t2v_cross_attn": WanT2VCrossAttention, + "i2v_cross_attn": WanI2VCrossAttention, +} + + +class WanAttentionBlock(nn.Module): + def __init__( + self, + cross_attn_type, + dim, + ffn_dim, + num_heads, + window_size=(-1, -1), + qk_norm=True, + cross_attn_norm=False, + eps=1e-6, + cp_comm_type="p2p", + attention_backend: str = "transformer_engine", + ): + super().__init__() + self.dim = dim + self.ffn_dim = ffn_dim + self.num_heads = num_heads + self.window_size = window_size + self.qk_norm = qk_norm + self.cross_attn_norm = cross_attn_norm + self.eps = eps + + # layers + self.norm1 = WanLayerNorm(dim, eps) + self.self_attn = WanSelfAttention(dim, num_heads, window_size, qk_norm, eps, cp_comm_type, attention_backend) + self.norm3 = WanLayerNorm(dim, eps, elementwise_affine=True) if cross_attn_norm else nn.Identity() + self.cross_attn = WAN_CROSSATTENTION_CLASSES[cross_attn_type]( + dim, num_heads, (-1, -1), qk_norm, eps, cp_comm_type, attention_backend + ) + self.norm2 = WanLayerNorm(dim, eps) + self.ffn = nn.Sequential(nn.Linear(dim, ffn_dim), nn.GELU(approximate="tanh"), nn.Linear(ffn_dim, dim)) + + # modulation + self.modulation = nn.Parameter(torch.randn(1, 6, dim) / dim**0.5) + + def init_weights(self): + self.self_attn.init_weights() + self.cross_attn.init_weights() + + self.norm1.reset_parameters() + self.norm2.reset_parameters() + self.norm3.reset_parameters() + + std = 1.0 / math.sqrt(self.dim) + torch.nn.init.trunc_normal_(self.modulation, std=std) + + def forward( + self, + x, + e, + seq_lens, + video_size: VideoSize, + freqs, + context, + context_lens, + ): + r""" + Args: + x(Tensor): Shape [B, L, C] + e(Tensor): Shape [B, 6, C] + seq_lens(Tensor): Shape [B], length of each sequence in batch + video_size(VideoSize): Shape [T, H, W] + freqs(Tensor): Rope freqs, shape [1024, C / num_heads / 2] + """ + assert e.dtype == torch.float32 + with amp.autocast("cuda", dtype=torch.float32): + e = (self.modulation + e).chunk(6, dim=1) + assert e[0].dtype == torch.float32 + + # self-attention + y = self.self_attn((self.norm1(x).float() * (1 + e[1]) + e[0]).type_as(x), seq_lens, video_size, freqs) + with amp.autocast("cuda", dtype=torch.float32): + x = x + y * e[2].type_as(x) + + # cross-attention & ffn function + def cross_attn_ffn(x, context, context_lens, e): + x = x + self.cross_attn(self.norm3(x), context, context_lens) + y = self.ffn((self.norm2(x).float() * (1 + e[4]) + e[3]).type_as(x)) + with amp.autocast("cuda", dtype=torch.float32): + x = x + y * e[5].type_as(x) + return x + + x = cross_attn_ffn(x, context, context_lens, e) + return x + + +class Head(nn.Module): + def __init__(self, dim, out_dim, patch_size, eps=1e-6): + super().__init__() + self.dim = dim + self.out_dim = out_dim + self.patch_size = patch_size + self.eps = eps + + # layers + out_dim = math.prod(patch_size) * out_dim + self.norm = WanLayerNorm(dim, eps) + self.head = nn.Linear(dim, out_dim) + + # modulation + self.modulation = nn.Parameter(torch.randn(1, 2, dim) / dim**0.5) + + def init_weights(self): + self.norm.reset_parameters() + + std = 1.0 / math.sqrt(self.dim) + torch.nn.init.trunc_normal_(self.modulation, std=std) + torch.nn.init.trunc_normal_(self.head.weight, std=std) + self.head.bias.data.zero_() + + def forward(self, x, e): + r""" + Args: + x(Tensor): Shape [B, L1, C] + e(Tensor): Shape [B, C] + """ + assert e.dtype == torch.float32 + with amp.autocast("cuda", dtype=torch.float32): + e = (self.modulation + e.unsqueeze(1)).chunk(2, dim=1) + x = self.head(self.norm(x) * (1 + e[1]) + e[0]) + return x + + +class MLPProj(torch.nn.Module): + def __init__(self, in_dim, out_dim, flf_pos_emb=False): + super().__init__() + + self.proj = torch.nn.Sequential( + torch.nn.LayerNorm(in_dim), + torch.nn.Linear(in_dim, in_dim), + torch.nn.GELU(), + torch.nn.Linear(in_dim, out_dim), + torch.nn.LayerNorm(out_dim), + ) + if flf_pos_emb: # NOTE: we only use this for `flf2v` + self.emb_pos = nn.Parameter(torch.zeros(1, FIRST_LAST_FRAME_CONTEXT_TOKEN_NUMBER, 1280)) + + def init_weights(self): + self.proj[0].reset_parameters() + self.proj[1].reset_parameters() + self.proj[3].reset_parameters() + self.proj[4].reset_parameters() + + if hasattr(self, "emb_pos"): + self.emb_pos.data.zero_() + + def forward(self, image_embeds): + if hasattr(self, "emb_pos"): + bs, n, d = image_embeds.shape + image_embeds = image_embeds.view(-1, 2 * n, d) + image_embeds = image_embeds + self.emb_pos + clip_extra_context_tokens = self.proj(image_embeds) + return clip_extra_context_tokens + + +class WanModel(WeightTrainingStat): + r""" + Wan diffusion backbone supporting both text-to-video and image-to-video. + """ + + def __init__( + self, + model_type="t2v", + patch_size=(1, 2, 2), + text_len=512, + in_dim=16, + dim=2048, + ffn_dim=8192, + freq_dim=256, + text_dim=4096, + out_dim=16, + num_heads=16, + num_layers=32, + window_size=(-1, -1), + qk_norm=True, + cross_attn_norm=True, + eps=1e-6, + concat_padding_mask: bool = False, + sac_config: SACConfig = SACConfig(), + cp_comm_type: str = "p2p", + attention_backend: str = "transformer_engine", + ): + r""" + Initialize the diffusion model backbone. + + Args: + model_type (`str`, *optional*, defaults to 't2v'): + Model variant - 't2v' (text-to-video) or 'i2v' (image-to-video) or 'flf2v' (first-last-frame-to-video) + patch_size (`tuple`, *optional*, defaults to (1, 2, 2)): + 3D patch dimensions for video embedding (t_patch, h_patch, w_patch) + text_len (`int`, *optional*, defaults to 512): + Fixed length for text embeddings + in_dim (`int`, *optional*, defaults to 16): + Input video channels (C_in) + dim (`int`, *optional*, defaults to 2048): + Hidden dimension of the transformer + ffn_dim (`int`, *optional*, defaults to 8192): + Intermediate dimension in feed-forward network + freq_dim (`int`, *optional*, defaults to 256): + Dimension for sinusoidal time embeddings + text_dim (`int`, *optional*, defaults to 4096): + Input dimension for text embeddings + out_dim (`int`, *optional*, defaults to 16): + Output video channels (C_out) + num_heads (`int`, *optional*, defaults to 16): + Number of attention heads + num_layers (`int`, *optional*, defaults to 32): + Number of transformer blocks + window_size (`tuple`, *optional*, defaults to (-1, -1)): + Window size for local attention (-1 indicates global attention) + qk_norm (`bool`, *optional*, defaults to True): + Enable query/key normalization + cross_attn_norm (`bool`, *optional*, defaults to False): + Enable cross-attention normalization + eps (`float`, *optional*, defaults to 1e-6): + Epsilon value for normalization layers + concat_padding_mask (`bool`, *optional*, defaults to False): + Enable concat padding mask + cp_comm_type (str, *optional*, defaults to 'p2p'): + CP communication type passed to TE. + attention_backend (str, defaults to 'transformer_engine', options are: ['transformer_engine', 'minimal_a2a']) + Backend used for attention + """ + + super().__init__() + + assert model_type in ["t2v", "i2v", "flf2v"] + self.model_type = model_type + + self.patch_size = patch_size + self.text_len = text_len + self.in_dim = in_dim + self.dim = dim + self.ffn_dim = ffn_dim + self.freq_dim = freq_dim + self.text_dim = text_dim + self.out_dim = out_dim + self.num_heads = num_heads + self.num_layers = num_layers + self.window_size = window_size + self.qk_norm = qk_norm + self.cross_attn_norm = cross_attn_norm + self.eps = eps + self.concat_padding_mask = concat_padding_mask + self.cp_comm_type = cp_comm_type + self.attention_backed = attention_backend + + # embeddings + in_dim = in_dim + 1 if self.concat_padding_mask else in_dim + self.patch_embedding = nn.Linear(in_dim * patch_size[0] * patch_size[1] * patch_size[2], dim) + + self.text_embedding = nn.Sequential(nn.Linear(text_dim, dim), nn.GELU(approximate="tanh"), nn.Linear(dim, dim)) + + self.time_embedding = nn.Sequential(nn.Linear(freq_dim, dim), nn.SiLU(), nn.Linear(dim, dim)) + self.time_projection = nn.Sequential(nn.SiLU(), nn.Linear(dim, dim * 6)) + + # blocks + cross_attn_type = "t2v_cross_attn" if model_type == "t2v" else "i2v_cross_attn" + self.blocks = nn.ModuleList( + [ + WanAttentionBlock( + cross_attn_type, + dim, + ffn_dim, + num_heads, + window_size, + qk_norm, + cross_attn_norm, + eps, + self.cp_comm_type, + attention_backend, + ) + for _ in range(num_layers) + ] + ) + + # head + self.head = Head(dim, out_dim, patch_size, eps) + + # buffers (don't use register_buffer otherwise dtype will be changed in to()) + assert (dim % num_heads) == 0 and (dim // num_heads) % 2 == 0 + + d = dim // num_heads + + self.rope_position_embedding = VideoRopePosition3DEmb( + head_dim=d, + len_h=128, + len_w=128, + len_t=32, + ) + + if model_type == "i2v" or model_type == "flf2v": + self.img_emb = MLPProj(1280, dim, flf_pos_emb=model_type == "flf2v") + + # initialize weights + self.init_weights() + + self.enable_selective_checkpoint(sac_config, self.blocks) + + def forward( + self, + x_B_C_T_H_W, + timesteps_B_T, + crossattn_emb, + seq_len=None, + frame_cond_crossattn_emb_B_L_D=None, + y_B_C_T_H_W=None, + padding_mask: Optional[torch.Tensor] = None, + is_uncond=False, + slg_layers=None, + **kwargs, + ): + r""" + Forward pass through the diffusion model + + Args: + x_B_C_T_H_W (Tensor): + Input video tensor with shape [B, C_in, T, H, W] + t (Tensor): + Diffusion timesteps tensor of shape [B] + context (List[Tensor]): + List of text embeddings each with shape [L, C] + seq_len (`int`): + Maximum sequence length for positional encoding + frame_cond_crossattn_emb_B_L_D (Tensor, *optional*): + CLIP image features for image-to-video mode or first-last-frame-to-video mode + y_B_C_T_H_W (Tensor, *optional*): + Conditional video inputs for image-to-video mode, shape [B, C_in, T, H, W] + + Returns: + Tensor: + Denoised video tensor with shape [B, C_out, T, H / 8, W / 8] + """ + assert timesteps_B_T.shape[1] == 1 + t_B = timesteps_B_T[:, 0] + del kwargs + if self.model_type == "i2v" or self.model_type == "flf2v": + assert frame_cond_crossattn_emb_B_L_D is not None and y_B_C_T_H_W is not None + + if y_B_C_T_H_W is not None: + x_B_C_T_H_W = torch.cat([x_B_C_T_H_W, y_B_C_T_H_W], dim=1) + + if self.concat_padding_mask: + padding_mask = transforms.functional.resize( + padding_mask, list(x_B_C_T_H_W.shape[-2:]), interpolation=transforms.InterpolationMode.NEAREST + ) + x_B_C_T_H_W = torch.cat( + [x_B_C_T_H_W, padding_mask.unsqueeze(1).repeat(1, 1, x_B_C_T_H_W.shape[2], 1, 1)], dim=1 + ) + + # embeddings + x_B_T_H_W_D = rearrange( + x_B_C_T_H_W, + "b c (t kt) (h kh) (w kw) -> b t h w (c kt kh kw)", + kt=self.patch_size[0], + kh=self.patch_size[1], + kw=self.patch_size[2], + ) + x_B_T_H_W_D = self.patch_embedding(x_B_T_H_W_D) + + video_size = VideoSize(T=x_B_T_H_W_D.shape[1], H=x_B_T_H_W_D.shape[2], W=x_B_T_H_W_D.shape[3]) + x_B_L_D = rearrange(x_B_T_H_W_D, "b t h w d -> b (t h w) d") + seq_lens = torch.tensor([u.size(0) for u in x_B_L_D], dtype=torch.long) + seq_len = seq_lens.max().item() + assert seq_lens.max() == seq_len + + # time embeddings + with amp.autocast("cuda", dtype=torch.float32): + e_B_D = self.time_embedding(sinusoidal_embedding_1d(self.freq_dim, t_B).float()) + e0_B_6_D = self.time_projection(e_B_D).unflatten(1, (6, self.dim)) + assert e_B_D.dtype == torch.float32 and e0_B_6_D.dtype == torch.float32 + + # context + context_lens = None + context_B_L_D = self.text_embedding(crossattn_emb) + + if frame_cond_crossattn_emb_B_L_D is not None: + context_clip = self.img_emb(frame_cond_crossattn_emb_B_L_D) # bs x 257 (x2) x dim + context_B_L_D = torch.concat([context_clip, context_B_L_D], dim=1) + + # arguments + kwargs = dict( + e=e0_B_6_D, + seq_lens=seq_lens, + video_size=video_size, + freqs=self.rope_position_embedding(x_B_T_H_W_D), + context=context_B_L_D, + context_lens=context_lens, + ) + + for block_idx, block in enumerate(self.blocks): + if slg_layers is not None and block_idx in slg_layers and is_uncond: + continue + x_B_L_D = block(x_B_L_D, **kwargs) + + # head + x_B_L_D = self.head(x_B_L_D, e_B_D) + + # unpatchify + t, h, w = video_size + x_B_C_T_H_W = rearrange( + x_B_L_D, + "b (t h w) (nt nh nw d) -> b d (t nt) (h nh) (w nw)", + nt=self.patch_size[0], + nh=self.patch_size[1], + nw=self.patch_size[2], + t=t, + h=h, + w=w, + d=self.out_dim, + ) + + return x_B_C_T_H_W + + def init_weights(self): + r""" + Initialize model parameters using Xavier initialization. + """ + + # basic init + for m in self.modules(): + if isinstance(m, nn.Linear): + nn.init.xavier_uniform_(m.weight) + if m.bias is not None: + nn.init.zeros_(m.bias) + + for block in self.blocks: + block.init_weights() + self.head.init_weights() + + # init embeddings + nn.init.xavier_uniform_(self.patch_embedding.weight.flatten(1)) + nn.init.zeros_(self.patch_embedding.bias) + + for m in self.text_embedding.modules(): + if isinstance(m, nn.Linear): + nn.init.normal_(m.weight, std=0.02) + if m.bias is not None: + nn.init.zeros_(m.bias) + + for m in self.time_embedding.modules(): + if isinstance(m, nn.Linear): + nn.init.normal_(m.weight, std=0.02) + if m.bias is not None: + nn.init.zeros_(m.bias) + + for m in self.time_projection.modules(): + if isinstance(m, nn.Linear): + nn.init.normal_(m.weight, std=0.02) + if m.bias is not None: + nn.init.zeros_(m.bias) + + # init output layer + nn.init.zeros_(self.head.head.weight) + if self.head.head.bias is not None: + nn.init.zeros_(self.head.head.bias) + + def fully_shard(self, mesh): + for i, block in enumerate(self.blocks): + fully_shard(block, mesh=mesh, reshard_after_forward=True) + fully_shard(self.head, mesh=mesh, reshard_after_forward=False) + fully_shard(self.text_embedding, mesh=mesh, reshard_after_forward=True) + fully_shard(self.time_embedding, mesh=mesh, reshard_after_forward=True) + fully_shard(self.patch_embedding, mesh=mesh, reshard_after_forward=True) + + def disable_context_parallel(self): + # pos_embedder + self.rope_position_embedding.disable_context_parallel() + # attention + for block in self.blocks: + block.self_attn.set_context_parallel_group( + process_group=None, + ranks=None, + stream=torch.cuda.Stream(), + ) + + self._is_context_parallel_enabled = False + + def enable_context_parallel(self, process_group: Optional[ProcessGroup] = None): + # pos_embedder + self.rope_position_embedding.enable_context_parallel(process_group=process_group) + cp_ranks = get_process_group_ranks(process_group) + for block in self.blocks: + block.self_attn.set_context_parallel_group( + process_group=process_group, + ranks=cp_ranks, + stream=torch.cuda.Stream(), + ) + + self._is_context_parallel_enabled = True + + @property + def is_context_parallel_enabled(self): + return self._is_context_parallel_enabled + + def enable_selective_checkpoint(self, sac_config: SACConfig, blocks: nn.ModuleList): + if sac_config.mode == CheckpointMode.NONE: + pass + + log.info( + f"Enable selective checkpoint with {sac_config.mode}, for every {sac_config.every_n_blocks} blocks. Total blocks: {len(blocks)}" + ) + _context_fn = sac_config.get_context_fn() + for block_id, block in blocks.named_children(): + if int(block_id) % sac_config.every_n_blocks == 0: + log.info(f"Enable selective checkpoint for block {block_id}") + block = ptd_checkpoint_wrapper( + block, + context_fn=_context_fn, + preserve_rng_state=False, + ) + blocks.register_module(block_id, block) + self.register_module( + "head", + ptd_checkpoint_wrapper( + self.head, + context_fn=_context_fn, + preserve_rng_state=False, + ), + ) + + def load_state_dict(self, state_dict, strict=True, assign=False) -> _IncompatibleKeys: + filtered_state_dict = {} + for k, v in state_dict.items(): + if "_extra_state" in k: # Key introduced by TransformerEngine for FP8 + log.warning(f"Skipping key {k} introduced by TransformerEngine for FP8 in the checkpoint.") + continue + filtered_state_dict[k] = v + + state_dict = filtered_state_dict + + missing_keys, unexpected_keys = super().load_state_dict(state_dict, strict=False, assign=assign) + + if strict is True: + # We don't use FP8 so we can ignore those keys + assert all("_extra_state" in k for k in missing_keys) + assert all("_extra_state" in k for k in unexpected_keys) + + return _IncompatibleKeys(missing_keys, unexpected_keys) diff --git a/REGEN-main/cosmos_policy/_src/predict2/networks/xlm_roberta.py b/REGEN-main/cosmos_policy/_src/predict2/networks/xlm_roberta.py new file mode 100644 index 0000000000000000000000000000000000000000..931683d3de62e0e5bc4034de518cb6e12bd425ad --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/networks/xlm_roberta.py @@ -0,0 +1,180 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Modified from transformers.models.xlm_roberta.modeling_xlm_roberta +# Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved. +import torch +import torch.nn as nn +import torch.nn.functional as F + +__all__ = ["XLMRoberta", "xlm_roberta_large"] + + +class SelfAttention(nn.Module): + def __init__(self, dim, num_heads, dropout=0.1, eps=1e-5): + assert dim % num_heads == 0 + super().__init__() + self.dim = dim + self.num_heads = num_heads + self.head_dim = dim // num_heads + self.eps = eps + + # layers + self.q = nn.Linear(dim, dim) + self.k = nn.Linear(dim, dim) + self.v = nn.Linear(dim, dim) + self.o = nn.Linear(dim, dim) + self.dropout = nn.Dropout(dropout) + + def forward(self, x, mask): + """ + x: [B, L, C]. + """ + b, s, c, n, d = *x.size(), self.num_heads, self.head_dim + + # compute query, key, value + q = self.q(x).reshape(b, s, n, d).permute(0, 2, 1, 3) + k = self.k(x).reshape(b, s, n, d).permute(0, 2, 1, 3) + v = self.v(x).reshape(b, s, n, d).permute(0, 2, 1, 3) + + # compute attention + p = self.dropout.p if self.training else 0.0 + x = F.scaled_dot_product_attention(q, k, v, mask, p) + x = x.permute(0, 2, 1, 3).reshape(b, s, c) + + # output + x = self.o(x) + x = self.dropout(x) + return x + + +class AttentionBlock(nn.Module): + def __init__(self, dim, num_heads, post_norm, dropout=0.1, eps=1e-5): + super().__init__() + self.dim = dim + self.num_heads = num_heads + self.post_norm = post_norm + self.eps = eps + + # layers + self.attn = SelfAttention(dim, num_heads, dropout, eps) + self.norm1 = nn.LayerNorm(dim, eps=eps) + self.ffn = nn.Sequential(nn.Linear(dim, dim * 4), nn.GELU(), nn.Linear(dim * 4, dim), nn.Dropout(dropout)) + self.norm2 = nn.LayerNorm(dim, eps=eps) + + def forward(self, x, mask): + if self.post_norm: + x = self.norm1(x + self.attn(x, mask)) + x = self.norm2(x + self.ffn(x)) + else: + x = x + self.attn(self.norm1(x), mask) + x = x + self.ffn(self.norm2(x)) + return x + + +class XLMRoberta(nn.Module): + """ + XLMRobertaModel with no pooler and no LM head. + """ + + def __init__( + self, + vocab_size=250002, + max_seq_len=514, + type_size=1, + pad_id=1, + dim=1024, + num_heads=16, + num_layers=24, + post_norm=True, + dropout=0.1, + eps=1e-5, + ): + super().__init__() + self.vocab_size = vocab_size + self.max_seq_len = max_seq_len + self.type_size = type_size + self.pad_id = pad_id + self.dim = dim + self.num_heads = num_heads + self.num_layers = num_layers + self.post_norm = post_norm + self.eps = eps + + # embeddings + self.token_embedding = nn.Embedding(vocab_size, dim, padding_idx=pad_id) + self.type_embedding = nn.Embedding(type_size, dim) + self.pos_embedding = nn.Embedding(max_seq_len, dim, padding_idx=pad_id) + self.dropout = nn.Dropout(dropout) + + # blocks + self.blocks = nn.ModuleList( + [AttentionBlock(dim, num_heads, post_norm, dropout, eps) for _ in range(num_layers)] + ) + + # norm layer + self.norm = nn.LayerNorm(dim, eps=eps) + + def forward(self, ids): + """ + ids: [B, L] of torch.LongTensor. + """ + b, s = ids.shape + mask = ids.ne(self.pad_id).long() + + # embeddings + x = ( + self.token_embedding(ids) + + self.type_embedding(torch.zeros_like(ids)) + + self.pos_embedding(self.pad_id + torch.cumsum(mask, dim=1) * mask) + ) + if self.post_norm: + x = self.norm(x) + x = self.dropout(x) + + # blocks + mask = torch.where(mask.view(b, 1, 1, s).gt(0), 0.0, torch.finfo(x.dtype).min) + for block in self.blocks: + x = block(x, mask) + + # output + if not self.post_norm: + x = self.norm(x) + return x + + +def xlm_roberta_large(pretrained=False, return_tokenizer=False, device="cpu", **kwargs): + """ + XLMRobertaLarge adapted from Huggingface. + """ + # params + cfg = dict( + vocab_size=250002, + max_seq_len=514, + type_size=1, + pad_id=1, + dim=1024, + num_heads=16, + num_layers=24, + post_norm=True, + dropout=0.1, + eps=1e-5, + ) + cfg.update(**kwargs) + + # init a model on device + with torch.device(device): + model = XLMRoberta(**cfg) + return model diff --git a/REGEN-main/cosmos_policy/_src/predict2/schedulers/rectified_flow.py b/REGEN-main/cosmos_policy/_src/predict2/schedulers/rectified_flow.py new file mode 100644 index 0000000000000000000000000000000000000000..a92e47061817e03b198b8dfd543ac3762bf3299c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/schedulers/rectified_flow.py @@ -0,0 +1,173 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Callable + +import torch +from diffusers import FlowMatchEulerDiscreteScheduler + + +class TrainTimeWeight: + def __init__( + self, + noise_scheduler, + weight: str = "uniform", + ): + # Map reweighting -> uniform to support inference for existing checkpoints. + if weight == "reweighting": + weight = "uniform" + + self.weight = weight + self.noise_scheduler = noise_scheduler + + assert self.weight == "uniform", "Only uniform loss weight is supported in RF" + + def __call__(self, t, tensor_kwargs) -> torch.Tensor: + if self.weight == "uniform": + wts = torch.ones_like(t) + else: + raise NotImplementedError(f"Time weight '{self.weight}' is not implemented.") + + return wts + + +class TrainTimeSampler: + def __init__( + self, + distribution: str = "uniform", + ): + self.distribution = distribution + + @torch.no_grad() + def __call__( + self, + batch_size: int, + device: torch.device = torch.device("cpu"), + dtype: torch.dtype = torch.float32, + ) -> torch.Tensor: + """ + Sample time tensor for training + + Returns: + torch.Tensor: Time tensor, shape (batch_size,) + """ + if self.distribution == "uniform": + t = torch.rand((batch_size,)).to(device=device, dtype=dtype) + elif self.distribution == "logitnormal": + t = torch.sigmoid(torch.randn((batch_size,))).to(device=device, dtype=dtype) + else: + raise NotImplementedError(f"Time distribution '{self.dist}' is not implemented.") + + return t + + +class RectifiedFlow: + def __init__( + self, + velocity_field: Callable, + train_time_distribution: TrainTimeSampler | str = "uniform", + train_time_weight_method: str = "uniform", + use_dynamic_shift: bool = False, + shift: int = 3, + device: torch.device = torch.device("cpu"), + dtype: torch.dtype = torch.float32, + ): + r"""Initialize the RectifiedFlow class. + + Args: + velocity_field (`Callable`): + A function that predicts the velocity given the current state and time. + train_time_distribution (`TrainTimeSampler` or `str`, *optional*, defaults to `"uniform"`): + Distribution for sampling training times. + Can be an instance of `TrainTimeSampler` or a string specifying the distribution type. + train_time_weight (`TrainTimeWeight` or `str`, *optional*, defaults to `"uniform"`): + Weight applied to training times. + Can be an instance of `TrainTimeWeight` or a string specifying the weight type. + """ + self.velocity_field = velocity_field + self.train_time_sampler: TrainTimeSampler = ( + train_time_distribution + if isinstance(train_time_distribution, TrainTimeSampler) + else TrainTimeSampler(train_time_distribution) + ) + + if use_dynamic_shift: + self.noise_scheduler = FlowMatchEulerDiscreteScheduler(use_dynamic_shifting=use_dynamic_shift) + else: + self.noise_scheduler = FlowMatchEulerDiscreteScheduler(shift=shift) + self.train_time_weight = TrainTimeWeight(self.noise_scheduler, train_time_weight_method) + + self.device = torch.device(device) if isinstance(device, str) else device + self.dtype = torch.dtype(dtype) if isinstance(dtype, str) else dtype + + def sample_train_time(self, batch_size: int): + r"""This method calls the `TrainTimeSampler` to sample training times. + + Returns: + t (`torch.Tensor`): + A tensor of sampled training times with shape `(batch_size,)`, + matching the class specified `device` and `dtype`. + """ + time = self.train_time_sampler(batch_size, device=self.device, dtype=self.dtype) + return time + + def get_discrete_timestamp(self, u, tensor_kwargs): + r"""This method map time from 0,1 to discrete steps""" + + indices = (u.squeeze() * self.noise_scheduler.config.num_train_timesteps).long() + timesteps = self.noise_scheduler.timesteps.to(**tensor_kwargs)[indices] + return timesteps.unsqueeze(0) if timesteps.ndim == 0 else timesteps + + def get_sigmas(self, timesteps, tensor_kwargs): + sigmas = self.noise_scheduler.sigmas.to(**tensor_kwargs) + schedule_timesteps = self.noise_scheduler.timesteps.to(**tensor_kwargs) + step_indices = [(schedule_timesteps == t).nonzero().squeeze().tolist() for t in timesteps] + assert len(step_indices) == timesteps.shape[0], "Number of indices do not match the given timesteps." + sigma = sigmas[step_indices].flatten() + + return sigma + + def get_interpolation( + self, + x_0: torch.Tensor, + x_1: torch.Tensor, + t: torch.Tensor, + ): + r""" + This method computes interpolation `X_t` and their time derivatives `dotX_t` at the specified time points `t`. + Note that `x_0` is the noise, and `x_1` is the clean data. This is aligned with the notation in the recified flow community, + but different from the notation in the diffusion community. + + Args: + x_0 (`torch.Tensor`): + noise, shape `(B, D1, D2, ..., Dn)`, where `B` is the batch size, and `D1, D2, ..., Dn` are the data dimensions. + x_1 (`torch.Tensor`): + clean data, with the same shape as `x_0` + t (`torch.Tensor`): + A tensor of time steps, with shape `(B,)`, where each value is in `[0, 1]`. + + Returns: + (x_t, dot_x_t) (`Tuple[torch.Tensor, torch.Tensor]`): + - x_t (`torch.Tensor`): The interpolated state, with shape `(B, D1, D2, ..., Dn)`. + - dot_x_t (torch.Tensor): The time derivative of the interpolated state, with the same shape as `x_t`. + """ + assert x_0.shape == x_1.shape, "x_0 and x_1 must have the same shape." + assert x_0.shape[0] == x_1.shape[0], "Batch size of x_0 and x_1 must match." + assert t.shape[0] == x_1.shape[0], "Batch size of t must match x_1." + # Reshape t to match dimensions of x_1 + t = t.view(t.shape[0], *([1] * (len(x_1.shape) - 1))) + x_t = x_0 * t + x_1 * (1 - t) + dot_x_t = x_0 - x_1 + return x_t, dot_x_t diff --git a/REGEN-main/cosmos_policy/_src/predict2/tests/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/tests/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/tests/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/tests/training_loss_test.py b/REGEN-main/cosmos_policy/_src/predict2/tests/training_loss_test.py new file mode 100644 index 0000000000000000000000000000000000000000..9f48d997a493504eeb9fc164498384c2ef4137e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/tests/training_loss_test.py @@ -0,0 +1,85 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Test that verifies specific training output patterns are present in the logs. +This test runs a training command and checks for expected iteration and loss patterns. + +Usage: + * [run all tests]: pytest -s cosmos_policy/_src/predict2/tests/training_loss_test.py --L1 --all 2>&1 | tee /tmp/err.log +""" + +import re +import subprocess + +import pytest + +from cosmos_policy._src.imaginaire.utils.helper_test import RunIf + + +@RunIf(min_gpus=1) +@pytest.mark.L1 +def test_training_output_patterns(): + """Test that verifies specific training output patterns are present in the logs.""" + + # Define the command to run + cmd = "torchrun --nproc_per_node=1 --master_port=12341 -m scripts.train --config=cosmos_policy/_src/predict2/configs/text2world/config.py -- experiment=error-free_ddp_mock-data_base-cb trainer.max_iter=3 trainer.cudnn.deterministic=True trainer.cudnn.benchmark=False" + + # Define the expected patterns to check for + expected_patterns = [ + r"Iteration 1: Hit counter: 1/5 \| Loss: 16\.7822", + r"Iteration 2: Hit counter: 2/5 \| Loss: 13\.3350", + r"Iteration 3: Hit counter: 3/5 \| Loss: 17\.5436", + ] + + # Run the command and capture output + try: + result = subprocess.run( + cmd, + shell=True, + capture_output=True, + text=True, + check=False, + timeout=300, # 5 minute timeout + ) + + # Get the combined output (stdout + stderr) + output = result.stdout + result.stderr + + # Check if command was successful + if result.returncode != 0: + pytest.fail(f"Command failed with return code {result.returncode}. Error: {result.stderr}") + + # Check for each expected pattern + missing_patterns = [] + for i, pattern in enumerate(expected_patterns, 1): + if not re.search(pattern, output): + missing_patterns.append(f"Pattern {i}: {pattern}") + + # If any patterns are missing, fail the test + if missing_patterns: + pytest.fail( + "Missing expected patterns in output:\n" + + "\n".join(missing_patterns) + + f"\n\nCommand output:\n{output}" + ) + + # If we reach here, all patterns were found + print("✓ All expected patterns found in training output") + + except subprocess.TimeoutExpired: + pytest.fail(f"Command timed out after 5 minutes: {cmd}") + except Exception as e: + pytest.fail(f"Unexpected error running command: {e}") diff --git a/REGEN-main/cosmos_policy/_src/predict2/text_encoders/__init__.py b/REGEN-main/cosmos_policy/_src/predict2/text_encoders/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/text_encoders/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/_src/predict2/text_encoders/reason1.py b/REGEN-main/cosmos_policy/_src/predict2/text_encoders/reason1.py new file mode 100644 index 0000000000000000000000000000000000000000..bfd7e280c8cf6bfb973b845981761043aee65f69 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/text_encoders/reason1.py @@ -0,0 +1,259 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +This file is modified from cosmos_policy/_src/reason1/models/vlm_qwen.py for extracting reason embeddings. +Main change is to return the hidden states from the language model. +""" + +from typing import List, Optional + +import torch +from torch.distributed._tensor import DTensor + +from cosmos_policy._src.reason1.models.vlm_qwen import QwenModel +from cosmos_policy._src.reason1.networks.qwen2_5_vl import get_rope_index as get_rope_index_v2_5 +from cosmos_policy._src.reason1.networks.qwen2_vl import get_rope_index as get_rope_index_v2 + + +class QwenVLBaseModel(QwenModel): + """ + This is a base class for QwenVL models. + Here we override the forward method and the training_step method to + obtain more intermediate results from the language model. + """ + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + + """ + Copy from QwenModel.forward with MODIFICATIONS + MODIFICATIONS: add "lm_outputs" to the batch output. + """ + + def _forward( + self, + input_ids: torch.LongTensor = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[List[torch.FloatTensor]] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + labels: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + pixel_values: Optional[torch.Tensor] = None, + pixel_values_videos: Optional[torch.FloatTensor] = None, + image_grid_thw: Optional[torch.LongTensor] = None, + video_grid_thw: Optional[torch.LongTensor] = None, + rope_deltas: Optional[torch.LongTensor] = None, + cache_position: Optional[torch.LongTensor] = None, + second_per_grid_ts: Optional[torch.Tensor] = None, + ) -> torch.Tensor: + r""" + Args: + labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): + Labels for computing the masked language modeling loss. Indices should either be in `[0, ..., + config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored + (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`. + + Returns: + + Example: + + ```python + >>> from PIL import Image + >>> import requests + >>> from transformers import AutoProcessor, Qwen2_5_VLForConditionalGeneration + + >>> model = Qwen2_5_VLForConditionalGeneration.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct") + >>> processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct") + + >>> messages = [ + { + "role": "user", + "content": [ + {"type": "image"}, + {"type": "text", "text": "What is shown in this image?"}, + ], + }, + ] + >>> url = "https://www.ilankelman.org/stopsigns/australia.jpg" + >>> image = Image.open(requests.get(url, stream=True).raw) + + >>> text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True) + >>> inputs = processor(text=[text], images=[image], vision_infos=[vision_infos]) + + >>> # Generate + >>> generate_ids = model.generate(inputs.input_ids, max_length=30) + >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0] + "The image shows a street scene with a red stop sign in the foreground. In the background, there is a large red gate with Chinese characters ..." + ```""" + + output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions + output_hidden_states = ( + output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states + ) + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + if inputs_embeds is None: + inputs_embeds = self.model.embed_tokens(input_ids) + # This is a trick to handle TP for LLM but no TP for vision encoder, we need to convert DTensor to regular tensor later + is_inputs_embeds_dtensor = isinstance(inputs_embeds, DTensor) # This is True for TP>1, False for TP=1 + if is_inputs_embeds_dtensor: + target_device_mesh = inputs_embeds.device_mesh + target_placements = inputs_embeds.placements + inputs_embeds = inputs_embeds.full_tensor() + + if pixel_values is not None: + pixel_values = pixel_values.type(self.visual.dtype) + image_embeds = self.visual(pixel_values, grid_thw=image_grid_thw) + n_image_tokens = (input_ids == self.config.image_token_id).sum().item() + n_image_features = image_embeds.shape[0] + if n_image_tokens != n_image_features: + raise ValueError( + f"Image features and image tokens do not match: tokens: {n_image_tokens}, features {n_image_features}" + ) + + mask = input_ids == self.config.image_token_id + mask_unsqueezed = mask.unsqueeze(-1) + mask_expanded = mask_unsqueezed.expand_as(inputs_embeds) + image_mask = mask_expanded.to(inputs_embeds.device) + + image_embeds = image_embeds.to(inputs_embeds.device, inputs_embeds.dtype) + inputs_embeds = inputs_embeds.masked_scatter(image_mask, image_embeds) + + if pixel_values_videos is not None: + pixel_values_videos = pixel_values_videos.type(self.visual.dtype) + video_embeds = self.visual(pixel_values_videos, grid_thw=video_grid_thw) + n_video_tokens = (input_ids == self.config.video_token_id).sum().item() + n_video_features = video_embeds.shape[0] + if n_video_tokens != n_video_features: + raise ValueError( + f"Video features and video tokens do not match: tokens: {n_video_tokens}, features {n_video_features}" + ) + + mask = input_ids == self.config.video_token_id + mask_unsqueezed = mask.unsqueeze(-1) + mask_expanded = mask_unsqueezed.expand_as(inputs_embeds) + video_mask = mask_expanded.to(inputs_embeds.device) + + video_embeds = video_embeds.to(inputs_embeds.device, inputs_embeds.dtype) + inputs_embeds = inputs_embeds.masked_scatter(video_mask, video_embeds) + + if is_inputs_embeds_dtensor: + inputs_embeds = ( + DTensor.from_local(inputs_embeds, device_mesh=target_device_mesh) + .redistribute(placements=target_placements) + .to_local() + ) + if attention_mask is not None: + attention_mask = attention_mask.to(inputs_embeds.device) + + # if we get 4D attention mask we cannot calculate rope deltas anymore. + if position_ids is None and (attention_mask is None or attention_mask.ndim == 2): + # calculate RoPE index once per generation in the pre-fill stage only + if ( + (cache_position is not None and cache_position[0] == 0) + or self.rope_deltas is None + or (past_key_values is None or past_key_values.get_seq_length() == 0) + ): + if self.config.model_type == "qwen2_5_vl": + position_ids, rope_deltas = get_rope_index_v2_5( + self.config, + input_ids, + image_grid_thw, + video_grid_thw, + second_per_grid_ts, + attention_mask, + ) + elif self.config.model_type == "qwen2_vl": + position_ids, rope_deltas = get_rope_index_v2( + self.config, + input_ids, + image_grid_thw, + video_grid_thw, + attention_mask, + ) + elif self.config.model_type == "qwen2_5": + position_ids = None + rope_deltas = None + else: + raise ValueError(f"Unsupported model type: {self.config.model_type}") + self.rope_deltas = rope_deltas + # then use the prev pre-calculated rope-deltas to get the correct position ids + else: + batch_size, seq_length, _ = inputs_embeds.shape + delta = ( + (cache_position[0] + self.rope_deltas).to(inputs_embeds.device) if cache_position is not None else 0 + ) + position_ids = torch.arange(seq_length, device=inputs_embeds.device) + position_ids = position_ids.view(1, -1).expand(batch_size, -1) + if cache_position is not None: # otherwise `deltas` is an int `0` + delta = delta.repeat_interleave(batch_size // delta.shape[0], dim=0) + position_ids = position_ids.add(delta) + position_ids = position_ids.unsqueeze(0).expand(3, -1, -1) + + outputs = self.model( # Qwen2_5_VLModel + input_ids=None, + position_ids=position_ids, + attention_mask=attention_mask, + past_key_values=past_key_values, + inputs_embeds=inputs_embeds, + use_cache=use_cache, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + cache_position=cache_position, + ) + + hidden_states = outputs[0] + logits = self.lm_head(hidden_states) + if self.cp_mesh is not None: + logits = DTensor.from_local(logits, device_mesh=self.cp_mesh, placements=[Shard(1)]).full_tensor() # noqa: F821 + return logits, outputs + + """ + Copy from QwenModel.forward with MODIFICATIONS + MODIFICATIONS: adding the hidden states to the output. + """ + + def forward(self, tokens, data_batch={}, start_pos: int = 0) -> torch.Tensor: + """ + The training step of the model, including the loss computation. + """ + assert "pixel_values" not in data_batch, "pixel_values should not be in data_batch, use images instead" + pixel_values = data_batch.get("images", None) + image_grid_thw = data_batch.get("image_grid_thw", None) + pixel_values_videos = data_batch.get("videos", None) + video_grid_thw = data_batch.get("video_grid_thw", None) + attention_mask = data_batch.get("padding_mask", None) + + if image_grid_thw is not None: + assert len(image_grid_thw) == 1, "Only batch=1 is supported for now, due to `get_rope_index`" + image_grid_thw = image_grid_thw[0] # 1, N_img, 3 -> N_img, 3 + if video_grid_thw is not None: + assert len(video_grid_thw) == 1, "Only batch=1 is supported for now, due to `get_rope_index`" + video_grid_thw = video_grid_thw[0] # 1, N_video, 3 -> N_video, 3 + logits, outputs = self._forward( + input_ids=tokens, + pixel_values=pixel_values, + image_grid_thw=image_grid_thw, + pixel_values_videos=pixel_values_videos, + video_grid_thw=video_grid_thw, + attention_mask=attention_mask, + ) + return logits, outputs diff --git a/REGEN-main/cosmos_policy/_src/predict2/text_encoders/text_encoder.py b/REGEN-main/cosmos_policy/_src/predict2/text_encoders/text_encoder.py new file mode 100644 index 0000000000000000000000000000000000000000..edd363b30c2417052a9038bd5e0e494252d26165 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/text_encoders/text_encoder.py @@ -0,0 +1,238 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os + +import attrs +import torch +from torch.distributed.checkpoint.state_dict import StateDictOptions, set_model_state_dict + +from cosmos_policy._src.imaginaire.flags import SMOKE +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import instantiate as lazy_instantiate +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.embedding_concat_strategy import ( + EmbeddingConcatStrategy as EmbeddingConcatStrategy, +) +from cosmos_policy._src.predict2.models.utils import load_state_dict, load_state_dict_from_folder +from cosmos_policy._src.predict2.text_encoders.reason1 import QwenVLBaseModel +from cosmos_policy._src.reason1.configs.default.model_config_qwen import QwenModelConfig, QwenVisionConfig +from cosmos_policy._src.reason1.tokenizer.processor import build_tokenizer + +NUM_EMBEDDING_PADDING_TOKENS = 512 + + +@attrs.define(slots=False) +class TextEncoderConfig: + """ + Config for the text encoder model + """ + + compute_online: bool = False + embedding_concat_strategy: str = str(EmbeddingConcatStrategy.MEAN_POOLING) + n_layers_per_group: int = 5 + ckpt_path: str = "s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/" + s3_credential_path: str = "credentials/s3_checkpoint.secret" + model_config: QwenVLBaseModel = L(QwenVLBaseModel)( + model_config=L(QwenModelConfig)( + tokenizer_type="Qwen/Qwen2.5-VL-7B-Instruct", + name_or_path="Qwen/Qwen2.5-VL-7B-Instruct", + hidden_size=3584, + intermediate_size=18944, + max_window_layers=28, + num_attention_heads=28, + num_hidden_layers=28, + num_key_value_heads=4, + tie_word_embeddings=False, + vocab_size=152064, + vision_config=L(QwenVisionConfig)(out_hidden_size=3584), + output_hidden_states=True, + ), + tokenizer=L(build_tokenizer)( + tokenizer_type="Qwen/Qwen2.5-VL-7B-Instruct", + ), + ) + + +class TextEncoder: + def __init__(self, config: TextEncoderConfig, device: str = "cuda"): + self.config = config + self.device = device + + log.info("Instantiating text encoder model...") + with torch.device("meta"): + self.model = lazy_instantiate(self.config.model_config) + self.model.to_empty(device=self.device) + if SMOKE: + return + with torch.no_grad(): + self.model.init_weights() + from cosmos_policy._src.imaginaire.utils.checkpoint_db import get_checkpoint_path + + log.info(f"Loading checkpoint from {self.config.ckpt_path}.") + ckpt_path = get_checkpoint_path(self.config.ckpt_path) + if torch.distributed.is_initialized(): + torch.distributed.barrier() + is_fsdp = torch.distributed.get_world_size() > 1 + else: + is_fsdp = False + if os.path.isdir(ckpt_path): + state_dict = load_state_dict_from_folder(ckpt_path) + else: + state_dict = load_state_dict(ckpt_path) + # remove _extra_state + state_dict = {k: v for k, v in state_dict.items() if not k.endswith("._extra_state")} + + # Load Regular weights. + if is_fsdp: + set_model_state_dict( + self.model, + state_dict, + options=StateDictOptions( + full_state_dict=True, + broadcast_from_rank0=True, + strict=False, + ), + ) + else: + self.model.load_state_dict(state_dict, strict=False) + + del state_dict + log.info(f"Finished loading checkpoint from {ckpt_path}.") + self.model.eval() + torch.cuda.empty_cache() + log.info("Text encoder model instantiated") + + @staticmethod + def mean_normalize(tensor: torch.Tensor) -> torch.Tensor: + """ + Mean normalize a tensor by subtracting the mean and dividing by the standard deviation. + + Args: + tensor (torch.tensor): The tensor to normalize + + Returns: + torch.tensor: The normalized tensor + """ + return (tensor - tensor.mean(dim=-1, keepdim=True)) / (tensor.std(dim=-1, keepdim=True) + 1e-8) + + def compute_text_embeddings_online( + self, data_batch: dict[str, torch.Tensor], input_caption_key: str + ) -> torch.Tensor: + """ + Compute text embeddings for the given prompts. + """ + assert self.model is not None, "Text encoder is not initialized" + + # Tokenize prompts + input_ids_batch = [] + + for sample_idx in range(len(data_batch[input_caption_key])): + conversations = [ + { + "role": "system", + "content": [ + { + "type": "text", + "text": "You are a helpful assistant who will provide prompts to an image generator.", + } + ], + }, + { + "role": "user", + "content": [ + { + "type": "text", + "text": data_batch[input_caption_key][sample_idx], + } + ], + }, + ] + tokenizer_output = self.model.tokenizer.apply_chat_template( + conversations, + tokenize=True, + add_generation_prompt=False, + add_vision_id=False, + ) + input_ids = tokenizer_output["input_ids"] + pad_id = self.model.tokenizer.pad_id + + # Do padding or truncation + if NUM_EMBEDDING_PADDING_TOKENS > len(input_ids): + # Do padding: + pad_len = NUM_EMBEDDING_PADDING_TOKENS - len(input_ids) + input_ids = input_ids.tolist() + [pad_id] * pad_len + else: + # Do truncation: + input_ids = input_ids.tolist()[:NUM_EMBEDDING_PADDING_TOKENS] + input_ids = torch.LongTensor(input_ids).to(device="cuda") + input_ids_batch.append(input_ids) + + input_ids_batch = torch.stack(input_ids_batch, dim=0) + + # Compute text embeddings + self.model = self.model.to(self.device) + with torch.no_grad(): + _, outputs_batch = self.model(input_ids_batch, {}) + hidden_states = outputs_batch["hidden_states"] + + # # Skip the embeddings of the system prompt + # hidden_states = hidden_states[:, num_system_prompt_tokens:] + + # Now compute the normalized embeddings + normalized_hidden_states = [] + for layer_idx in range(1, len(hidden_states)): + normalized_state = self.mean_normalize(hidden_states[layer_idx]) + normalized_hidden_states.append(normalized_state) + + text_embeddings = None + if self.config.embedding_concat_strategy == str(EmbeddingConcatStrategy.FULL_CONCAT): + text_embeddings = torch.cat(normalized_hidden_states, dim=-1) + elif self.config.embedding_concat_strategy == str(EmbeddingConcatStrategy.MEAN_POOLING): + # Stack the normalized hidden states and calculate the mean + text_embeddings = torch.stack(normalized_hidden_states) + text_embeddings = text_embeddings.mean(dim=0) + elif self.config.embedding_concat_strategy == str(EmbeddingConcatStrategy.POOL_EVERY_N_LAYERS_AND_CONCAT): + # Split the l + n_layers_per_group = self.config.n_layers_per_group + text_embeddings = [] + for i in range(0, len(normalized_hidden_states), n_layers_per_group): + group_embeddings = normalized_hidden_states[i : i + n_layers_per_group] + group_embedding = torch.stack(group_embeddings) + group_embedding = group_embedding.mean(dim=0) + text_embeddings.append(group_embedding) + text_embeddings = torch.cat(text_embeddings, dim=-1) + else: + raise ValueError(f"Invalid embedding_concat_strategy: {self.config.embedding_concat_strategy}") + + return text_embeddings + + +def get_reason1_embeddings(text: str): + """ + Get reason1 embeddings for a given text. + Output (1, seq len, d) embeddings + """ + config = TextEncoderConfig( + embedding_concat_strategy="full_concat", + ) + text_encoder = TextEncoder(config) + text_embeddings = text_encoder.compute_text_embeddings_online( + { + "text": [text], + }, + "text", + ) + return text_embeddings diff --git a/REGEN-main/cosmos_policy/_src/predict2/text_encoders/text_encoder_test.py b/REGEN-main/cosmos_policy/_src/predict2/text_encoders/text_encoder_test.py new file mode 100644 index 0000000000000000000000000000000000000000..9db3e0fc8bd0a2cb044837dea45442e62f2ad153 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/text_encoders/text_encoder_test.py @@ -0,0 +1,110 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Unit tests for the TextEncoder class. + +Usage: + pytest -s cosmos_policy/_src/predict2/text_encoders/text_encoder_test.py --L0 + pytest -s cosmos_policy/_src/predict2/text_encoders/text_encoder_test.py --L1 +""" + +import unittest + +import pytest +import torch + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.embedding_concat_strategy import EmbeddingConcatStrategy +from cosmos_policy._src.predict2.text_encoders.text_encoder import TextEncoder, TextEncoderConfig + + +class TestTextEncoder(unittest.TestCase): + """Test the TextEncoder class.""" + + def setUp(self): + """Set up test fixtures.""" + self.config = TextEncoderConfig( + compute_online=True, + embedding_concat_strategy=str(EmbeddingConcatStrategy.MEAN_POOLING), + n_layers_per_group=2, + ) + + @pytest.mark.L0 + def test_mean_normalize(self): + """Test the mean_normalize static method.""" + # Create a test tensor + tensor = torch.tensor([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]]) + + # Apply mean normalization + normalized = TextEncoder.mean_normalize(tensor) + + # Check that the result has the same shape + assert normalized.shape == tensor.shape + + # Check that each row has mean close to 0 and std close to 1 + for i in range(tensor.shape[0]): + assert abs(normalized[i].mean().item()) < 1e-6 + assert abs(normalized[i].std().item() - 1.0) < 1e-6 + + @pytest.mark.L1 + def test_compute_text_embeddings_online_full_concat(self): + """Test text embedding computation with FULL_CONCAT strategy.""" + # Skip if CUDA is not available + if not torch.cuda.is_available(): + pytest.skip("CUDA not available") + + # Create encoder with FULL_CONCAT strategy + config = TextEncoderConfig( + compute_online=True, + embedding_concat_strategy=str(EmbeddingConcatStrategy.FULL_CONCAT), + n_layers_per_group=2, + ) + encoder = TextEncoder(config) + + # Test data + data_batch = {"input_caption": ["A beautiful sunset", "A cat playing"]} + + # Compute embeddings + embeddings = encoder.compute_text_embeddings_online(data_batch, "input_caption") + + log.info(f"Embeddings shape: {embeddings.shape}") + + # Verify the output is a tensor + assert isinstance(embeddings, torch.Tensor) + assert embeddings.dim() == 3 # [batch_size, seq_len, hidden_dim] + assert embeddings.shape[0] == 2 # batch_size + assert embeddings.shape[1] == 512 # sequence length + assert embeddings.shape[2] == 3584 * 28 # hidden dimension (num_layers * hidden_dim) + + @pytest.mark.L0 + def test_config_defaults(self): + """Test TextEncoderConfig default values.""" + config = TextEncoderConfig() + + assert config.compute_online is False + assert config.embedding_concat_strategy == str(EmbeddingConcatStrategy.MEAN_POOLING) + assert config.n_layers_per_group == 5 + assert "s3://bucket/cosmos_reasoning1" in config.ckpt_path + assert config.model_config is not None + + +if __name__ == "__main__": + # Set up test environment + torch.backends.cudnn.deterministic = True + torch.backends.cudnn.benchmark = False + + # Run tests + unittest.main(verbosity=2) diff --git a/REGEN-main/cosmos_policy/_src/predict2/tokenizers/base_vae.py b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/base_vae.py new file mode 100644 index 0000000000000000000000000000000000000000..f131db9c2f1591c76ced23ec7afd6d2c1856317c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/base_vae.py @@ -0,0 +1,462 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +import os +from abc import ABC, abstractmethod +from typing import Optional + +import torch +import torch.nn.functional as F + +from cosmos_policy._src.imaginaire.utils.distributed import rank0_first +from cosmos_policy._src.imaginaire.utils.env_parsers.cred_env_parser import CRED_ENVS +from cosmos_policy._src.imaginaire.utils.s3_utils import load_from_s3_with_cache + + +class BaseVAE(torch.nn.Module, ABC): + """ + Abstract base class for a Variational Autoencoder (VAE). + + All subclasses should implement the methods to define the behavior for encoding + and decoding, along with specifying the latent channel size. + """ + + def __init__(self, channel: int = 3, name: str = "vae"): + super().__init__() + self.channel = channel + self.name = name + + @property + def latent_ch(self) -> int: + """ + Returns the number of latent channels in the VAE. + """ + return self.channel + + @abstractmethod + def encode(self, state: torch.Tensor) -> torch.Tensor: + """ + Encodes the input tensor into a latent representation. + + Args: + - state (torch.Tensor): The input tensor to encode. + + Returns: + - torch.Tensor: The encoded latent tensor. + """ + pass + + @abstractmethod + def decode(self, latent: torch.Tensor) -> torch.Tensor: + """ + Decodes the latent representation back to the original space. + + Args: + - latent (torch.Tensor): The latent tensor to decode. + + Returns: + - torch.Tensor: The decoded tensor. + """ + pass + + @property + def spatial_compression_factor(self) -> int: + """ + Returns the spatial reduction factor for the VAE. + """ + raise NotImplementedError("The spatial_compression_factor property must be implemented in the derived class.") + + +class BasePretrainedImageVAE(BaseVAE): + """ + A base class for pretrained Variational Autoencoder (VAE) that loads mean and standard deviation values + from a remote store, handles data type conversions, and normalization + using provided mean and standard deviation values for latent space representation. + Derived classes should load pre-trained encoder and decoder components from a remote store + + Attributes: + latent_mean (Tensor): The mean used for normalizing the latent representation. + latent_std (Tensor): The standard deviation used for normalizing the latent representation. + dtype (dtype): Data type for model tensors, determined by whether bf16 is enabled. + + Args: + mean_std_fp (str): File path to the pickle file containing mean and std of the latent space. + latent_ch (int, optional): Number of latent channels (default is 16). + is_image (bool, optional): Flag to indicate whether the output is an image (default is True). + is_bf16 (bool, optional): Flag to use Brain Floating Point 16-bit data type (default is True). + """ + + def __init__( + self, + name: str, + mean_std_fp: str, + latent_ch: int = 16, + is_image: bool = True, + is_bf16: bool = True, + s3_credential_path=Optional[str], + load_mean_std: bool = True, + ) -> None: + super().__init__(latent_ch, name) + dtype = torch.bfloat16 if is_bf16 else torch.float32 + self.dtype = dtype + self.is_image = is_image + self.mean_std_fp = mean_std_fp + self.name = name + self.load_mean_std = load_mean_std + + assert s3_credential_path is None or isinstance(s3_credential_path, str) + if s3_credential_path is None: + self.backend_args = None + elif os.path.exists(s3_credential_path) or CRED_ENVS.APP_ENV in ["prod", "dev", "stg"]: + self.backend_args = { + "backend": "s3", + "path_mapping": None, + "s3_credential_path": s3_credential_path, + } + else: + raise FileNotFoundError(f"Invalid s3_credential_path: {s3_credential_path} and APP_ENV is not prod/dev/stg") + + self.register_mean_std(mean_std_fp) + + def register_mean_std(self, mean_std_fp: str) -> None: + target_shape = [1, self.latent_ch, 1, 1] if self.is_image else [1, self.latent_ch, 1, 1, 1] + + if self.load_mean_std: + extention = mean_std_fp.split(".")[-1] + latent_mean, latent_std = load_from_s3_with_cache( + mean_std_fp, + f"vae/{self.name}_mean_std.{extention}", + easy_io_kwargs={"map_location": torch.device(torch.cuda.current_device())}, + backend_args=self.backend_args, + ) + self.register_buffer( + "latent_mean", + latent_mean.to(self.dtype).reshape(*target_shape), + persistent=False, + ) + self.register_buffer( + "latent_std", + latent_std.to(self.dtype).reshape(*target_shape), + persistent=False, + ) + else: + # Use zeros for mean and ones for std when load_mean_std=False + device = torch.device(torch.cuda.current_device()) if torch.cuda.is_available() else torch.device("cpu") + self.register_buffer( + "latent_mean", + torch.zeros(*target_shape, dtype=self.dtype, device=device), + persistent=False, + ) + self.register_buffer( + "latent_std", + torch.ones(*target_shape, dtype=self.dtype, device=device), + persistent=False, + ) + + @torch.no_grad() + def encode(self, state: torch.Tensor) -> torch.Tensor: + """ + Encode the input state to latent space; also handle the dtype conversion, mean and std scaling + """ + in_dtype = state.dtype + latent_mean = self.latent_mean.to(in_dtype) + latent_std = self.latent_std.to(in_dtype) + encoded_state = self.encoder(state.to(self.dtype)) + if isinstance(encoded_state, torch.Tensor): + pass + elif isinstance(encoded_state, tuple): + assert isinstance(encoded_state[0], torch.Tensor) + encoded_state = encoded_state[0] + else: + raise ValueError("Invalid type of encoded state") + return (encoded_state.to(in_dtype) - latent_mean) / latent_std + + @torch.no_grad() + def decode(self, latent: torch.Tensor) -> torch.Tensor: + """ + Decode the input latent to state; also handle the dtype conversion, mean and std scaling + """ + in_dtype = latent.dtype + latent = latent * self.latent_std.to(in_dtype) + self.latent_mean.to(in_dtype) + return self.decoder(latent.to(self.dtype)).to(in_dtype) + + def reset_dtype(self, *args, **kwargs): + """ + Resets the data type of the encoder and decoder to the model's default data type. + + Args: + *args, **kwargs: Unused, present to allow flexibility in method calls. + """ + del args, kwargs + self.decoder.to(self.dtype) + self.encoder.to(self.dtype) + + +class JITVAE(BasePretrainedImageVAE): + """ + A JIT compiled Variational Autoencoder (VAE) that loads pre-trained encoder + and decoder components from a remote store, handles data type conversions, and normalization + using provided mean and standard deviation values for latent space representation. + + Attributes: + encoder (Module): The JIT compiled encoder loaded from storage. + decoder (Module): The JIT compiled decoder loaded from storage. + latent_mean (Tensor): The mean used for normalizing the latent representation. + latent_std (Tensor): The standard deviation used for normalizing the latent representation. + dtype (dtype): Data type for model tensors, determined by whether bf16 is enabled. + + Args: + enc_fp (str): File path to the encoder's JIT file on the remote store. + dec_fp (str): File path to the decoder's JIT file on the remote store. + name (str): Name of the model, used for differentiating cache file paths. + mean_std_fp (str): File path to the pickle file containing mean and std of the latent space. + latent_ch (int, optional): Number of latent channels (default is 16). + is_image (bool, optional): Flag to indicate whether the output is an image (default is True). + is_bf16 (bool, optional): Flag to use Brain Floating Point 16-bit data type (default is True). + """ + + def __init__( + self, + enc_fp: str, + dec_fp: str, + name: str, + mean_std_fp: str, + s3_credential_path: Optional[str] = None, + latent_ch: int = 16, + is_image: bool = True, + is_bf16: bool = True, + load_mean_std: bool = True, + ): + super().__init__( + name, + mean_std_fp, + latent_ch, + is_image, + is_bf16, + s3_credential_path=s3_credential_path, + load_mean_std=load_mean_std, + ) + self.load_encoder(enc_fp) + self.load_decoder(dec_fp) + + def load_encoder(self, enc_fp: str) -> None: + """ + Load the encoder from the remote store. + + Args: + - enc_fp (str): File path to the encoder's JIT file on the remote store. + """ + self.encoder = load_from_s3_with_cache( + enc_fp, + f"vae/{self.name}_enc.jit", + easy_io_kwargs={"map_location": torch.device(torch.cuda.current_device())}, + backend_args=self.backend_args, + ) + self.encoder.eval() + for param in self.encoder.parameters(): + param.requires_grad = False + self.encoder.to(self.dtype) + + def load_decoder(self, dec_fp: str) -> None: + """ + Load the decoder from the remote store. + + Args: + - dec_fp (str): File path to the decoder's JIT file on the remote store. + """ + self.decoder = load_from_s3_with_cache( + dec_fp, + f"vae/{self.name}_dec.jit", + easy_io_kwargs={"map_location": torch.device(torch.cuda.current_device())}, + backend_args=self.backend_args, + ) + self.decoder.eval() + for param in self.decoder.parameters(): + param.requires_grad = False + self.decoder.to(self.dtype) + + +class StateDictVAE(BasePretrainedImageVAE): + """ + A Variational Autoencoder (VAE) that loads pre-trained weights into + provided encoder and decoder components from a remote store, handles data type conversions, + and normalization using provided mean and standard deviation values for latent space representation. + + Attributes: + encoder (Module): The encoder with weights loaded from storage. + decoder (Module): The decoder with weights loaded from storage. + latent_mean (Tensor): The mean used for normalizing the latent representation. + latent_std (Tensor): The standard deviation used for normalizing the latent representation. + dtype (dtype): Data type for model tensors, determined by whether bf16 is enabled. + + Args: + enc_fp (str): File path to the encoder's JIT file on the remote store. + dec_fp (str): File path to the decoder's JIT file on the remote store. + vae (Module): Instance of VAE with not loaded weights + name (str): Name of the model, used for differentiating cache file paths. + mean_std_fp (str): File path to the pickle file containing mean and std of the latent space. + latent_ch (int, optional): Number of latent channels (default is 16). + is_image (bool, optional): Flag to indicate whether the output is an image (default is True). + is_bf16 (bool, optional): Flag to use Brain Floating Point 16-bit data type (default is True). + """ + + def __init__( + self, + enc_fp: str, + dec_fp: str, + vae: torch.nn.Module, + name: str, + mean_std_fp: str, + s3_credential_path: Optional[str] = None, + latent_ch: int = 16, + is_image: bool = True, + is_bf16: bool = True, + ): + super().__init__(name, mean_std_fp, latent_ch, is_image, is_bf16, s3_credential_path=s3_credential_path) + + self.load_encoder_and_decoder(enc_fp, dec_fp, vae) + + def load_encoder_and_decoder(self, enc_fp: str, dec_fp: str, vae: torch.nn.Module) -> None: + """ + Load the encoder from the remote store. + + Args: + - vae_fp (str): File path to the vae's state dict file on the remote store. + - vae (str): VAE module into which weights will be loaded. + """ + state_dict_enc = load_from_s3_with_cache( + enc_fp, + f"vae/{self.name}_enc.jit", + easy_io_kwargs={"map_location": torch.device(torch.cuda.current_device())}, + backend_args=self.backend_args, + ) + + state_dict_dec = load_from_s3_with_cache( + dec_fp, + f"vae/{self.name}_dec.jit", + easy_io_kwargs={"map_location": torch.device(torch.cuda.current_device())}, + backend_args=self.backend_args, + ) + + jit_weights_state_dict = state_dict_enc.state_dict() | state_dict_dec.state_dict() + jit_weights_state_dict = { + k: v + for k, v in jit_weights_state_dict.items() + # Global variables captured by JIT + if k + not in ( + "encoder.patcher.wavelets", + "encoder.patcher._arange", + "decoder.unpatcher.wavelets", + "decoder.unpatcher._arange", + ) + } + + vae.load_state_dict(jit_weights_state_dict) + vae.eval() + for param in vae.parameters(): + param.requires_grad = False + vae.to(self.dtype) + + self.vae = vae + self.encoder = self.vae.encode + self.decoder = self.vae.decode + + def reset_dtype(self, *args, **kwargs): + """ + Resets the data type of the encoder and decoder to the model's default data type. + + Args: + *args, **kwargs: Unused, present to allow flexibility in method calls. + """ + del args, kwargs + self.vae.to(self.dtype) + + +class SDVAE(BaseVAE): + def __init__(self, batch_size=16, count_std: bool = False, is_downsample: bool = True) -> None: + super().__init__(channel=4, name="sd_vae") + self.dtype = torch.bfloat16 + self.register_buffer( + "scale", + torch.tensor([4.17, 4.62, 3.71, 3.28], dtype=self.dtype).reciprocal().reshape(1, -1, 1, 1), + persistent=False, + ) + self.register_buffer( + "bias", + -1.0 * torch.tensor([5.81, 3.25, 0.12, -2.15], dtype=self.dtype).reshape(1, -1, 1, 1) * self.scale, + persistent=False, + ) + self.batch_size = batch_size + self.count_std = count_std + self.is_downsample = is_downsample + self.load_vae() + self.reset_dtype() + + def reset_dtype(self, *args, **kwargs): + del args, kwargs + self.vae.to(self.dtype) + + @rank0_first + def load_vae(self) -> None: + os.environ["HF_HUB_DISABLE_SYMLINKS_WARNING"] = "1" + os.environ["HF_HUB_DISABLE_PROGRESS_BARS"] = "1" + import diffusers + + vae_name = "stabilityai/sd-vae-ft-mse" + try: + vae = diffusers.models.AutoencoderKL.from_pretrained(vae_name, local_files_only=True) + except: # noqa: E722 + # Could not load the model from cache; try without local_files_only. + vae = diffusers.models.AutoencoderKL.from_pretrained(vae_name) + self.vae = vae.eval().requires_grad_(False) + + @torch.no_grad() + def encode(self, state: torch.Tensor) -> torch.Tensor: + """ + state : pixel range [-1, 1] + """ + if self.is_downsample: + _h, _w = state.shape[-2:] + state = F.interpolate(state, size=(_h // 2, _w // 2), mode="bilinear", align_corners=False) + in_dtype = state.dtype + state = state.to(self.dtype) + state = (state + 1.0) / 2.0 + latent_dist = self.vae.encode(state)["latent_dist"] + mean, std = latent_dist.mean, latent_dist.std + if self.count_std: + latent = mean + torch.randn_like(mean) * std + else: + latent = mean + latent = latent * self.scale + latent = latent + self.bias + return latent.to(in_dtype) + + @torch.no_grad() + def decode(self, latent: torch.Tensor) -> torch.Tensor: + in_dtype = latent.dtype + latent = latent.to(self.dtype) + latent = latent - self.bias + latent = latent / self.scale + latent = torch.cat([self.vae.decode(batch)["sample"] for batch in latent.split(self.batch_size)]) + if self.is_downsample: + _h, _w = latent.shape[-2:] + latent = F.interpolate(latent, size=(_h * 2, _w * 2), mode="bilinear", align_corners=False) + return latent.to(in_dtype) * 2 - 1.0 + + @property + def spatial_compression_factor(self) -> int: + return 8 diff --git a/REGEN-main/cosmos_policy/_src/predict2/tokenizers/cosmos.py b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/cosmos.py new file mode 100644 index 0000000000000000000000000000000000000000..dd30211e7107c94edfee636b89bd893320d69961 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/cosmos.py @@ -0,0 +1,28 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.predict2.tokenizers.wan2pt1 import Wan2pt1VAEInterface +from cosmos_policy._src.predict2.tokenizers.wan2pt2 import Wan2pt2VAEInterface + +Wan2pt1VAEConfig: LazyDict = L(Wan2pt1VAEInterface)(name="wan2pt1_tokenizer") +Wan2pt1VAEConfig_GCP: LazyDict = L(Wan2pt1VAEInterface)( + name="wan2pt1_tokenizer_gcp", + s3_credential_path="credentials/gcp_training.secret", + vae_pth="s3://bucket/cosmos_diffusion_v2/pretrain_weights/tokenizer/wan2pt1/Wan2.1_VAE.pth", +) +Wan2pt2VAEConfig: LazyDict = L(Wan2pt2VAEInterface)(name="wan2pt2_tokenizer") diff --git a/REGEN-main/cosmos_policy/_src/predict2/tokenizers/interface.py b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/interface.py new file mode 100644 index 0000000000000000000000000000000000000000..eb8a3f83293c760356b835a15b6939533215b97b --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/interface.py @@ -0,0 +1,98 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os +from abc import ABC, abstractmethod +from typing import Optional + +import torch + +from cosmos_policy._src.imaginaire.utils.env_parsers.cred_env_parser import CRED_ENVS + + +class VideoTokenizerInterface(ABC): + def __init__(self, s3_credential_path: Optional[str] = None): + assert s3_credential_path is None or isinstance(s3_credential_path, str) + if s3_credential_path is None: + self.backend_args = None + elif os.path.exists(s3_credential_path) or CRED_ENVS.APP_ENV in ["prod", "dev", "stg"]: + self.backend_args = { + "backend": "s3", + "path_mapping": None, + "s3_credential_path": s3_credential_path, + } + else: + raise FileNotFoundError(f"Invalid s3_credential_path: {s3_credential_path} and APP_ENV is not prod/dev/stg") + + @abstractmethod + def reset_dtype(self): + """ + Reset the dtype of the model to the dtype its weights were trained with or quantized to. + """ + pass + + @abstractmethod + def encode(self, state: torch.Tensor) -> torch.Tensor: + pass + + @abstractmethod + def decode(self, latent: torch.Tensor) -> torch.Tensor: + pass + + @abstractmethod + def get_latent_num_frames(self, num_pixel_frames: int) -> int: + pass + + @abstractmethod + def get_pixel_num_frames(self, num_latent_frames: int) -> int: + pass + + @property + @abstractmethod + def spatial_compression_factor(self): + pass + + @property + @abstractmethod + def temporal_compression_factor(self): + pass + + @property + @abstractmethod + def spatial_resolution(self): + pass + + @property + @abstractmethod + def pixel_chunk_duration(self): + pass + + @property + @abstractmethod + def latent_chunk_duration(self): + pass + + @property + @abstractmethod + def latent_ch(self) -> int: + pass + + @property + def is_chunk_overlap(self): + return False + + @property + def is_causal(self): + return True diff --git a/REGEN-main/cosmos_policy/_src/predict2/tokenizers/wan2pt1.py b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/wan2pt1.py new file mode 100644 index 0000000000000000000000000000000000000000..3f028e8c9a7623aa5104aad64f4681315630bf71 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/wan2pt1.py @@ -0,0 +1,1060 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved. + +import time +from contextlib import nullcontext +from typing import Optional + +import torch +import torch.distributed as distributed +import torch.nn as nn +import torch.nn.functional as F +from einops import rearrange +from megatron.core import parallel_state + +from cosmos_policy._src.imaginaire.flags import INTERNAL, SMOKE +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.distributed import broadcast, get_rank, sync_model_states +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.predict2.tokenizers.interface import VideoTokenizerInterface +from cosmos_policy._src.predict2.tokenizers.wan2pt1_2d_plugins import plugin_mount +from cosmos_policy._src.predict2.utils.tokenizer_benchmarking import BenchmarkTimes + +__all__ = [ + "WanVAE", +] + +CACHE_T = 2 + + +class CausalConv3d(nn.Conv3d): + """ + Causal 3d convolusion. + """ + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + self._padding = (self.padding[2], self.padding[2], self.padding[1], self.padding[1], 2 * self.padding[0], 0) + self.padding = (0, 0, 0) + + def forward(self, x, cache_x=None): + padding = list(self._padding) + if cache_x is not None and self._padding[4] > 0: + cache_x = cache_x.to(x.device) + x = torch.cat([cache_x, x], dim=2) + padding[4] -= cache_x.shape[2] + x = F.pad(x, padding) + + return super().forward(x) + + +class RMS_norm(nn.Module): + def __init__(self, dim, channel_first=True, images=True, bias=False): + super().__init__() + broadcastable_dims = (1, 1, 1) if not images else (1, 1) + shape = (dim, *broadcastable_dims) if channel_first else (dim,) + + self.channel_first = channel_first + self.scale = dim**0.5 + self.gamma = nn.Parameter(torch.ones(shape)) + self.bias = nn.Parameter(torch.zeros(shape)) if bias else 0.0 + + def forward(self, x): + return F.normalize(x, dim=(1 if self.channel_first else -1)) * self.scale * self.gamma + self.bias + + +class Upsample(nn.Upsample): + def forward(self, x): + """ + Fix bfloat16 support for nearest neighbor interpolation. + """ + return super().forward(x.float()).type_as(x) + + +class Resample(nn.Module): + def __init__(self, dim, mode): + assert mode in ("none", "upsample2d", "upsample3d", "downsample2d", "downsample3d") + super().__init__() + self.dim = dim + self.mode = mode + + # layers + if mode == "upsample2d": + self.resample = nn.Sequential( + Upsample(scale_factor=(2.0, 2.0), mode="nearest-exact"), nn.Conv2d(dim, dim // 2, 3, padding=1) + ) + elif mode == "upsample3d": + self.resample = nn.Sequential( + Upsample(scale_factor=(2.0, 2.0), mode="nearest-exact"), nn.Conv2d(dim, dim // 2, 3, padding=1) + ) + self.time_conv = CausalConv3d(dim, dim * 2, (3, 1, 1), padding=(1, 0, 0)) + + elif mode == "downsample2d": + self.resample = nn.Sequential(nn.ZeroPad2d((0, 1, 0, 1)), nn.Conv2d(dim, dim, 3, stride=(2, 2))) + elif mode == "downsample3d": + self.resample = nn.Sequential(nn.ZeroPad2d((0, 1, 0, 1)), nn.Conv2d(dim, dim, 3, stride=(2, 2))) + self.time_conv = CausalConv3d(dim, dim, (3, 1, 1), stride=(2, 1, 1), padding=(0, 0, 0)) + + else: + self.resample = nn.Identity() + + def forward(self, x, feat_cache=None, feat_idx=[0]): + b, c, t, h, w = x.size() + if self.mode == "upsample3d": + if feat_cache is not None: + idx = feat_idx[0] + if feat_cache[idx] is None: + feat_cache[idx] = "Rep" + feat_idx[0] += 1 + else: + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None and feat_cache[idx] != "Rep": + # cache last frame of last two chunk + cache_x = torch.cat( + [feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x], dim=2 + ) + if cache_x.shape[2] < 2 and feat_cache[idx] is not None and feat_cache[idx] == "Rep": + cache_x = torch.cat([torch.zeros_like(cache_x).to(cache_x.device), cache_x], dim=2) + if feat_cache[idx] == "Rep": + x = self.time_conv(x) + else: + x = self.time_conv(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + + x = x.reshape(b, 2, c, t, h, w) + x = torch.stack((x[:, 0, :, :, :, :], x[:, 1, :, :, :, :]), 3) + x = x.reshape(b, c, t * 2, h, w) + t = x.shape[2] + x = rearrange(x, "b c t h w -> (b t) c h w") + x = self.resample(x) + x = rearrange(x, "(b t) c h w -> b c t h w", t=t) + + if self.mode == "downsample3d": + if feat_cache is not None: + idx = feat_idx[0] + if feat_cache[idx] is None: + feat_cache[idx] = x.clone() + feat_idx[0] += 1 + else: + cache_x = x[:, :, -1:, :, :].clone() + # if cache_x.shape[2] < 2 and feat_cache[idx] is not None and feat_cache[idx]!='Rep': + # # cache last frame of last two chunk + # cache_x = torch.cat([feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x], dim=2) + + x = self.time_conv(torch.cat([feat_cache[idx][:, :, -1:, :, :], x], 2)) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + return x + + def init_weight(self, conv): + conv_weight = conv.weight + nn.init.zeros_(conv_weight) + c1, c2, t, h, w = conv_weight.size() + one_matrix = torch.eye(c1, c2) + init_matrix = one_matrix + nn.init.zeros_(conv_weight) + # conv_weight.data[:,:,-1,1,1] = init_matrix * 0.5 + conv_weight.data[:, :, 1, 0, 0] = init_matrix # * 0.5 + conv.weight.data.copy_(conv_weight) + nn.init.zeros_(conv.bias.data) + + def init_weight2(self, conv): + conv_weight = conv.weight.data + nn.init.zeros_(conv_weight) + c1, c2, t, h, w = conv_weight.size() + init_matrix = torch.eye(c1 // 2, c2) + # init_matrix = repeat(init_matrix, 'o ... -> (o 2) ...').permute(1,0,2).contiguous().reshape(c1,c2) + conv_weight[: c1 // 2, :, -1, 0, 0] = init_matrix + conv_weight[c1 // 2 :, :, -1, 0, 0] = init_matrix + conv.weight.data.copy_(conv_weight) + nn.init.zeros_(conv.bias.data) + + +class ResidualBlock(nn.Module): + def __init__(self, in_dim, out_dim, dropout=0.0): + super().__init__() + self.in_dim = in_dim + self.out_dim = out_dim + + # layers + self.residual = nn.Sequential( + RMS_norm(in_dim, images=False), + nn.SiLU(), + CausalConv3d(in_dim, out_dim, 3, padding=1), + RMS_norm(out_dim, images=False), + nn.SiLU(), + nn.Dropout(dropout), + CausalConv3d(out_dim, out_dim, 3, padding=1), + ) + self.shortcut = CausalConv3d(in_dim, out_dim, 1) if in_dim != out_dim else nn.Identity() + + def forward(self, x, feat_cache=None, feat_idx=[0]): + h = self.shortcut(x) + for layer in self.residual: + if isinstance(layer, CausalConv3d) and feat_cache is not None: + idx = feat_idx[0] + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None: + # cache last frame of last two chunk + cache_x = torch.cat( + [feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x], dim=2 + ) + x = layer(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + else: + x = layer(x) + return x + h + + +class AttentionBlock(nn.Module): + """ + Causal self-attention with a single head. + """ + + def __init__(self, dim): + super().__init__() + self.dim = dim + + # layers + self.norm = RMS_norm(dim) + self.to_qkv = nn.Conv2d(dim, dim * 3, 1) + self.proj = nn.Conv2d(dim, dim, 1) + + # zero out the last layer params + nn.init.zeros_(self.proj.weight) + + def forward(self, x): + identity = x + b, c, t, h, w = x.size() + x = rearrange(x, "b c t h w -> (b t) c h w") + x = self.norm(x) + # compute query, key, value + q, k, v = self.to_qkv(x).reshape(b * t, 1, c * 3, -1).permute(0, 1, 3, 2).contiguous().chunk(3, dim=-1) + + # apply attention + x = F.scaled_dot_product_attention( + q, + k, + v, + ) + x = x.squeeze(1).permute(0, 2, 1).reshape(b * t, c, h, w) + + # output + x = self.proj(x) + x = rearrange(x, "(b t) c h w-> b c t h w", t=t) + return x + identity + + +class Encoder3d(nn.Module): + def __init__( + self, + dim=128, + z_dim=4, + dim_mult=[1, 2, 4, 4], + num_res_blocks=2, + attn_scales=[], + temperal_downsample=[True, True, False], + dropout=0.0, + ): + super().__init__() + self.dim = dim + self.z_dim = z_dim + self.dim_mult = dim_mult + self.num_res_blocks = num_res_blocks + self.attn_scales = attn_scales + self.temperal_downsample = temperal_downsample + + # dimensions + dims = [dim * u for u in [1] + dim_mult] + scale = 1.0 + + # init block + self.conv1 = CausalConv3d(3, dims[0], 3, padding=1) + + # downsample blocks + downsamples = [] + for i, (in_dim, out_dim) in enumerate(zip(dims[:-1], dims[1:])): + # residual (+attention) blocks + for _ in range(num_res_blocks): + downsamples.append(ResidualBlock(in_dim, out_dim, dropout)) + if scale in attn_scales: + downsamples.append(AttentionBlock(out_dim)) + in_dim = out_dim + + # downsample block + if i != len(dim_mult) - 1: + mode = "downsample3d" if temperal_downsample[i] else "downsample2d" + downsamples.append(Resample(out_dim, mode=mode)) + scale /= 2.0 + self.downsamples = nn.Sequential(*downsamples) + + # middle blocks + self.middle = nn.Sequential( + ResidualBlock(out_dim, out_dim, dropout), AttentionBlock(out_dim), ResidualBlock(out_dim, out_dim, dropout) + ) + + # output blocks + self.head = nn.Sequential( + RMS_norm(out_dim, images=False), nn.SiLU(), CausalConv3d(out_dim, z_dim, 3, padding=1) + ) + + def forward(self, x, feat_cache=None, feat_idx=[0]): + if feat_cache is not None: + idx = feat_idx[0] + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None: + # cache last frame of last two chunk + cache_x = torch.cat([feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x], dim=2) + x = self.conv1(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + else: + x = self.conv1(x) + + # downsamples + for layer in self.downsamples: + if feat_cache is not None: + x = layer(x, feat_cache, feat_idx) + else: + x = layer(x) + + # middle + for layer in self.middle: + if isinstance(layer, ResidualBlock) and feat_cache is not None: + x = layer(x, feat_cache, feat_idx) + else: + x = layer(x) + + # head + for layer in self.head: + if isinstance(layer, CausalConv3d) and feat_cache is not None: + idx = feat_idx[0] + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None: + # cache last frame of last two chunk + cache_x = torch.cat( + [feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x], dim=2 + ) + x = layer(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + else: + x = layer(x) + return x + + +class Decoder3d(nn.Module): + def __init__( + self, + dim=128, + z_dim=4, + dim_mult=[1, 2, 4, 4], + num_res_blocks=2, + attn_scales=[], + temperal_upsample=[False, True, True], + dropout=0.0, + ): + super().__init__() + self.dim = dim + self.z_dim = z_dim + self.dim_mult = dim_mult + self.num_res_blocks = num_res_blocks + self.attn_scales = attn_scales + self.temperal_upsample = temperal_upsample + + # dimensions + dims = [dim * u for u in [dim_mult[-1]] + dim_mult[::-1]] + scale = 1.0 / 2 ** (len(dim_mult) - 2) + + # init block + self.conv1 = CausalConv3d(z_dim, dims[0], 3, padding=1) + + # middle blocks + self.middle = nn.Sequential( + ResidualBlock(dims[0], dims[0], dropout), AttentionBlock(dims[0]), ResidualBlock(dims[0], dims[0], dropout) + ) + + # upsample blocks + upsamples = [] + for i, (in_dim, out_dim) in enumerate(zip(dims[:-1], dims[1:])): + # residual (+attention) blocks + if i == 1 or i == 2 or i == 3: + in_dim = in_dim // 2 + for _ in range(num_res_blocks + 1): + upsamples.append(ResidualBlock(in_dim, out_dim, dropout)) + if scale in attn_scales: + upsamples.append(AttentionBlock(out_dim)) + in_dim = out_dim + + # upsample block + if i != len(dim_mult) - 1: + mode = "upsample3d" if temperal_upsample[i] else "upsample2d" + upsamples.append(Resample(out_dim, mode=mode)) + scale *= 2.0 + self.upsamples = nn.Sequential(*upsamples) + + # output blocks + self.head = nn.Sequential(RMS_norm(out_dim, images=False), nn.SiLU(), CausalConv3d(out_dim, 3, 3, padding=1)) + + def forward(self, x, feat_cache=None, feat_idx=[0]): + # conv1 + if feat_cache is not None: + idx = feat_idx[0] + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None: + # cache last frame of last two chunk + cache_x = torch.cat([feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x], dim=2) + x = self.conv1(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + else: + x = self.conv1(x) + + # middle + for layer in self.middle: + if isinstance(layer, ResidualBlock) and feat_cache is not None: + x = layer(x, feat_cache, feat_idx) + else: + x = layer(x) + + # upsamples + for layer in self.upsamples: + if feat_cache is not None: + x = layer(x, feat_cache, feat_idx) + else: + x = layer(x) + + # head + for layer in self.head: + if isinstance(layer, CausalConv3d) and feat_cache is not None: + idx = feat_idx[0] + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None: + # cache last frame of last two chunk + cache_x = torch.cat( + [feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), cache_x], dim=2 + ) + x = layer(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + else: + x = layer(x) + return x + + +def count_conv3d(model): + count = 0 + for m in model.modules(): + if isinstance(m, CausalConv3d): + count += 1 + return count + + +class WanVAE_(nn.Module): + def __init__( + self, + dim=128, + z_dim=4, + dim_mult=[1, 2, 4, 4], + num_res_blocks=2, + attn_scales=[], + temperal_downsample=[True, True, False], + dropout=0.0, + temporal_window=4, + ): + super().__init__() + self.dim = dim + self.z_dim = z_dim + self.dim_mult = dim_mult + self.num_res_blocks = num_res_blocks + self.attn_scales = attn_scales + self.temperal_downsample = temperal_downsample + self.temperal_upsample = temperal_downsample[::-1] + self.temporal_window = temporal_window + # modules + self.encoder = Encoder3d( + dim, z_dim * 2, dim_mult, num_res_blocks, attn_scales, self.temperal_downsample, dropout + ) + self.conv1 = CausalConv3d(z_dim * 2, z_dim * 2, 1) + self.conv2 = CausalConv3d(z_dim, z_dim, 1) + self.decoder = Decoder3d(dim, z_dim, dim_mult, num_res_blocks, attn_scales, self.temperal_upsample, dropout) + + def forward(self, x): + mu, log_var = self.encode(x) + z = self.reparameterize(mu, log_var) + x_recon = self.decode(z) + return x_recon, mu, log_var + + def encode(self, x, scale, clear_encoder_cache=True): + if clear_encoder_cache: + self.clear_cache() + # cache + t = x.shape[2] + iter_ = 1 + (t - 1) // self.temporal_window + # 对encode输入的x，按时间拆分为1、self.temporal_stride、self.temporal_stride、self.temporal_window.... + for i in range(iter_): + self._enc_conv_idx = [0] + if i == 0: + out = self._i0_encode(x) + else: + out_ = self.encoder( + x[:, :, 1 + self.temporal_window * (i - 1) : 1 + self.temporal_window * i, :, :], + feat_cache=self._enc_feat_map, + feat_idx=self._enc_conv_idx, + ) + out = torch.cat([out, out_], 2) + if (t - 1) % self.temporal_window: + self._enc_conv_idx = [0] + out_ = self.encoder( + x[:, :, 1 + self.temporal_window * (iter_ - 1) :, :, :], + feat_cache=self._enc_feat_map, + feat_idx=self._enc_conv_idx, + ) + out = torch.cat([out, out_], 2) + mu, log_var = self.conv1(out).chunk(2, dim=1) + if isinstance(scale[0], torch.Tensor): + mu = (mu - scale[0].view(1, self.z_dim, 1, 1, 1)) * scale[1].view(1, self.z_dim, 1, 1, 1) + else: + mu = (mu - scale[0]) * scale[1] + if clear_encoder_cache: + self.clear_cache() + return mu + + @torch.compiler.disable + def _i0_encode(self, x): + """ + If enabled torch.compile uses significantly more memory for this step, so we disable it + """ + out = self.encoder(x[:, :, :1, :, :], feat_cache=self._enc_feat_map, feat_idx=self._enc_conv_idx) + return out + + @torch.compiler.disable + def _i0_decode(self, x): + return self.decoder(x[:, :, 0:1, :, :], feat_cache=self._feat_map, feat_idx=self._conv_idx) + + def decode(self, z, scale, clear_decoder_cache=True): + if clear_decoder_cache: + self.clear_cache() + # z: [b,c,t,h,w] + if isinstance(scale[0], torch.Tensor): + z = z / scale[1].view(1, self.z_dim, 1, 1, 1) + scale[0].view(1, self.z_dim, 1, 1, 1) + else: + z = z / scale[1] + scale[0] + iter_ = z.shape[2] + x = self.conv2(z) + for i in range(iter_): + self._conv_idx = [0] + if i == 0: + out = self._i0_decode(x) + else: + out_ = self.decoder(x[:, :, i : i + 1, :, :], feat_cache=self._feat_map, feat_idx=self._conv_idx) + out = torch.cat([out, out_], 2) + if clear_decoder_cache: + self.clear_cache() + return out + + def reparameterize(self, mu, log_var): + std = torch.exp(0.5 * log_var) + eps = torch.randn_like(std) + return eps * std + mu + + def sample(self, imgs, deterministic=False): + mu, log_var = self.encode(imgs) + if deterministic: + return mu + std = torch.exp(0.5 * log_var.clamp(-30.0, 20.0)) + return mu + std * torch.randn_like(std) + + def clear_cache(self): + self._conv_num = count_conv3d(self.decoder) + self._conv_idx = [0] + self._feat_map = [None] * self._conv_num + # cache encode + self._enc_conv_num = count_conv3d(self.encoder) + self._enc_conv_idx = [0] + self._enc_feat_map = [None] * self._enc_conv_num + + +def _video_vae( + pretrained_path=None, + z_dim=None, + device="cpu", + s3_credential_path: str = "credentials/s3_training.secret", + load_mean_std=False, + mean_std_path: str = "s3://bucket/cosmos_diffusion_v2/pretrain_weights/tokenizer/wan2pt1/Wan2.1_VAE.pth", + **kwargs, +): + """ + Autoencoder3d adapted from Stable Diffusion 1.x, 2.x and XL. + """ + # params + cfg = dict( + dim=96, + z_dim=z_dim, + dim_mult=[1, 2, 4, 4], + num_res_blocks=2, + attn_scales=[], + temperal_downsample=[False, True, True], + dropout=0.0, + ) + cfg.update(**kwargs) + + # init model + with torch.device("meta"): + model = WanVAE_(**cfg) + + if SMOKE or pretrained_path is None: + model.to_empty(device=device) + if load_mean_std: + img_mean, img_std = torch.randn(1, 16, 1, 1, 1, device=device), torch.randn(1, 16, 1, 1, 1, device=device) + video_mean, video_std = ( + torch.randn(1, 16, 32, 1, 1, device=device), + torch.randn(1, 16, 32, 1, 1, device=device), + ) + else: + if get_rank() == 0: + if not INTERNAL: + from cosmos_policy._src.imaginaire.utils.checkpoint_db import get_checkpoint_path + + pretrained_path = get_checkpoint_path(pretrained_path) + if pretrained_path.startswith("s3://"): + backend_key = "wan2pt1_vae" + easy_io.set_s3_backend( + key=backend_key, + backend_args={ + "backend": "s3", + "s3_credential_path": s3_credential_path, + }, + ) + else: + backend_key = None + + ckpt = easy_io.load( + pretrained_path, + backend_key=backend_key, + map_location=device, + ) + if load_mean_std: + img_mean_std = mean_std_path.replace("mean_std.pt", "images_mean_std.pt") + video_mean_std = mean_std_path.replace("mean_std.pt", "video_mean_std.pt") + if not INTERNAL: + from cosmos_policy._src.imaginaire.utils.checkpoint_db import get_checkpoint_path + + img_mean_std = get_checkpoint_path(img_mean_std) + video_mean_std = get_checkpoint_path(video_mean_std) + img_mean, img_std = easy_io.load(img_mean_std, backend_key=backend_key, map_location=device) + video_mean, video_std = easy_io.load(video_mean_std, backend_key=backend_key, map_location=device) + img_mean = img_mean.reshape(1, 16, 1, 1, 1) + img_std = img_std.reshape(1, 16, 1, 1, 1) + video_mean = video_mean.reshape(1, 16, 32, 1, 1) + video_std = video_std.reshape(1, 16, 32, 1, 1) + + # load checkpoint + log.info(f"loading {pretrained_path}") + model.load_state_dict(ckpt, assign=True) + else: + model.to_empty(device=device) + if load_mean_std: + img_mean, img_std = ( + torch.randn(1, 16, 1, 1, 1, device=device), + torch.randn(1, 16, 1, 1, 1, device=device), + ) + video_mean, video_std = ( + torch.randn(1, 16, 32, 1, 1, device=device), + torch.randn(1, 16, 32, 1, 1, device=device), + ) + sync_model_states(model) + + if load_mean_std: + log.info("broadcast mean and std for wan2pt1") + broadcast(img_mean, 0) + broadcast(img_std, 0) + broadcast(video_mean, 0) + broadcast(video_std, 0) + return model, img_mean, img_std, video_mean, video_std + + return ( + model, + torch.zeros(1, 1, 1, 1, 1, device=device), + torch.ones(1, 1, 1, 1, 1, device=device), + torch.zeros(1, 1, 50, 1, 1, device=device), + torch.ones(1, 1, 50, 1, 1, device=device), + ) + + +class WanVAE: + def __init__( + self, + z_dim=16, + vae_pth="s3://bucket/cosmos_diffusion_v2/pretrain_weights/tokenizer/wan2pt1/Wan2.1_VAE.pth", + s3_credential_path: str = "credentials/s3_training.secret", + load_mean_std=False, + mean_std_path: str = "s3://bucket/cosmos_diffusion_v2/pretrain_weights/tokenizer/wan2pt1/mean_std.pt", + dtype=torch.bfloat16, + device="cuda", + is_amp=True, + benchmark: bool = False, + temporal_window: int = 4, + is_parallel: bool = False, + cp_grid_shape: Optional[tuple[int, int]] = None, + ): + self.dtype = dtype + self.device = device + self.benchmark = benchmark + self.temporal_window = temporal_window + self.is_parallel = is_parallel + self.cp_grid_shape = cp_grid_shape + self.context_parallel_enabled = False + self.cp_group_initialized = False + + mean = [ + -0.7571, + -0.7089, + -0.9113, + 0.1075, + -0.1745, + 0.9653, + -0.1517, + 1.5508, + 0.4134, + -0.0715, + 0.5517, + -0.3632, + -0.1922, + -0.9497, + 0.2503, + -0.2921, + ] + std = [ + 2.8184, + 1.4541, + 2.3275, + 2.6558, + 1.2196, + 1.7708, + 2.6052, + 2.0743, + 3.2687, + 2.1526, + 2.8652, + 1.5579, + 1.6382, + 1.1253, + 2.8251, + 1.9160, + ] + self.mean = torch.tensor(mean, dtype=dtype, device=device) + self.std = torch.tensor(std, dtype=dtype, device=device) + self.scale = [self.mean, 1.0 / self.std] + + # init model + self.model, self.img_mean, self.img_std, self.video_mean, self.video_std = _video_vae( + pretrained_path=vae_pth, + z_dim=z_dim, + s3_credential_path=s3_credential_path, + load_mean_std=load_mean_std, + mean_std_path=mean_std_path, + device=device, + temporal_window=temporal_window, + ) + + if is_parallel: + cp_group = None + if parallel_state.is_initialized(): + cp_group = parallel_state.get_context_parallel_group() + if cp_grid_shape is None: + cp_grid_shape = (1, cp_group.size()) + else: + assert False, "is_parallel set, but context parallelism is initialized" + + self._initialize_context_parallel(cp_group, cp_grid_shape) + + self.model = self.model.eval().requires_grad_(False) + self.is_amp = is_amp + if not is_amp: + self.model = self.model.to(dtype=dtype) + self.context = nullcontext() + else: + self.context = torch.amp.autocast("cuda", dtype=dtype) + + def count_param(self): + return sum(p.numel() for p in self.model.parameters()) + + @torch.no_grad() + def encode(self, videos, clear_encoder_cache=True): + """ + videos: A list of videos each with shape [C, T, H, W]. + """ + if self.is_parallel: + if self._is_image_batch(videos): + self._disable_context_parallel() + else: + # Latents are concatenated before attention so we won't need to gather chunks after execution + try: + videos = self._broadcast_split_for_model_parallelsim(videos) + self._enable_context_parallel() + except ValueError as e: + log.warning(str(e)) + self._disable_context_parallel() + if self.benchmark: + torch.cuda.synchronize() + benchmark_times = BenchmarkTimes() + total_time = time.perf_counter() + in_dtype = videos.dtype + with self.context: + if not self.is_amp: + videos = videos.to(self.dtype) + if self.benchmark: + torch.cuda.synchronize() + model_time = time.perf_counter() + latent = self.model.encode(videos, self.scale, clear_encoder_cache) + if self.benchmark: + torch.cuda.synchronize() + benchmark_times.model_invocation = time.perf_counter() - model_time + latent = latent.to(in_dtype) + if self.benchmark: + torch.cuda.synchronize() + benchmark_times.total = time.perf_counter() - total_time + return latent, benchmark_times + return latent + + @torch.no_grad() + def decode(self, zs, clear_decoder_cache=True): + if self.benchmark: + torch.cuda.synchronize() + benchmark_times = BenchmarkTimes() + total_time = time.perf_counter() + if self.is_parallel: + if self._is_image_batch(zs): + self._disable_context_parallel() + else: + # Make sure height and width divisible by CP factors + can_apply_cp = (zs.shape[3] % self.cp_grid_shape[0] == 0) and (zs.shape[4] % self.cp_grid_shape[1] == 0) + if not can_apply_cp: + log.warning( + f"For parallel encoding with grid_shape {self.cp_grid_shape} latent height should be divisible by grid_shape[0], got {zs.shape[3]} / {self.cp_grid_shape[0]} and width should be divisible by grid_shape[1], got {zs.shape[4]} / {self.cp_grid_shape[1]}, falling back to non CP" + ) + self._disable_context_parallel() + else: + self._enable_context_parallel() + in_dtype = zs.dtype + with self.context: + if not self.is_amp: + zs = zs.to(self.dtype) + if self.benchmark: + torch.cuda.synchronize() + model_time = time.perf_counter() + video_recon = self.model.decode(zs, self.scale, clear_decoder_cache) + if self.benchmark: + torch.cuda.synchronize() + benchmark_times.model_invocation = time.perf_counter() - model_time + video_recon = video_recon.to(in_dtype) + if self.is_parallel and self.context_parallel_enabled: + # Decoder splits tensors into CP chunks after attention (it is assumed all ranks in CP group have same data before execution), so we only need to gather at the end + video_recon = self._cat_outputs_cp(video_recon) + if self.benchmark: + torch.cuda.synchronize() + benchmark_times.total = time.perf_counter() - total_time + return video_recon, benchmark_times + return video_recon + + @property + def spatial_compression_factor(self): + return 8 + + @property + def temporal_compression_factor(self): + return 4 + + @property + def _cp_dim(self): + return 3 + + def _broadcast_split_for_model_parallelsim(self, state: torch.Tensor) -> torch.Tensor: + # All ranks from CP group get different data to encode, later when data is split before calling `compute_loss_with_epsilon_and_sigma`, they get data broadcasted from min rank in group + # So we have to broadcast data now + assert len(state.shape) == 5, "State should be of shape BCTHW" + cp_rows, cp_cols = self.cp_grid_shape + can_cp_be_applied_to_shape = ( + state.shape[3] % (cp_rows * self.spatial_compression_factor) == 0 + and state.shape[4] % (cp_cols * self.spatial_compression_factor) == 0 + ) + + if not can_cp_be_applied_to_shape: + raise ValueError( + f"For parallel encoding with grid_shape {self.cp_grid_shape} height should be divisible by compression_factor*grid_shape[0], got {state.shape[3]} / ({self.cp_grid_shape[0]} * {self.spatial_compression_factor}) and width should be divisible by compression_factor*grid_shape[1], got {state.shape[4]} / ({self.cp_grid_shape[1]} * {self.spatial_compression_factor}), falling back to non CP" + ) + + # distributed.broadcast doesn't work with torch.export so we use distributed.all_gather + state = state.contiguous() + state_list = [torch.zeros_like(state) for _ in range(cp_rows * cp_cols)] + distributed.all_gather(state_list, state, group=self.cp_group) + state = state_list[0] + # state = context_parallel.broadcast(state.contiguous(), self.cp_group) + + chunk_h = state.shape[3] // cp_rows + chunk_w = state.shape[4] // cp_cols + group_rank = distributed.get_rank(group=self.cp_group) + + row_id = group_rank // cp_cols + col_id = group_rank % cp_cols + + return state[:, :, :, row_id * chunk_h : (row_id + 1) * chunk_h, col_id * chunk_w : (col_id + 1) * chunk_w] + + def _cat_outputs_cp(self, local_video_recon: torch.Tensor): + video_recon_chunks = [torch.zeros_like(local_video_recon) for _ in range(self.cp_group_size)] + distributed.all_gather(video_recon_chunks, local_video_recon, group=self.cp_group) + + # Concatenate chunks vertically then horizontaly + video_recon = torch.cat( + [torch.cat(video_recon_chunks[c :: self.cp_grid_shape[1]], dim=3) for c in range(self.cp_grid_shape[1])], + dim=4, + ) + + return video_recon + + def _enable_context_parallel(self): + self.context_parallel_enabled = True + for _, plugin_list in self.plugins.items(): + for _, plugin in plugin_list.items(): + plugin.set_enable(True) + + def _disable_context_parallel(self): + self.context_parallel_enabled = False + for _, plugin_list in self.plugins.items(): + for _, plugin in plugin_list.items(): + plugin.set_enable(False) + + def _is_image_batch(self, x: torch.Tensor) -> bool: + assert len(x.shape) == 5, "Expected tensor's shape to be BCTHW" + return x.shape[2] == 1 + + def _initialize_context_parallel(self, cp_group: distributed.ProcessGroup, cp_grid_shape) -> None: + assert self.cp_group_initialized is False + self.is_parallel = True + self.cp_grid_shape = cp_grid_shape + self.context_parallel_enabled = False + self.cp_group = cp_group + + self.cp_group_size = len(distributed.get_process_group_ranks(self.cp_group)) + self.plugins: dict = plugin_mount(self.model, self.cp_group, cp_grid_shape) + self._enable_context_parallel() + log.info(f"Enabled CP with grid_shape: {cp_grid_shape} for Wan2.1 tokenizer") + + +class Wan2pt1VAEInterface(VideoTokenizerInterface): + def __init__(self, chunk_duration: int = 81, load_mean_std=False, **kwargs): + self.keep_decoder_cache = kwargs.get("keep_decoder_cache", False) + self.keep_encoder_cache = kwargs.get("keep_encoder_cache", False) + self.model = WanVAE( + dtype=torch.bfloat16, + is_amp=False, + load_mean_std=load_mean_std, + vae_pth=kwargs.get( + "vae_pth", + "s3://bucket/cosmos_diffusion_v2/pretrain_weights/tokenizer/wan2pt1/Wan2.1_VAE.pth", + ), + s3_credential_path=kwargs.get("s3_credential_path", "credentials/s3_training.secret"), + temporal_window=kwargs.get("temporal_window", 4), + is_parallel=kwargs.get("is_parallel", False), + cp_grid_shape=kwargs.get("cp_grid_shape", None), + ) + del kwargs + self.chunk_duration = chunk_duration + self.cp_initialized = False + + def initialize_context_parallel(self, cp_group: distributed.ProcessGroup, cp_grid_shape: tuple[int, int]) -> None: + assert self.cp_initialized is False + self.cp_initialized = True + self.model._initialize_context_parallel(cp_group, cp_grid_shape) + + @property + def dtype(self): + return self.model.dtype + + def reset_dtype(self): + pass + + def clear_cache(self): + """Clear the feature cache for both encoder and decoder.""" + self.model.model.clear_cache() + + def encode(self, state: torch.Tensor) -> torch.Tensor: + latents = self.model.encode(state, clear_encoder_cache=not self.keep_encoder_cache) + num_frames = latents.shape[2] + if num_frames == 1: + return (latents - self.model.img_mean.type_as(latents)) / self.model.img_std.type_as(latents) + else: + return (latents - self.model.video_mean[:, :, :num_frames].type_as(latents)) / self.model.video_std[ + :, :, :num_frames + ].type_as(latents) + + def decode(self, latent: torch.Tensor) -> torch.Tensor: + num_frames = latent.shape[2] + if num_frames == 1: + recon = self.model.decode( + ((latent * self.model.img_std.type_as(latent)) + self.model.img_mean.type_as(latent)).contiguous() + ) + else: + recon = self.model.decode( + ( + (latent * self.model.video_std[:, :, :num_frames].type_as(latent)) + + self.model.video_mean[:, :, :num_frames].type_as(latent) + ).contiguous() + ) + + if isinstance(recon, list): + # torch.export makes batch_size=1 to be returned as list so we take first element and create batch dimension back + assert len(recon) == 1, "Assuming batch_size=1 was used" + recon = recon[0].unsqueeze(0) + return recon + + def get_latent_num_frames(self, num_pixel_frames: int) -> int: + return 1 + (num_pixel_frames - 1) // 4 + + def get_pixel_num_frames(self, num_latent_frames: int) -> int: + return (num_latent_frames - 1) * 4 + 1 + + @property + def spatial_compression_factor(self): + return 8 + + @property + def temporal_compression_factor(self): + return 4 + + @property + def pixel_chunk_duration(self): + return self.chunk_duration + + @property + def latent_chunk_duration(self): + return self.get_latent_num_frames(self.chunk_duration) + + @property + def latent_ch(self): + return 16 + + @property + def spatial_resolution(self): + return 512 + + @property + def name(self): + return "wan2pt1_tokenizer" diff --git a/REGEN-main/cosmos_policy/_src/predict2/tokenizers/wan2pt1_2d_plugins.py b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/wan2pt1_2d_plugins.py new file mode 100644 index 0000000000000000000000000000000000000000..55cc7732fe5e161ca404eecc1372051ecb8e3f43 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/wan2pt1_2d_plugins.py @@ -0,0 +1,930 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Original code author is Yan Wang, yanwa@nvidia.com +# +# The purpose of those plugins is to enable CP (at the moment splitting by video height) for Wan2.1 tokenizer inference, +# they work by wrapping original Conv/Attention modules' forward calls. +# In case of convolutional layers ranks receive/send required data from/to ranks that are handling adjacent parts of the video +# to properly compute convolution, pad their input with that data, run original module, and truncate output if needed. +# Attention is not parallelized, so all ranks gather (if needed) all parts of the video run attention and split result (if needed) back into chunks +# +# Only function `plugin_mount` should be used outside this file + +from itertools import chain + +import torch +import torch.distributed as distributed + + +def _create_adj_groups( + grid_shape: tuple[int, int], cp_group: distributed.ProcessGroup +) -> tuple[list[distributed.ProcessGroup], list[distributed.ProcessGroup]]: + grid_rows, grid_cols = grid_shape + + all_rank_groups = [None for _ in range(distributed.get_world_size())] + group_ranks = distributed.get_process_group_ranks(cp_group) if cp_group is not None else [] + distributed.all_gather_object(all_rank_groups, group_ranks) + + all_groups = list(set(tuple(rank_group) for rank_group in all_rank_groups)) + global_rank = distributed.get_rank() + in_row_groups = None + in_col_groups = None + + for cp_group_ranks in all_groups: + if len(cp_group_ranks) == 0: + continue + tmp_in_row_groups = [] + tmp_in_col_groups = [] + scaling_factor = min(cp_group_ranks) + for row in range(grid_rows): + in_row_adj_ranks_list = [ + (cp_group_ranks[row * grid_cols + col], cp_group_ranks[row * grid_cols + col + 1]) + for col in range(grid_cols - 1) + ] + + adj_groups = [distributed.new_group(in_row_adj_ranks) for in_row_adj_ranks in in_row_adj_ranks_list] + # print(f'{global_rank=}\t{adj_groups=}') + # if all(adj_group == distributed.GroupMember.NON_GROUP_MEMBER for adj_group in adj_groups): + # continue + tmp_in_row_groups.append(adj_groups) + + for col in range(grid_cols): + in_col_adj_ranks_list = [ + (row * grid_cols + col + scaling_factor, (row + 1) * grid_cols + col + scaling_factor) + for row in range(grid_rows - 1) + ] + + adj_groups = [distributed.new_group(in_col_adj_ranks) for in_col_adj_ranks in in_col_adj_ranks_list] + + # if all(adj_group == distributed.GroupMember.NON_GROUP_MEMBER for adj_group in adj_groups): + # continue + tmp_in_col_groups.append(adj_groups) + + if global_rank in cp_group_ranks: + in_row_groups = tmp_in_row_groups + in_col_groups = tmp_in_col_groups + return in_row_groups, in_col_groups + + +class _ModulePlugin: + def __init__(self, module, module_id, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups): + self.module = module + self.module_id = module_id + self.enable = True + self.implement_forward() + self.plugin_config = plugin_config + + self.in_row_adj_groups = in_row_adj_groups + self.in_col_adj_groups = in_col_adj_groups + self.cp_group = cp_group + self.group_rank = distributed.get_rank(group=cp_group) + self.group_rank_to_global_rank = distributed.get_process_group_ranks(cp_group) + self.cp_group_size = len(self.group_rank_to_global_rank) + + self.grid_shape = grid_shape + self.grid_rows, self.grid_cols = grid_shape + + self.row_id = self.group_rank // self.grid_cols + self.col_id = self.group_rank % self.grid_cols + + # Neighbor rank calculation + self.left_neighbor = self.group_rank - 1 if self.col_id > 0 else None + self.right_neighbor = self.group_rank + 1 if self.col_id < self.grid_cols - 1 else None + self.top_neighbor = self.group_rank - self.grid_cols if self.row_id > 0 else None + self.bottom_neighbor = self.group_rank + self.grid_cols if self.row_id < self.grid_rows - 1 else None + + self.my_row_groups = self.in_row_adj_groups[self.row_id] + self.my_col_groups = self.in_col_adj_groups[self.col_id] + + def implement_forward(self): + module = self.module + if not hasattr(module, "old_forward"): + module.old_forward = module.forward + + self.new_forward = self.get_new_forward() + + def forward(*args, **kwargs): + self.update_config() + return self.new_forward(*args, **kwargs) if self.enable else module.old_forward(*args, **kwargs) + + module.forward = forward + + def set_enable(self, enable=True): + self.enable = enable + + def get_new_forward(self): + raise NotImplementedError + + def update_config(self, config: dict | None = None): + if config is None: + config = self.plugin_config.get(self.module_id[0], {}) + + for key, value in config.items(): + setattr(self, key, value) + + +class _Conv3DSafeNewPlugin(_ModulePlugin): + def __init__( + self, + module, + module_id, + plugin_config=None, + cp_group=None, + grid_shape=None, + in_row_adj_groups=None, + in_col_adj_groups=None, + ): + super().__init__(module, module_id, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups) + + self.kernel_size = getattr(module, "kernel_size", (1, 1, 1)) + + if isinstance(self.kernel_size, int): + self.kernel_size = (self.kernel_size, self.kernel_size, self.kernel_size) + + kernel_height = self.kernel_size[1] + d_height = kernel_height - 1 + self.padding_left_height = d_height // 2 + self.padding_right_height = d_height - self.padding_left_height + self.height_padding_flag = self.padding_left_height if d_height > 0 else 0 + + kernel_width = self.kernel_size[2] + d_width = kernel_width - 1 + self.padding_left_width = d_width // 2 + self.padding_right_width = d_width - self.padding_left_width + self.width_padding_flag = self.padding_left_width if d_width > 0 else 0 + + def pad_context_2d(self, h): + """2D context padding: simultaneously perform padding on both height and width dimensions""" + + if self.width_padding_flag == 0 and self.height_padding_flag == 0: + return h + + # First step: perform padding on width dimension (dim=4) + if self.width_padding_flag > 0: + h = self._pad_width_dimension(h) + + # Second step: perform padding on height dimension (dim=3) + if self.height_padding_flag > 0: + h = self._pad_height_dimension(h) + + return h + + def _pad_width_dimension(self, h): + """Perform padding on width dimension (dim=4)""" + # Only pad in necessary directions, no padding at boundaries + contexts_to_concat = [] + + # Left padding: only needed for non-leftmost columns + if self.left_neighbor is not None: + share_to_left = h[:, :, :, :, : self.padding_left_width].contiguous() + if self.col_id % 2: + # Odd column, handle left first + padding_list = [torch.zeros_like(share_to_left) for _ in range(2)] + distributed.all_gather(padding_list, share_to_left, group=self.my_row_groups[self.col_id - 1]) + left_context = padding_list[0].to(h.device, non_blocking=True) + else: + # Even column, handle left later + padding_list = [torch.zeros_like(share_to_left) for _ in range(2)] + distributed.all_gather(padding_list, share_to_left, group=self.my_row_groups[self.col_id - 1]) + left_context = padding_list[0].to(h.device, non_blocking=True) + contexts_to_concat.append(left_context) + + # Add original data + contexts_to_concat.append(h) + + # Right padding: only needed for non-rightmost columns + if self.right_neighbor is not None: + share_to_right = h[:, :, :, :, -self.padding_right_width :].contiguous() + if self.col_id % 2: + # Odd column, handle right later + padding_list = [torch.zeros_like(share_to_right) for _ in range(2)] + distributed.all_gather(padding_list, share_to_right, group=self.my_row_groups[self.col_id]) + right_context = padding_list[1].to(h.device, non_blocking=True) + else: + # Even column, handle right first + padding_list = [torch.zeros_like(share_to_right) for _ in range(2)] + distributed.all_gather(padding_list, share_to_right, group=self.my_row_groups[self.col_id]) + right_context = padding_list[1].to(h.device, non_blocking=True) + contexts_to_concat.append(right_context) + + h_with_width_context = torch.cat(contexts_to_concat, dim=4) + return h_with_width_context + + def _pad_height_dimension(self, h): + """Perform padding on height dimension (dim=3)""" + # Only pad in necessary directions, no padding at boundaries + contexts_to_concat = [] + + # Top padding: only needed for non-topmost rows + if self.top_neighbor is not None: + share_to_top = h[:, :, :, : self.padding_left_height].contiguous() + if self.row_id % 2: + # Odd row, handle top first + padding_list = [torch.zeros_like(share_to_top) for _ in range(2)] + distributed.all_gather(padding_list, share_to_top, group=self.my_col_groups[self.row_id - 1]) + top_context = padding_list[0].to(h.device, non_blocking=True) + else: + # Even row, handle top later + padding_list = [torch.zeros_like(share_to_top) for _ in range(2)] + distributed.all_gather(padding_list, share_to_top, group=self.my_col_groups[self.row_id - 1]) + top_context = padding_list[0].to(h.device, non_blocking=True) + contexts_to_concat.append(top_context) + + # Add original data + contexts_to_concat.append(h) + + # Bottom padding: only needed for non-bottommost rows + if self.bottom_neighbor is not None: + share_to_bottom = h[:, :, :, -self.padding_right_height :].contiguous() + if self.row_id % 2: + # Odd row, handle bottom later + padding_list = [torch.zeros_like(share_to_bottom) for _ in range(2)] + distributed.all_gather(padding_list, share_to_bottom, group=self.my_col_groups[self.row_id]) + bottom_context = padding_list[1].to(h.device, non_blocking=True) + else: + # Even row, handle bottom first + padding_list = [torch.zeros_like(share_to_bottom) for _ in range(2)] + distributed.all_gather(padding_list, share_to_bottom, group=self.my_col_groups[self.row_id]) + bottom_context = padding_list[1].to(h.device, non_blocking=True) + contexts_to_concat.append(bottom_context) + + h_with_height_context = torch.cat(contexts_to_concat, dim=3) + return h_with_height_context + + def get_new_forward(self): + module = self.module + + def new_forward(hidden_states, cache_x=None): + # If no padding is needed, directly use original forward + if self.width_padding_flag == 0 and self.height_padding_flag == 0: + return module.old_forward(hidden_states, cache_x) + + # Perform 2D context padding + hidden_states = self.pad_context_2d(hidden_states) + if cache_x is not None: + cache_x = self.pad_context_2d(cache_x) + + # Execute convolution operation + result = module.old_forward(hidden_states, cache_x) + + # Crop results, remove padding (only crop parts that actually had padding added) + # First crop height dimension + if self.height_padding_flag > 0: + # Calculate actual cropping range + start_h = self.padding_left_height if self.top_neighbor is not None else 0 + end_h = ( + -self.padding_right_height + if (self.padding_right_height > 0 and self.bottom_neighbor is not None) + else None + ) + if start_h > 0 or end_h is not None: + result = result[:, :, :, start_h:end_h] + + # Then crop width dimension + if self.width_padding_flag > 0: + # Calculate actual cropping range + start_w = self.padding_left_width if self.left_neighbor is not None else 0 + end_w = ( + -self.padding_right_width + if (self.padding_right_width > 0 and self.right_neighbor is not None) + else None + ) + if start_w > 0 or end_w is not None: + result = result[:, :, :, :, start_w:end_w] + return result + + return new_forward + + +class _Conv2DSafeNewPlugin(_ModulePlugin): + def __init__( + self, + module, + module_id, + plugin_config=None, + cp_group=None, + grid_shape=None, + in_row_adj_groups=None, + in_col_adj_groups=None, + ): + super().__init__(module, module_id, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups) + + self.kernel_size = getattr(module, "kernel_size", (1, 1)) + self.stride = getattr(module, "stride", (1, 1)) + + if isinstance(self.kernel_size, int): + self.kernel_size = (self.kernel_size, self.kernel_size) + if isinstance(self.stride, int): + self.stride = (self.stride, self.stride) + + kernel_height = self.kernel_size[0] + d_height = kernel_height - 1 + self.padding_left_height = d_height // 2 + self.padding_right_height = d_height - self.padding_left_height + self.height_padding_flag = self.padding_left_height if d_height > 0 else 0 + + kernel_width = self.kernel_size[1] + d_width = kernel_width - 1 + self.padding_left_width = d_width // 2 + self.padding_right_width = d_width - self.padding_left_width + self.width_padding_flag = self.padding_left_width if d_width > 0 else 0 + + def pad_context_2d(self, h): + if self.width_padding_flag == 0 and self.height_padding_flag == 0: + return h + + if self.width_padding_flag > 0: + h = self._pad_width_dimension(h) + if self.height_padding_flag > 0: + h = self._pad_height_dimension(h) + + return h + + def _pad_width_dimension(self, h): + """Perform padding on width dimension (dim=3)""" + # Only pad in necessary directions, no padding at boundaries + contexts_to_concat = [] + + # Left padding: only needed for non-leftmost columns + if self.left_neighbor is not None: + share_to_left = h[:, :, :, : self.padding_left_width].contiguous() + if self.col_id % 2: + # Odd column, handle left first + padding_list = [torch.zeros_like(share_to_left) for _ in range(2)] + distributed.all_gather(padding_list, share_to_left, group=self.my_row_groups[self.col_id - 1]) + left_context = padding_list[0].to(h.device, non_blocking=True) + else: + # Even column, handle left later + padding_list = [torch.zeros_like(share_to_left) for _ in range(2)] + distributed.all_gather(padding_list, share_to_left, group=self.my_row_groups[self.col_id - 1]) + left_context = padding_list[0].to(h.device, non_blocking=True) + contexts_to_concat.append(left_context) + + # Add original data + contexts_to_concat.append(h) + + # Right padding: only needed for non-rightmost columns + if self.right_neighbor is not None: + share_to_right = h[:, :, :, -self.padding_right_width :].contiguous() + if self.col_id % 2: + # Odd column, handle right later + padding_list = [torch.zeros_like(share_to_right) for _ in range(2)] + distributed.all_gather(padding_list, share_to_right, group=self.my_row_groups[self.col_id]) + right_context = padding_list[1].to(h.device, non_blocking=True) + else: + # Even column, handle right first + padding_list = [torch.zeros_like(share_to_right) for _ in range(2)] + distributed.all_gather(padding_list, share_to_right, group=self.my_row_groups[self.col_id]) + right_context = padding_list[1].to(h.device, non_blocking=True) + contexts_to_concat.append(right_context) + + h_with_width_context = torch.cat(contexts_to_concat, dim=3) + return h_with_width_context + + def _pad_height_dimension(self, h): + """Perform padding on height dimension (dim=2)""" + # Only pad in necessary directions, no padding at boundaries + contexts_to_concat = [] + + # Top padding: only needed for non-topmost rows + if self.top_neighbor is not None: + share_to_top = h[:, :, : self.padding_left_height].contiguous() + if self.row_id % 2: + # Odd row, handle top first + padding_list = [torch.zeros_like(share_to_top) for _ in range(2)] + distributed.all_gather(padding_list, share_to_top, group=self.my_col_groups[self.row_id - 1]) + top_context = padding_list[0].to(h.device, non_blocking=True) + else: + # Even row, handle top later + padding_list = [torch.zeros_like(share_to_top) for _ in range(2)] + distributed.all_gather(padding_list, share_to_top, group=self.my_col_groups[self.row_id - 1]) + top_context = padding_list[0].to(h.device, non_blocking=True) + contexts_to_concat.append(top_context) + + # Add original data + contexts_to_concat.append(h) + + # Bottom padding: only needed for non-bottommost rows + if self.bottom_neighbor is not None: + share_to_bottom = h[:, :, -self.padding_right_height :].contiguous() + if self.row_id % 2: + padding_list = [torch.zeros_like(share_to_bottom) for _ in range(2)] + distributed.all_gather(padding_list, share_to_bottom, group=self.my_col_groups[self.row_id]) + bottom_context = padding_list[1].to(h.device, non_blocking=True) + else: + padding_list = [torch.zeros_like(share_to_bottom) for _ in range(2)] + distributed.all_gather(padding_list, share_to_bottom, group=self.my_col_groups[self.row_id]) + bottom_context = padding_list[1].to(h.device, non_blocking=True) + contexts_to_concat.append(bottom_context) + + h_with_height_context = torch.cat(contexts_to_concat, dim=2) + return h_with_height_context + + def get_new_forward(self): + module = self.module + + def new_forward(hidden_states: torch.Tensor) -> torch.Tensor: + hidden_states = self.pad_context_2d(hidden_states) + + result = module.old_forward(hidden_states) + + if self.height_padding_flag: + result = result[ + :, + :, + (self.padding_left_height if self.top_neighbor is not None else None) : ( + (-self.padding_right_height if self.padding_right_height > 0 else None) + if self.bottom_neighbor is not None + else None + ), + ] + + if self.width_padding_flag: + result = result[ + :, + :, + :, + (self.padding_left_width if self.left_neighbor is not None else None) : ( + (-self.padding_right_width if self.padding_right_width > 0 else None) + if self.right_neighbor is not None + else None + ), + ] + + return result + + return new_forward + + +class _Conv2DSafeNewPluginStride2(_ModulePlugin): + def __init__( + self, + module, + module_id, + plugin_config=None, + cp_group=None, + grid_shape=None, + in_row_adj_groups=None, + in_col_adj_groups=None, + ): + super().__init__(module, module_id, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups) + + self.kernel_size = getattr(module, "kernel_size", (1, 1)) + self.stride = getattr(module, "stride", (1, 1)) + + if isinstance(self.kernel_size, int): + self.kernel_size = (self.kernel_size, self.kernel_size) + if isinstance(self.stride, int): + self.stride = (self.stride, self.stride) + + kernel_height, kernel_width = self.kernel_size + self.padding_height = (kernel_height - 1) // 2 if kernel_height > 1 else 0 + self.padding_width = (kernel_width - 1) // 2 if kernel_width > 1 else 0 + + self.diagonal_sender = None + self.diagonal_receiver = None + + self.right_sender = self.group_rank + 1 if self.col_id < self.grid_cols - 1 else None + self.bottom_sender = self.group_rank + self.grid_cols if self.row_id < self.grid_rows - 1 else None + self.left_receiver = self.group_rank - 1 if self.col_id > 0 else None + self.top_receiver = self.group_rank - self.grid_cols if self.row_id > 0 else None + + if self.row_id < self.grid_rows - 1 and self.col_id < self.grid_cols - 1: + self.diagonal_sender = self.group_rank + self.grid_cols + 1 + + if self.row_id > 0 and self.col_id > 0: + self.diagonal_receiver = self.group_rank - self.grid_cols - 1 + + def pad_context_2d(self, h): + if h is None or self.cp_group_size == 1: + return h + + if self.padding_height == 0 and self.padding_width == 0: + return h + + return self._pad_with_unidirectional_transfer(h) + + def _pad_with_unidirectional_transfer(self, h): + if self.padding_width > 0: + h = self._pad_width_with_concat(h) + + if self.padding_height > 0: + h = self._pad_height_with_concat(h) + + if self.padding_width > 0 and self.padding_height > 0: + h = self._handle_diagonal_transfer(h) + + return h + + def _pad_width_with_concat(self, h): + contexts_to_concat = [] + contexts_to_concat.append(h) + if self.right_sender is not None: + tmp = torch.zeros(h.shape[0], h.shape[1], h.shape[2], self.padding_width, dtype=h.dtype, device=h.device) + padding_list = [torch.zeros_like(tmp) for _ in range(2)] + distributed.all_gather(padding_list, tmp, group=self.my_row_groups[self.col_id]) + contexts_to_concat.append(padding_list[1].to(h.device, non_blocking=True)) + + if self.left_receiver is not None: + left_boundary = h[:, :, :, : self.padding_width].contiguous() + padding_list = [torch.zeros_like(left_boundary) for _ in range(2)] + distributed.all_gather(padding_list, left_boundary, group=self.my_row_groups[self.col_id - 1]) + + return torch.cat(contexts_to_concat, dim=3) + + def _pad_height_with_concat(self, h): + contexts_to_concat = [] + contexts_to_concat.append(h) + + if self.bottom_sender is not None: + tmp = torch.zeros(h.shape[0], h.shape[1], self.padding_height, h.shape[3], dtype=h.dtype, device=h.device) + padding_list = [torch.zeros_like(tmp) for _ in range(2)] + distributed.all_gather(padding_list, tmp, group=self.my_col_groups[self.row_id]) + contexts_to_concat.append(padding_list[1]) + + if self.top_receiver is not None: + top_boundary = h[:, :, : self.padding_height, :].contiguous() + padding_list = [torch.zeros_like(top_boundary) for _ in range(2)] + distributed.all_gather(padding_list, top_boundary, group=self.my_col_groups[self.row_id - 1]) + + return torch.cat(contexts_to_concat, dim=2) + + def _handle_diagonal_transfer(self, h): + if self.diagonal_sender is not None and self.right_sender is not None and self.bottom_sender is not None: + # Receive data from bottom-right diagonal neighbor + diagonal_data = torch.zeros( + h.shape[0], h.shape[1], self.padding_height, self.padding_width, dtype=h.dtype, device=h.device + ) + distributed.recv(diagonal_data, src=self.diagonal_sender) + + # Fill diagonal data to bottom-right corner + # Since width and height padding have been performed earlier, h's size has increased + # Need to place diagonal data at the bottom-right position + original_h = h.shape[2] - self.padding_height + original_w = h.shape[3] - self.padding_width + h[:, :, original_h:, original_w:] = diagonal_data + + # Send data to top-left diagonal neighbor + # Only devices that are not in the first row and not in the first column need to send diagonal data + if self.diagonal_receiver is not None: + # Send own top-left corner data to top-left diagonal neighbor + # Take the top-left part of original data (data before padding) + corner_data = h[:, :, : self.padding_height, : self.padding_width].contiguous() + distributed.send(corner_data, dst=self.diagonal_receiver) + + return h + + def get_new_forward(self): + module = self.module + + def new_forward(hidden_states): + # Doing CP execution here causes torch.compile + CUDA graphs to deadlock, not sure why, so we just gather all data and, run convolution like we are not using CP, and return appropriate chunk + # hidden_states = self.pad_context_2d(hidden_states) + # return module.old_forward(hidden_states) + + is_last_row = self.row_id == self.grid_rows - 1 + is_last_col = self.col_id == self.grid_cols - 1 + + # Remove ZeroPad from chunks + if is_last_col: + hidden_states = hidden_states[:, :, :, :-1] + if is_last_row: + hidden_states = hidden_states[:, :, :-1] + + gathered_tensors = [torch.zeros_like(hidden_states) for _ in range(self.cp_group_size)] + distributed.all_gather(gathered_tensors, hidden_states.contiguous(), group=self.cp_group) + + combined_tensor = torch.cat( + [torch.cat(gathered_tensors[c :: self.grid_cols], dim=2) for c in range(self.grid_cols)], dim=3 + ) + + # Reapply ZeroPad to whole video + combined_tensor = torch.nn.ZeroPad2d((0, 1, 0, 1)).eval()(combined_tensor) + + forward_output = module.old_forward(combined_tensor) + + chunk_h = forward_output.shape[2] // self.grid_rows + chunk_w = forward_output.shape[3] // self.grid_cols + + local_output = forward_output[ + :, + :, + self.row_id * chunk_h : (self.row_id + 1) * chunk_h, + self.col_id * chunk_w : (self.col_id + 1) * chunk_w, + ].contiguous() + + return local_output + + return new_forward + + +class _WanAttentionPlugin(_ModulePlugin): + def __init__( + self, + module, + module_id, + plugin_config=None, + cp_group=None, + grid_shape=None, + in_row_adj_groups=None, + in_col_adj_groups=None, + all_gather_before_attention=False, + cp_split_after_attention=True, + ): + self.all_gather_before_attention = all_gather_before_attention + self.cp_split_after_attention = cp_split_after_attention + + super().__init__(module, module_id, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups) + + def get_new_forward(self): + module = self.module + + def new_forward(hidden_states: torch.Tensor) -> torch.Tensor: + if self.all_gather_before_attention: + gathered_tensors = [torch.zeros_like(hidden_states) for _ in range(self.cp_group_size)] + distributed.all_gather(gathered_tensors, hidden_states, group=self.cp_group) + + combined_tensor = torch.cat( + [torch.cat(gathered_tensors[c :: self.grid_cols], dim=3) for c in range(self.grid_cols)], dim=4 + ) + else: + combined_tensor = hidden_states + + forward_output = module.old_forward(combined_tensor) + + if self.cp_split_after_attention: + chunk_h = forward_output.shape[3] // self.grid_rows + chunk_w = forward_output.shape[4] // self.grid_cols + + local_output = forward_output[ + :, + :, + :, + self.row_id * chunk_h : (self.row_id + 1) * chunk_h, + self.col_id * chunk_w : (self.col_id + 1) * chunk_w, + ].contiguous() + else: + local_output = forward_output + + return local_output + + return new_forward + + +class _ResamplePlugin(_ModulePlugin): + def __init__( + self, + module, + module_id, + plugin_config=None, + cp_group=None, + grid_shape=None, + in_row_adj_groups=None, + in_col_adj_groups=None, + ): + super().__init__(module, module_id, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups) + + def get_new_forward(self): + module = self.module + + def new_forward(*args, **kwargs) -> torch.Tensor: + return module.old_forward(*args, **kwargs) + + return new_forward + + def set_enable(self, enable=True): + self.enable = enable + + if not (self.module.mode in ["downsample2d", "downsample3d"] and self.group_rank < self.cp_group_size - 1): + return + + if self.enable is True: + # Check if at boundaries + is_last_row = self.row_id == self.grid_rows - 1 + is_last_col = self.col_id == self.grid_cols - 1 + + # Determine padding based on position + left_pad = 0 + right_pad = 1 if is_last_col else 0 # Last column needs right padding + top_pad = 0 + bottom_pad = 1 if is_last_row else 0 # Last row needs bottom padding + self.module.resample[0] = torch.nn.ZeroPad2d((left_pad, right_pad, top_pad, bottom_pad)).eval() + else: + self.module.resample[0] = torch.nn.ZeroPad2d((0, 1, 0, 1)).eval() + + +def plugin_mount(model, cp_group, grid_shape): + """ + Register plugins and allow CP execution of Wan2.1 tokenizer + + Args: + model (torch.nn.Module): instance of `projects.diffusion.v2.tokenizers.wan2pt1.WanVAE_` + cp_group (distributed.ProcessGroup): CP group that will be used + grid_shape (tuple[int, int]): + + Returns: + plugins (dict[str, dict[str, _ModulePlugin]]): dict[layer_name, dict[plugin_id, _ModulePlugin]] dictionarly with plugins, allowing to turn them on/off + """ + + PLUGIN_CONFIG = { + "attn": { + "padding": 24, + "top_k": 24, + "top_k_chunk_size": 24, + "attn_scale": 1.0, + "token_num_scale": True, + "dynamic_scale": True, + }, + "conv_3d": { + "padding": 1, + }, + "conv_layer": {}, + } + + if cp_group is not None: + group_rank_to_global_rank = distributed.get_process_group_ranks(cp_group) + cp_group_size = len(group_rank_to_global_rank) + assert cp_group_size == grid_shape[0] * grid_shape[1] + + in_row_adj_groups, in_col_adj_groups = _create_adj_groups(grid_shape, cp_group) + if cp_group is None: + return {} + assert len(in_col_adj_groups) == grid_shape[1] and all( + len(group) == grid_shape[0] - 1 for group in in_col_adj_groups + ) + assert len(in_row_adj_groups) == grid_shape[0] and all( + len(group) == grid_shape[1] - 1 for group in in_row_adj_groups + ) + + plugins = {} + _conv_3d_plugin_mount(plugins, model, PLUGIN_CONFIG, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups) + _conv_2d_plugin_stride2_mount( + plugins, model, PLUGIN_CONFIG, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups + ) # only for wan vae encoder + _conv_2d_plugin_mount(plugins, model, PLUGIN_CONFIG, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups) + _wanattention_plugin_mount( + plugins, model, PLUGIN_CONFIG, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups + ) + _resample_plugin_mount(plugins, model, PLUGIN_CONFIG, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups) + + return plugins + + +def _wanattention_plugin_mount( + plugins: dict, model, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups +): + plugins["wanattention"] = {} + + wanattention_gather_after = [] + wanattention_split_after = [] + for name, module in chain(model.named_modules()): + if "middle" in name and module.__class__.__name__ == "AttentionBlock": + (wanattention_gather_after if "encoder" in name else wanattention_split_after).append(module) + + for i, wanattention in enumerate(wanattention_gather_after): + plugin_id = "wanattention", i + plugins["wanattention"][plugin_id] = _WanAttentionPlugin( + wanattention, + plugin_id, + plugin_config, + cp_group, + grid_shape, + in_row_adj_groups, + in_col_adj_groups, + all_gather_before_attention=True, + cp_split_after_attention=False, + ) + + for i, wanattention in enumerate(wanattention_split_after): + plugin_id = "wanattention", i + len(wanattention_gather_after) + plugins["wanattention"][plugin_id] = _WanAttentionPlugin( + wanattention, + plugin_id, + plugin_config, + cp_group, + grid_shape, + in_row_adj_groups, + in_col_adj_groups, + all_gather_before_attention=False, + cp_split_after_attention=True, + ) + + +def _conv_3d_plugin_mount( + plugins: dict, model, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups +): + plugins["conv_3d"] = {} + conv3d_s = [] + from cosmos_policy._src.predict2.tokenizers.wan2pt1 import CausalConv3d + + for name, module in chain(model.named_modules()): + if any( + p in name + for p in [ + "encoder.middle.2", + "encoder.head", + "decoder.conv1", + "decoder.middle.0", + ] + ): + continue + + if ( + any( + p in name + for p in [ + "conv1", + "conv2", + ] + ) + and "encoder" not in name + and "decoder" not in name + ): + continue + + if isinstance(module, CausalConv3d) and module.kernel_size[1] > 1: + conv3d_s.append(module) + + for i, conv in enumerate(conv3d_s): + plugin_id = "conv_3d", i + plugins["conv_3d"][plugin_id] = _Conv3DSafeNewPlugin( + conv, plugin_id, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups + ) + + +def _conv_2d_plugin_stride2_mount( + plugins: dict, model, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups +): + plugins["conv_2d_stride2"] = {} + conv2d_stride2_s = [] + + for name, module in model.encoder.named_modules(): + if ( + any( + p in name + for p in [ + "middle.2", + "head", + ] + ) + and ".resample" in name + and module.__class__.__name__ == "Conv2d" + ): + continue + if ".resample" in name and module.__class__.__name__ == "Conv2d": + conv2d_stride2_s.append(module) + + for i, conv in enumerate(conv2d_stride2_s): + plugin_id = "conv_2d_stride2", i + plugins["conv_2d_stride2"][plugin_id] = _Conv2DSafeNewPluginStride2( + conv, plugin_id, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups + ) + + +def _conv_2d_plugin_mount( + plugins: dict, model, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups +): + plugins["conv_2d"] = {} + conv2d_s = [] + for name, module in model.decoder.named_modules(): + if any(p in name for p in ["conv1", "middle.0"]): + continue + if ".resample" in name and module.__class__.__name__ == "Conv2d": + conv2d_s.append(module) + + for i, conv in enumerate(conv2d_s): + plugin_id = "conv_2d", i + plugins["conv_2d"][plugin_id] = _Conv2DSafeNewPlugin( + conv, plugin_id, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups + ) + + +def _resample_plugin_mount( + plugins: dict, model, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups +): + from cosmos_policy._src.predict2.tokenizers.wan2pt1 import Resample + + plugins["resample"] = {} + resamples = [] + for name, module in model.named_modules(): + if isinstance(module, Resample) and module.mode in ["downsample2d", "downsample3d"]: + resamples.append(module) + + for i, resample in enumerate(resamples): + plugin_id = "resample", i + plugins["resample"][plugin_id] = _ResamplePlugin( + resample, plugin_id, plugin_config, cp_group, grid_shape, in_row_adj_groups, in_col_adj_groups + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/tokenizers/wan2pt2.py b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/wan2pt2.py new file mode 100644 index 0000000000000000000000000000000000000000..7e129e44b778137a519d6cd599d9723d9c1a7dff --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/tokenizers/wan2pt2.py @@ -0,0 +1,1133 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved. + +import time +from contextlib import nullcontext + +import torch +import torch.nn as nn +import torch.nn.functional as F +from einops import rearrange + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.distributed import get_rank, sync_model_states +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.predict2.tokenizers.interface import VideoTokenizerInterface +from cosmos_policy._src.predict2.utils.tokenizer_benchmarking import BenchmarkTimes + +__all__ = [ + "WanVAE", +] + +CACHE_T = 2 + + +class CausalConv3d(nn.Conv3d): + """ + Causal 3d convolusion. + """ + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + self._padding = ( + self.padding[2], + self.padding[2], + self.padding[1], + self.padding[1], + 2 * self.padding[0], + 0, + ) + self.padding = (0, 0, 0) + + def forward(self, x, cache_x=None): + padding = list(self._padding) + if cache_x is not None and self._padding[4] > 0: + cache_x = cache_x.to(x.device) + x = torch.cat([cache_x, x], dim=2) + padding[4] -= cache_x.shape[2] + x = F.pad(x, padding) + + return super().forward(x) + + +class RMS_norm(nn.Module): + def __init__(self, dim, channel_first=True, images=True, bias=False): + super().__init__() + broadcastable_dims = (1, 1, 1) if not images else (1, 1) + shape = (dim, *broadcastable_dims) if channel_first else (dim,) + + self.channel_first = channel_first + self.scale = dim**0.5 + self.gamma = nn.Parameter(torch.ones(shape)) + self.bias = nn.Parameter(torch.zeros(shape)) if bias else 0.0 + + def forward(self, x): + return F.normalize(x, dim=(1 if self.channel_first else -1)) * self.scale * self.gamma + self.bias + + +class Upsample(nn.Upsample): + def forward(self, x): + """ + Fix bfloat16 support for nearest neighbor interpolation. + """ + return super().forward(x.float()).type_as(x) + + +class Resample(nn.Module): + def __init__(self, dim, mode): + assert mode in ( + "none", + "upsample2d", + "upsample3d", + "downsample2d", + "downsample3d", + ) + super().__init__() + self.dim = dim + self.mode = mode + + # layers + if mode == "upsample2d": + self.resample = nn.Sequential( + Upsample(scale_factor=(2.0, 2.0), mode="nearest-exact"), + nn.Conv2d(dim, dim, 3, padding=1), + ) + elif mode == "upsample3d": + self.resample = nn.Sequential( + Upsample(scale_factor=(2.0, 2.0), mode="nearest-exact"), + nn.Conv2d(dim, dim, 3, padding=1), + ) + self.time_conv = CausalConv3d(dim, dim * 2, (3, 1, 1), padding=(1, 0, 0)) + elif mode == "downsample2d": + self.resample = nn.Sequential(nn.ZeroPad2d((0, 1, 0, 1)), nn.Conv2d(dim, dim, 3, stride=(2, 2))) + elif mode == "downsample3d": + self.resample = nn.Sequential(nn.ZeroPad2d((0, 1, 0, 1)), nn.Conv2d(dim, dim, 3, stride=(2, 2))) + self.time_conv = CausalConv3d(dim, dim, (3, 1, 1), stride=(2, 1, 1), padding=(0, 0, 0)) + else: + self.resample = nn.Identity() + + def forward(self, x, feat_cache=None, feat_idx=[0]): + b, c, t, h, w = x.size() + if self.mode == "upsample3d": + if feat_cache is not None: + idx = feat_idx[0] + if feat_cache[idx] is None: + feat_cache[idx] = "Rep" + feat_idx[0] += 1 + else: + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None and feat_cache[idx] != "Rep": + # cache last frame of last two chunk + cache_x = torch.cat( + [ + feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), + cache_x, + ], + dim=2, + ) + if cache_x.shape[2] < 2 and feat_cache[idx] is not None and feat_cache[idx] == "Rep": + cache_x = torch.cat( + [torch.zeros_like(cache_x).to(cache_x.device), cache_x], + dim=2, + ) + if feat_cache[idx] == "Rep": + x = self.time_conv(x) + else: + x = self.time_conv(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + x = x.reshape(b, 2, c, t, h, w) + x = torch.stack((x[:, 0, :, :, :, :], x[:, 1, :, :, :, :]), 3) + x = x.reshape(b, c, t * 2, h, w) + t = x.shape[2] + x = rearrange(x, "b c t h w -> (b t) c h w") + x = self.resample(x) + x = rearrange(x, "(b t) c h w -> b c t h w", t=t) + + if self.mode == "downsample3d": + if feat_cache is not None: + idx = feat_idx[0] + if feat_cache[idx] is None: + feat_cache[idx] = x.clone() + feat_idx[0] += 1 + else: + cache_x = x[:, :, -1:, :, :].clone() + x = self.time_conv(torch.cat([feat_cache[idx][:, :, -1:, :, :], x], 2)) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + return x + + +class ResidualBlock(nn.Module): + def __init__(self, in_dim, out_dim, dropout=0.0): + super().__init__() + self.in_dim = in_dim + self.out_dim = out_dim + + # layers + self.residual = nn.Sequential( + RMS_norm(in_dim, images=False), + nn.SiLU(), + CausalConv3d(in_dim, out_dim, 3, padding=1), + RMS_norm(out_dim, images=False), + nn.SiLU(), + nn.Dropout(dropout), + CausalConv3d(out_dim, out_dim, 3, padding=1), + ) + self.shortcut = CausalConv3d(in_dim, out_dim, 1) if in_dim != out_dim else nn.Identity() + + def forward(self, x, feat_cache=None, feat_idx=[0]): + h = self.shortcut(x) + for layer in self.residual: + if isinstance(layer, CausalConv3d) and feat_cache is not None: + idx = feat_idx[0] + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None: + # cache last frame of last two chunk + cache_x = torch.cat( + [ + feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), + cache_x, + ], + dim=2, + ) + x = layer(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + else: + x = layer(x) + return x + h + + +class AttentionBlock(nn.Module): + """ + Causal self-attention with a single head. + """ + + def __init__(self, dim): + super().__init__() + self.dim = dim + + # layers + self.norm = RMS_norm(dim) + self.to_qkv = nn.Conv2d(dim, dim * 3, 1) + self.proj = nn.Conv2d(dim, dim, 1) + + # zero out the last layer params + nn.init.zeros_(self.proj.weight) + + def forward(self, x): + identity = x + b, c, t, h, w = x.size() + x = rearrange(x, "b c t h w -> (b t) c h w") + x = self.norm(x) + # compute query, key, value + q, k, v = self.to_qkv(x).reshape(b * t, 1, c * 3, -1).permute(0, 1, 3, 2).contiguous().chunk(3, dim=-1) + + # apply attention + x = F.scaled_dot_product_attention( + q, + k, + v, + ) + x = x.squeeze(1).permute(0, 2, 1).reshape(b * t, c, h, w) + + # output + x = self.proj(x) + x = rearrange(x, "(b t) c h w-> b c t h w", t=t) + return x + identity + + +def patchify(x, patch_size): + if patch_size == 1: + return x + if x.dim() == 4: + x = rearrange(x, "b c (h q) (w r) -> b (c r q) h w", q=patch_size, r=patch_size) + elif x.dim() == 5: + x = rearrange( + x, + "b c f (h q) (w r) -> b (c r q) f h w", + q=patch_size, + r=patch_size, + ) + else: + raise ValueError(f"Invalid input shape: {x.shape}") + + return x + + +def unpatchify(x, patch_size): + if patch_size == 1: + return x + + if x.dim() == 4: + x = rearrange(x, "b (c r q) h w -> b c (h q) (w r)", q=patch_size, r=patch_size) + elif x.dim() == 5: + x = rearrange( + x, + "b (c r q) f h w -> b c f (h q) (w r)", + q=patch_size, + r=patch_size, + ) + return x + + +class AvgDown3D(nn.Module): + def __init__( + self, + in_channels, + out_channels, + factor_t, + factor_s=1, + ): + super().__init__() + self.in_channels = in_channels + self.out_channels = out_channels + self.factor_t = factor_t + self.factor_s = factor_s + self.factor = self.factor_t * self.factor_s * self.factor_s + + assert in_channels * self.factor % out_channels == 0 + self.group_size = in_channels * self.factor // out_channels + + def forward(self, x: torch.Tensor) -> torch.Tensor: + pad_t = (self.factor_t - x.shape[2] % self.factor_t) % self.factor_t + pad = (0, 0, 0, 0, pad_t, 0) + x = F.pad(x, pad) + B, C, T, H, W = x.shape + x = x.view( + B, + C, + T // self.factor_t, + self.factor_t, + H // self.factor_s, + self.factor_s, + W // self.factor_s, + self.factor_s, + ) + x = x.permute(0, 1, 3, 5, 7, 2, 4, 6).contiguous() + x = x.view( + B, + C * self.factor, + T // self.factor_t, + H // self.factor_s, + W // self.factor_s, + ) + x = x.view( + B, + self.out_channels, + self.group_size, + T // self.factor_t, + H // self.factor_s, + W // self.factor_s, + ) + x = x.mean(dim=2) + return x + + +class DupUp3D(nn.Module): + def __init__( + self, + in_channels: int, + out_channels: int, + factor_t, + factor_s=1, + ): + super().__init__() + self.in_channels = in_channels + self.out_channels = out_channels + + self.factor_t = factor_t + self.factor_s = factor_s + self.factor = self.factor_t * self.factor_s * self.factor_s + + assert out_channels * self.factor % in_channels == 0 + self.repeats = out_channels * self.factor // in_channels + + def forward(self, x: torch.Tensor, first_chunk=False) -> torch.Tensor: + x = x.repeat_interleave(self.repeats, dim=1) + x = x.view( + x.size(0), + self.out_channels, + self.factor_t, + self.factor_s, + self.factor_s, + x.size(2), + x.size(3), + x.size(4), + ) + x = x.permute(0, 1, 5, 2, 6, 3, 7, 4).contiguous() + x = x.view( + x.size(0), + self.out_channels, + x.size(2) * self.factor_t, + x.size(4) * self.factor_s, + x.size(6) * self.factor_s, + ) + if first_chunk: + x = x[:, :, self.factor_t - 1 :, :, :] + return x + + +class Down_ResidualBlock(nn.Module): + def __init__(self, in_dim, out_dim, dropout, mult, temperal_downsample=False, down_flag=False): + super().__init__() + + # Shortcut path with downsample + self.avg_shortcut = AvgDown3D( + in_dim, + out_dim, + factor_t=2 if temperal_downsample else 1, + factor_s=2 if down_flag else 1, + ) + + # Main path with residual blocks and downsample + downsamples = [] + for _ in range(mult): + downsamples.append(ResidualBlock(in_dim, out_dim, dropout)) + in_dim = out_dim + + # Add the final downsample block + if down_flag: + mode = "downsample3d" if temperal_downsample else "downsample2d" + downsamples.append(Resample(out_dim, mode=mode)) + + self.downsamples = nn.Sequential(*downsamples) + + def forward(self, x, feat_cache=None, feat_idx=[0]): + x_copy = x.clone() + for module in self.downsamples: + x = module(x, feat_cache, feat_idx) + + return x + self.avg_shortcut(x_copy) + + +class Up_ResidualBlock(nn.Module): + def __init__(self, in_dim, out_dim, dropout, mult, temperal_upsample=False, up_flag=False): + super().__init__() + # Shortcut path with upsample + if up_flag: + self.avg_shortcut = DupUp3D( + in_dim, + out_dim, + factor_t=2 if temperal_upsample else 1, + factor_s=2 if up_flag else 1, + ) + else: + self.avg_shortcut = None + + # Main path with residual blocks and upsample + upsamples = [] + for _ in range(mult): + upsamples.append(ResidualBlock(in_dim, out_dim, dropout)) + in_dim = out_dim + + # Add the final upsample block + if up_flag: + mode = "upsample3d" if temperal_upsample else "upsample2d" + upsamples.append(Resample(out_dim, mode=mode)) + + self.upsamples = nn.Sequential(*upsamples) + + def forward(self, x, feat_cache=None, feat_idx=[0], first_chunk=False): + x_main = x.clone() + for module in self.upsamples: + x_main = module(x_main, feat_cache, feat_idx) + if self.avg_shortcut is not None: + x_shortcut = self.avg_shortcut(x, first_chunk) + return x_main + x_shortcut + else: + return x_main + + +class Encoder3d(nn.Module): + def __init__( + self, + dim=128, + z_dim=4, + dim_mult=[1, 2, 4, 4], + num_res_blocks=2, + attn_scales=[], + temperal_downsample=[False, True, True], + dropout=0.0, + ): + super().__init__() + self.dim = dim + self.z_dim = z_dim + self.dim_mult = dim_mult + self.num_res_blocks = num_res_blocks + self.attn_scales = attn_scales + self.temperal_downsample = temperal_downsample + + # dimensions + dims = [dim * u for u in [1] + dim_mult] + scale = 1.0 + + # init block + self.conv1 = CausalConv3d(12, dims[0], 3, padding=1) + + # downsample blocks + downsamples = [] + for i, (in_dim, out_dim) in enumerate(zip(dims[:-1], dims[1:])): + t_down_flag = temperal_downsample[i] if i < len(temperal_downsample) else False + downsamples.append( + Down_ResidualBlock( + in_dim=in_dim, + out_dim=out_dim, + dropout=dropout, + mult=num_res_blocks, + temperal_downsample=t_down_flag, + down_flag=i != len(dim_mult) - 1, + ) + ) + scale /= 2.0 + self.downsamples = nn.Sequential(*downsamples) + + # middle blocks + self.middle = nn.Sequential( + ResidualBlock(out_dim, out_dim, dropout), + AttentionBlock(out_dim), + ResidualBlock(out_dim, out_dim, dropout), + ) + + # output blocks + self.head = nn.Sequential( + RMS_norm(out_dim, images=False), + nn.SiLU(), + CausalConv3d(out_dim, z_dim, 3, padding=1), + ) + + def forward(self, x, feat_cache=None, feat_idx=[0]): + if feat_cache is not None: + idx = feat_idx[0] + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None: + cache_x = torch.cat( + [ + feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), + cache_x, + ], + dim=2, + ) + x = self.conv1(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + else: + x = self.conv1(x) + + # downsamples + for layer in self.downsamples: + if feat_cache is not None: + x = layer(x, feat_cache, feat_idx) + else: + x = layer(x) + + # middle + for layer in self.middle: + if isinstance(layer, ResidualBlock) and feat_cache is not None: + x = layer(x, feat_cache, feat_idx) + else: + x = layer(x) + + # head + for layer in self.head: + if isinstance(layer, CausalConv3d) and feat_cache is not None: + idx = feat_idx[0] + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None: + cache_x = torch.cat( + [ + feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), + cache_x, + ], + dim=2, + ) + x = layer(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + else: + x = layer(x) + + return x + + +class Decoder3d(nn.Module): + def __init__( + self, + dim=128, + z_dim=4, + dim_mult=[1, 2, 4, 4], + num_res_blocks=2, + attn_scales=[], + temperal_upsample=[True, True, False], + dropout=0.0, + ): + super().__init__() + self.dim = dim + self.z_dim = z_dim + self.dim_mult = dim_mult + self.num_res_blocks = num_res_blocks + self.attn_scales = attn_scales + self.temperal_upsample = temperal_upsample + + # dimensions + dims = [dim * u for u in [dim_mult[-1]] + dim_mult[::-1]] + scale = 1.0 / 2 ** (len(dim_mult) - 2) + # init block + self.conv1 = CausalConv3d(z_dim, dims[0], 3, padding=1) + + # middle blocks + self.middle = nn.Sequential( + ResidualBlock(dims[0], dims[0], dropout), + AttentionBlock(dims[0]), + ResidualBlock(dims[0], dims[0], dropout), + ) + + # upsample blocks + upsamples = [] + for i, (in_dim, out_dim) in enumerate(zip(dims[:-1], dims[1:])): + t_up_flag = temperal_upsample[i] if i < len(temperal_upsample) else False + upsamples.append( + Up_ResidualBlock( + in_dim=in_dim, + out_dim=out_dim, + dropout=dropout, + mult=num_res_blocks + 1, + temperal_upsample=t_up_flag, + up_flag=i != len(dim_mult) - 1, + ) + ) + self.upsamples = nn.Sequential(*upsamples) + + # output blocks + self.head = nn.Sequential( + RMS_norm(out_dim, images=False), + nn.SiLU(), + CausalConv3d(out_dim, 12, 3, padding=1), + ) + + def forward(self, x, feat_cache=None, feat_idx=[0], first_chunk=False): + if feat_cache is not None: + idx = feat_idx[0] + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None: + cache_x = torch.cat( + [ + feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), + cache_x, + ], + dim=2, + ) + x = self.conv1(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + else: + x = self.conv1(x) + + for layer in self.middle: + if isinstance(layer, ResidualBlock) and feat_cache is not None: + x = layer(x, feat_cache, feat_idx) + else: + x = layer(x) + + # upsamples + for layer in self.upsamples: + if feat_cache is not None: + x = layer(x, feat_cache, feat_idx, first_chunk) + else: + x = layer(x) + + # head + for layer in self.head: + if isinstance(layer, CausalConv3d) and feat_cache is not None: + idx = feat_idx[0] + cache_x = x[:, :, -CACHE_T:, :, :].clone() + if cache_x.shape[2] < 2 and feat_cache[idx] is not None: + cache_x = torch.cat( + [ + feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(cache_x.device), + cache_x, + ], + dim=2, + ) + x = layer(x, feat_cache[idx]) + feat_cache[idx] = cache_x + feat_idx[0] += 1 + else: + x = layer(x) + return x + + +def count_conv3d(model): + count = 0 + for m in model.modules(): + if isinstance(m, CausalConv3d): + count += 1 + return count + + +class WanVAE_(nn.Module): + def __init__( + self, + dim=160, + dec_dim=256, + z_dim=48, + dim_mult=[1, 2, 4, 4], + num_res_blocks=2, + attn_scales=[], + temperal_downsample=[False, True, True], + dropout=0.0, + temporal_window=4, + ): + super().__init__() + self.dim = dim + self.z_dim = z_dim + self.dim_mult = dim_mult + self.num_res_blocks = num_res_blocks + self.attn_scales = attn_scales + self.temperal_downsample = temperal_downsample + self.temperal_upsample = temperal_downsample[::-1] + self.temporal_window = temporal_window + + # modules + self.encoder = Encoder3d( + dim, + z_dim * 2, + dim_mult, + num_res_blocks, + attn_scales, + self.temperal_downsample, + dropout, + ) + self.conv1 = CausalConv3d(z_dim * 2, z_dim * 2, 1) + self.conv2 = CausalConv3d(z_dim, z_dim, 1) + self.decoder = Decoder3d( + dec_dim, + z_dim, + dim_mult, + num_res_blocks, + attn_scales, + self.temperal_upsample, + dropout, + ) + + def forward(self, x, scale=[0, 1]): + mu = self.encode(x, scale) + x_recon = self.decode(mu, scale) + return x_recon, mu + + def encode(self, x, scale): + self.clear_cache() + x = patchify(x, patch_size=2) + t = x.shape[2] + iter_ = 1 + (t - 1) // self.temporal_window + for i in range(iter_): + self._enc_conv_idx = [0] + if i == 0: + out = self.encoder( + x[:, :, :1, :, :], + feat_cache=self._enc_feat_map, + feat_idx=self._enc_conv_idx, + ) + else: + out_ = self.encoder( + x[:, :, 1 + self.temporal_window * (i - 1) : 1 + self.temporal_window * i, :, :], + feat_cache=self._enc_feat_map, + feat_idx=self._enc_conv_idx, + ) + out = torch.cat([out, out_], 2) + if (t - 1) % self.temporal_window: + self._enc_conv_idx = [0] + out_ = self.encoder( + x[:, :, 1 + self.temporal_window * (iter_ - 1) :, :, :], + feat_cache=self._enc_feat_map, + feat_idx=self._enc_conv_idx, + ) + out = torch.cat([out, out_], 2) + mu, log_var = self.conv1(out).chunk(2, dim=1) + if isinstance(scale[0], torch.Tensor): + mu = (mu - scale[0].view(1, self.z_dim, 1, 1, 1)) * scale[1].view(1, self.z_dim, 1, 1, 1) + else: + mu = (mu - scale[0]) * scale[1] + self.clear_cache() + return mu + + @torch.compiler.disable + def _i0_encode(self, x): + """ + If enabled torch.compile uses significantly more memory for this step, so we disable it + """ + out = self.encoder(x[:, :, :1, :, :], feat_cache=self._enc_feat_map, feat_idx=self._enc_conv_idx) + return out + + def decode(self, z, scale): + self.clear_cache() + if isinstance(scale[0], torch.Tensor): + z = z / scale[1].view(1, self.z_dim, 1, 1, 1) + scale[0].view(1, self.z_dim, 1, 1, 1) + else: + z = z / scale[1] + scale[0] + iter_ = z.shape[2] + x = self.conv2(z) + for i in range(iter_): + self._conv_idx = [0] + if i == 0: + out = self.decoder( + x[:, :, i : i + 1, :, :], + feat_cache=self._feat_map, + feat_idx=self._conv_idx, + first_chunk=True, + ) + else: + out_ = self.decoder( + x[:, :, i : i + 1, :, :], + feat_cache=self._feat_map, + feat_idx=self._conv_idx, + ) + out = torch.cat([out, out_], 2) + out = unpatchify(out, patch_size=2) + self.clear_cache() + return out + + def reparameterize(self, mu, log_var): + std = torch.exp(0.5 * log_var) + eps = torch.randn_like(std) + return eps * std + mu + + def sample(self, imgs, deterministic=False): + mu, log_var = self.encode(imgs) + if deterministic: + return mu + std = torch.exp(0.5 * log_var.clamp(-30.0, 20.0)) + return mu + std * torch.randn_like(std) + + def clear_cache(self): + self._conv_num = count_conv3d(self.decoder) + self._conv_idx = [0] + self._feat_map = [None] * self._conv_num + # cache encode + self._enc_conv_num = count_conv3d(self.encoder) + self._enc_conv_idx = [0] + self._enc_feat_map = [None] * self._enc_conv_num + + +def _video_vae( + pretrained_path=None, + device="cpu", + s3_credential_path: str = "credentials/s3_training.secret", + **kwargs, +): + """ + Autoencoder3d adapted from Wan 2.2. + """ + # params + cfg = dict( + temperal_downsample=[False, True, True], + ) + cfg.update(**kwargs) + + # init model + with torch.device("meta"): + model = WanVAE_(**cfg) + + if pretrained_path is None: + model.to_empty(device=device) + else: + if get_rank() == 0: + if pretrained_path.startswith("s3://"): + backend_key = "wan2pt2_vae" + easy_io.set_s3_backend( + key=backend_key, + backend_args={ + "backend": "s3", + "s3_credential_path": s3_credential_path, + }, + ) + else: + backend_key = None + + ckpt = easy_io.load( + pretrained_path, + backend_key=backend_key, + map_location=device, + ) + + # load checkpoint + log.info(f"loading {pretrained_path}") + model.load_state_dict(ckpt, assign=True) + else: + model.to_empty(device=device) + sync_model_states(model) + + return model + + +class WanVAE: + def __init__( + self, + z_dim=48, + vae_pth="s3://bucket/cosmos_diffusion_v2/pretrain_weights/tokenizer/wan2pt2/Wan2.2_VAE.pth", + s3_credential_path: str = "credentials/s3_training.secret", + dtype=torch.bfloat16, + device="cuda", + is_amp=True, + benchmark: bool = False, + temporal_window: int = 4, + ): + self.dtype = dtype + self.device = device + self.benchmark = benchmark + self.temporal_window = temporal_window + + # Wan 2.2 mean and std values (48 dimensions) + mean = [ + -0.2289, + -0.0052, + -0.1323, + -0.2339, + -0.2799, + 0.0174, + 0.1838, + 0.1557, + -0.1382, + 0.0542, + 0.2813, + 0.0891, + 0.1570, + -0.0098, + 0.0375, + -0.1825, + -0.2246, + -0.1207, + -0.0698, + 0.5109, + 0.2665, + -0.2108, + -0.2158, + 0.2502, + -0.2055, + -0.0322, + 0.1109, + 0.1567, + -0.0729, + 0.0899, + -0.2799, + -0.1230, + -0.0313, + -0.1649, + 0.0117, + 0.0723, + -0.2839, + -0.2083, + -0.0520, + 0.3748, + 0.0152, + 0.1957, + 0.1433, + -0.2944, + 0.3573, + -0.0548, + -0.1681, + -0.0667, + ] + std = [ + 0.4765, + 1.0364, + 0.4514, + 1.1677, + 0.5313, + 0.4990, + 0.4818, + 0.5013, + 0.8158, + 1.0344, + 0.5894, + 1.0901, + 0.6885, + 0.6165, + 0.8454, + 0.4978, + 0.5759, + 0.3523, + 0.7135, + 0.6804, + 0.5833, + 1.4146, + 0.8986, + 0.5659, + 0.7069, + 0.5338, + 0.4889, + 0.4917, + 0.4069, + 0.4999, + 0.6866, + 0.4093, + 0.5709, + 0.6065, + 0.6415, + 0.4944, + 0.5726, + 1.2042, + 0.5458, + 1.6887, + 0.3971, + 1.0600, + 0.3943, + 0.5537, + 0.5444, + 0.4089, + 0.7468, + 0.7744, + ] + + self.mean = torch.tensor(mean, dtype=dtype, device=device) + self.std = torch.tensor(std, dtype=dtype, device=device) + self.scale = [self.mean, 1.0 / self.std] + + # init model + self.model = _video_vae( + pretrained_path=vae_pth, + s3_credential_path=s3_credential_path, + device=device, + temporal_window=temporal_window, + ) + self.model = self.model.eval().requires_grad_(False) + self.is_amp = is_amp + if not is_amp: + self.model = self.model.to(dtype=dtype) + self.context = nullcontext() + else: + self.context = torch.amp.autocast("cuda", dtype=dtype) + + def count_param(self): + return sum(p.numel() for p in self.model.parameters()) + + @torch.no_grad() + def encode(self, videos): + """ + videos: A list of videos each with shape [C, T, H, W]. + """ + if self.benchmark: + torch.cuda.synchronize() + benchmark_times = BenchmarkTimes() + total_time = time.perf_counter() + in_dtype = videos.dtype + with self.context: + if not self.is_amp: + videos = videos.to(self.dtype) + if self.benchmark: + torch.cuda.synchronize() + model_time = time.perf_counter() + latent = self.model.encode(videos, self.scale) + if self.benchmark: + torch.cuda.synchronize() + benchmark_times.model_invocation = time.perf_counter() - model_time + latent = latent.to(in_dtype) + if self.benchmark: + torch.cuda.synchronize() + benchmark_times.total = time.perf_counter() - total_time + return latent, benchmark_times + return latent + + @torch.no_grad() + def decode(self, zs): + if self.benchmark: + torch.cuda.synchronize() + benchmark_times = BenchmarkTimes() + total_time = time.perf_counter() + in_dtype = zs.dtype + with self.context: + if not self.is_amp: + zs = zs.to(self.dtype) + if self.benchmark: + torch.cuda.synchronize() + model_time = time.perf_counter() + video_recon = self.model.decode(zs, self.scale) + if self.benchmark: + torch.cuda.synchronize() + benchmark_times.model_invocation = time.perf_counter() - model_time + video_recon = video_recon.to(in_dtype) + if self.benchmark: + torch.cuda.synchronize() + benchmark_times.total = time.perf_counter() - total_time + return video_recon, benchmark_times + return video_recon + + +class Wan2pt2VAEInterface(VideoTokenizerInterface): + def __init__(self, chunk_duration: int = 93, **kwargs): + self.model = WanVAE( + dtype=torch.bfloat16, + is_amp=False, + vae_pth=kwargs.get( + "vae_pth", + "s3://bucket/cosmos_diffusion_v2/pretrain_weights/tokenizer/wan2pt2/Wan2.2_VAE.pth", + ), + s3_credential_path=kwargs.get("s3_credential_path", "credentials/s3_training.secret"), + temporal_window=kwargs.get("temporal_window", 4), + ) + + del kwargs + self.chunk_duration = chunk_duration + + @property + def dtype(self): + return self.model.dtype + + def reset_dtype(self): + pass + + def encode(self, state: torch.Tensor) -> torch.Tensor: + latents = self.model.encode(state) + return latents + + def decode(self, latent: torch.Tensor) -> torch.Tensor: + return self.model.decode(latent) + + def get_latent_num_frames(self, num_pixel_frames: int) -> int: + return 1 + (num_pixel_frames - 1) // 4 + + def get_pixel_num_frames(self, num_latent_frames: int) -> int: + return (num_latent_frames - 1) * 4 + 1 + + @property + def spatial_compression_factor(self): + return 16 # 2x from patchify + 8x from spatial downsampling + + @property + def temporal_compression_factor(self): + return 4 + + @property + def pixel_chunk_duration(self): + return self.chunk_duration + + @property + def latent_chunk_duration(self): + return self.get_latent_num_frames(self.chunk_duration) + + @property + def latent_ch(self): + return 48 + + @property + def spatial_resolution(self): + return 512 + + @property + def name(self): + return "wan2pt2_tokenizer" diff --git a/REGEN-main/cosmos_policy/_src/predict2/utils/dtensor_helper.py b/REGEN-main/cosmos_policy/_src/predict2/utils/dtensor_helper.py new file mode 100644 index 0000000000000000000000000000000000000000..5a6e7e61f6016822c32c241ffc1cd31787482de9 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/utils/dtensor_helper.py @@ -0,0 +1,88 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +import itertools +from typing import Any + +import torch +import torch.distributed as dist +from torch.distributed.device_mesh import DeviceMesh + +from cosmos_policy._src.imaginaire.utils.misc import get_local_tensor_if_DTensor + + +class DTensorFastEmaModelUpdater: + """ + Similar as FastEmaModelUpdater + """ + + def __init__(self): + # Flag to indicate whether the cache is taken or not. Useful to avoid cache overwrite + self.is_cached = False + + def copy_to(self, src_model: torch.nn.Module, tgt_model: torch.nn.Module) -> None: + with torch.no_grad(): + for tgt_params, src_params in zip(tgt_model.parameters(), src_model.parameters()): + tgt_params.to_local().data.copy_(src_params.to_local().data) + + @torch.no_grad() + def update_average(self, src_model: torch.nn.Module, tgt_model: torch.nn.Module, beta: float = 0.9999) -> None: + target_list = [] + source_list = [] + for tgt_params, src_params in zip(tgt_model.parameters(), src_model.parameters()): + assert tgt_params.dtype == torch.float32, ( + f"EMA model only works in FP32 dtype, got {tgt_params.dtype} instead." + ) + target_list.append(tgt_params.to_local()) + source_list.append(src_params.to_local().data) + torch._foreach_mul_(target_list, beta) + torch._foreach_add_(target_list, source_list, alpha=1.0 - beta) + + @torch.no_grad() + def cache(self, parameters: Any, is_cpu: bool = False) -> None: + assert self.is_cached is False, "EMA cache is already taken. Did you forget to restore it?" + device = "cpu" if is_cpu else "cuda" + self.collected_params = [param.to_local().clone().to(device) for param in parameters] + self.is_cached = True + + @torch.no_grad() + def restore(self, parameters: Any) -> None: + assert self.is_cached, "EMA cache is not taken yet." + for c_param, param in zip(self.collected_params, parameters, strict=False): + param.to_local().copy_(c_param.data.type_as(param.data)) + self.collected_params = [] + # Release the cache after we call restore + self.is_cached = False + + +def broadcast_dtensor_model_states(model: torch.nn.Module, mesh: DeviceMesh): + """Broadcast model states from replicate mesh's rank 0.""" + replicate_group = mesh.get_group("replicate") + all_ranks = dist.get_process_group_ranks(replicate_group) + if len(all_ranks) == 1: + return + + for _, tensor in itertools.chain(model.named_parameters(), model.named_buffers()): + # Get src rank which is the first rank in each replication group + src_rank = all_ranks[0] + # Broadcast the local tensor + local_tensor = get_local_tensor_if_DTensor(tensor) + dist.broadcast( + local_tensor, + src=src_rank, + group=replicate_group, + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/utils/flash_attention_jvp_triton.py b/REGEN-main/cosmos_policy/_src/predict2/utils/flash_attention_jvp_triton.py new file mode 100644 index 0000000000000000000000000000000000000000..6db3a4f56490347ecd3383e06e44ca48f48aabb1 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/utils/flash_attention_jvp_triton.py @@ -0,0 +1,824 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Flash Attention with JVP +=============== + +This is a Triton implementation of the Flash Attention v2 algorithm from Tri Dao (https://tridao.me/publications/flash2/flash2.pdf) + +Taken from https://github.com/triton-lang/triton/blob/main/python/tutorials/06-fused-attention.py (2025/03) + +Simplified version, combining Triton forward and official backward (Kaiwen Zheng) + +Modified to support Jacobian-vector-product (JVP) computation (Kaiwen Zheng) + +Credits: OpenAI kernel team + +Extra Credits: + +* Original flash attention paper (https://arxiv.org/abs/2205.14135) +* Rabe and Staats (https://arxiv.org/pdf/2112.05682v2.pdf) + +""" + +import torch +import triton +import triton.language as tl +from einops import rearrange +from flash_attn.flash_attn_interface import _flash_attn_backward, _flash_attn_varlen_backward + +DEVICE = "cuda" + + +@triton.jit +def _attn_fwd_inner( + acc, + acc_A, + acc_B, + l_i, + m_i, + r_i, + q, + tq, # + K_block_ptr, + V_block_ptr, + tK_block_ptr, + tV_block_ptr, # + start_m, + sm_scale, # + BLOCK_M: tl.constexpr, + BLOCK_N: tl.constexpr, # + STAGE: tl.constexpr, + offs_m: tl.constexpr, + offs_n: tl.constexpr, # + SEQ_LEN_KV: tl.constexpr, + HEAD_DIM_V: tl.constexpr, + bf16_v: tl.constexpr, +): + # range of values handled by this stage + if STAGE == 1: + lo, hi = 0, min(start_m * BLOCK_M, SEQ_LEN_KV) + elif STAGE == 2: + lo, hi = start_m * BLOCK_M, min((start_m + 1) * BLOCK_M, SEQ_LEN_KV) + lo = tl.multiple_of(lo, BLOCK_M) + # causal = False + else: + lo, hi = 0, SEQ_LEN_KV + qk_scale = sm_scale * 1.44269504 + K_block_ptr = tl.advance(K_block_ptr, (0, lo)) + V_block_ptr = tl.advance(V_block_ptr, (lo, 0)) + tK_block_ptr = tl.advance(tK_block_ptr, (0, lo)) + tV_block_ptr = tl.advance(tV_block_ptr, (lo, 0)) + # loop over k, v and update accumulator + for start_n in range(lo, hi, BLOCK_N): + start_n = tl.multiple_of(start_n, BLOCK_N) + # -- compute qk ---- + k, tk = ( + tl.load(K_block_ptr, boundary_check=(0, 1), padding_option="zero"), + tl.load(tK_block_ptr, boundary_check=(0, 1), padding_option="zero"), + ) + qk = tl.dot(q, k) + + tS_ij = tl.dot(tq, k) + tS_ij = tl.dot(q, tk, tS_ij) + tS_ij *= sm_scale + if STAGE == 2: + causal_mask = offs_m[:, None] >= (start_n + offs_n[None, :]) + qk = qk * qk_scale + tl.where(causal_mask, 0, -1.0e6) + m_ij = tl.maximum(m_i, tl.max(qk, 1)) + qk -= m_ij[:, None] + else: + m_ij = tl.maximum(m_i, tl.max(qk, 1) * qk_scale) + qk = qk * qk_scale - m_ij[:, None] + # mask if SEQ_LEN_KV % BLOCK_N != 0 + boundary_m = tl.full([BLOCK_M], hi, dtype=tl.int32) + size_n = start_n + offs_n[None, :] + mask = size_n < boundary_m[:, None] + qk = tl.where(mask, qk, float("-inf")) + p = tl.math.exp2(qk) + l_ij = tl.sum(p, 1) + tS_ij = tl.where(mask, tS_ij, float("0")) + H_ij = p * tS_ij + r_ij = tl.sum(H_ij, 1) + # -- update m_i and l_i + alpha = tl.math.exp2(m_i - m_ij) + l_i = l_i * alpha + l_ij + r_i = r_i * alpha + r_ij + # -- update output accumulator -- + acc = acc * alpha[:, None] + acc_A = acc_A * alpha[:, None] + acc_B = acc_B * alpha[:, None] + # update acc + v, tv = ( + tl.load(V_block_ptr, boundary_check=(0, 1), padding_option="zero"), + tl.load(tV_block_ptr, boundary_check=(0, 1), padding_option="zero"), + ) + # boundary_v = tl.full([HEAD_DIM_V], hi, dtype=tl.int32) + # size_n = start_n + offs_n + # mask_v = size_n[:, None] < boundary_v[None, :] + # v = tl.where(mask_v, v, float("0")) + # tv = tl.where(mask_v, tv, float("0")) + if bf16_v: + p = p.to(tl.bfloat16) + H_ij = H_ij.to(tl.bfloat16) + v = v.to(tl.bfloat16) + tv = tv.to(tl.bfloat16) + else: + p = p.to(tl.float16) + H_ij = H_ij.to(tl.float16) + v = v.to(tl.float16) + tv = tv.to(tl.float16) + acc = tl.dot(p, v, acc) + acc_A = tl.dot(p, tv, acc_A) + acc_B = tl.dot(H_ij, v, acc_B) + # update m_i and l_i + m_i = m_ij + V_block_ptr = tl.advance(V_block_ptr, (BLOCK_N, 0)) + K_block_ptr = tl.advance(K_block_ptr, (0, BLOCK_N)) + tV_block_ptr = tl.advance(tV_block_ptr, (BLOCK_N, 0)) + tK_block_ptr = tl.advance(tK_block_ptr, (0, BLOCK_N)) + return acc, acc_A, acc_B, l_i, m_i, r_i + + +# We don't run auto-tuning every time to keep the tutorial fast. Keeping +# the code below and commenting out the equivalent parameters is convenient for +# re-tuning. +configs = [ + triton.Config({"BLOCK_M": BM, "BLOCK_N": BN}, num_stages=s, num_warps=w) + for BM in [64, 128] + for BN in [16, 32, 64] + for s in [3, 4, 7] + for w in [4, 8] +] + + +@triton.autotune(configs, key=["SEQ_LEN_Q", "SEQ_LEN_KV", "HEAD_DIM_QK", "HEAD_DIM_V"]) +@triton.jit +def _attn_fwd( + Q, + K, + V, + tQ, + tK, + tV, + sm_scale, + M, + Out, + tOut, # + stride_qz, + stride_qh, + stride_qm, + stride_qd, # + stride_kz, + stride_kh, + stride_kn, + stride_kd, # + stride_vz, + stride_vh, + stride_vn, + stride_vd, # + stride_oz, + stride_oh, + stride_om, + stride_od, # + Z, + H, # + SEQ_LEN_Q, + SEQ_LEN_KV, # + HEAD_DIM_QK: tl.constexpr, + HEAD_DIM_V: tl.constexpr, # + BLOCK_M: tl.constexpr, # + BLOCK_N: tl.constexpr, # + STAGE: tl.constexpr, # +): + start_m = tl.program_id(0) + off_hz = tl.program_id(1) + off_z = off_hz // H + off_h = off_hz % H + q_offset = off_z.to(tl.int64) * stride_qz + off_h.to(tl.int64) * stride_qh + k_offset = off_z.to(tl.int64) * stride_kz + off_h.to(tl.int64) * stride_kh + v_offset = off_z.to(tl.int64) * stride_vz + off_h.to(tl.int64) * stride_vh + o_offset = off_z.to(tl.int64) * stride_oz + off_h.to(tl.int64) * stride_oh + + start_m_idx = start_m * BLOCK_M + # end_m_idx = (start_m + 1) * BLOCK_M + + # block pointers + Q_block_ptr = tl.make_block_ptr( + base=Q + q_offset, + shape=(SEQ_LEN_Q, HEAD_DIM_QK), + strides=(stride_qm, stride_qd), + offsets=(start_m_idx, 0), + block_shape=(BLOCK_M, HEAD_DIM_QK), + order=(1, 0), + ) + V_block_ptr = tl.make_block_ptr( + base=V + v_offset, + shape=(SEQ_LEN_KV, HEAD_DIM_V), + strides=(stride_vn, stride_vd), + offsets=(0, 0), + block_shape=(BLOCK_N, HEAD_DIM_V), + order=(1, 0), + ) + # load transposed K + K_block_ptr = tl.make_block_ptr( + base=K + k_offset, + shape=(HEAD_DIM_QK, SEQ_LEN_KV), + strides=(stride_kd, stride_kn), + offsets=(0, 0), + block_shape=(HEAD_DIM_QK, BLOCK_N), + order=(0, 1), + ) + O_block_ptr = tl.make_block_ptr( + base=Out + o_offset, + shape=(SEQ_LEN_Q, HEAD_DIM_V), + strides=(stride_om, stride_od), + offsets=(start_m_idx, 0), + block_shape=(BLOCK_M, HEAD_DIM_V), + order=(1, 0), + ) + tQ_block_ptr = tl.make_block_ptr( + base=tQ + q_offset, + shape=(SEQ_LEN_Q, HEAD_DIM_QK), + strides=(stride_qm, stride_qd), + offsets=(start_m_idx, 0), + block_shape=(BLOCK_M, HEAD_DIM_QK), + order=(1, 0), + ) + tV_block_ptr = tl.make_block_ptr( + base=tV + v_offset, + shape=(SEQ_LEN_KV, HEAD_DIM_V), + strides=(stride_vn, stride_vd), + offsets=(0, 0), + block_shape=(BLOCK_N, HEAD_DIM_V), + order=(1, 0), + ) + # load transposed K + tK_block_ptr = tl.make_block_ptr( + base=tK + k_offset, + shape=(HEAD_DIM_QK, SEQ_LEN_KV), + strides=(stride_kd, stride_kn), + offsets=(0, 0), + block_shape=(HEAD_DIM_QK, BLOCK_N), + order=(0, 1), + ) + tO_block_ptr = tl.make_block_ptr( + base=tOut + o_offset, + shape=(SEQ_LEN_Q, HEAD_DIM_V), + strides=(stride_om, stride_od), + offsets=(start_m_idx, 0), + block_shape=(BLOCK_M, HEAD_DIM_V), + order=(1, 0), + ) + # initialize offsets + offs_m = start_m_idx + tl.arange(0, BLOCK_M) + offs_n = tl.arange(0, BLOCK_N) + offs_d_qk, offs_d_v = tl.arange(0, HEAD_DIM_QK), tl.arange(0, HEAD_DIM_V) + # initialize pointer to m and l + m_i = tl.zeros([BLOCK_M], dtype=tl.float32) - float("inf") + l_i = tl.zeros([BLOCK_M], dtype=tl.float32) # + 1.0 + acc = tl.zeros([BLOCK_M, HEAD_DIM_V], dtype=tl.float32) + r_i = tl.zeros([BLOCK_M], dtype=tl.float32) + acc_A = tl.zeros([BLOCK_M, HEAD_DIM_V], dtype=tl.float32) + acc_B = tl.zeros([BLOCK_M, HEAD_DIM_V], dtype=tl.float32) + # load q: it will stay in SRAM throughout + q, tq = ( + tl.load(Q_block_ptr, boundary_check=(0, 1), padding_option="zero"), + tl.load(tQ_block_ptr, boundary_check=(0, 1), padding_option="zero"), + ) + # stage 1: off-band + # For causal = True, STAGE = 3 and _attn_fwd_inner gets 1 as its STAGE + # For causal = False, STAGE = 1, and _attn_fwd_inner gets 3 as its STAGE + if STAGE & 1: + acc, acc_A, acc_B, l_i, m_i, r_i = _attn_fwd_inner( + acc, + acc_A, + acc_B, + l_i, + m_i, + r_i, + q, + tq, # + K_block_ptr, + V_block_ptr, + tK_block_ptr, + tV_block_ptr, # + start_m, + sm_scale, # + BLOCK_M, + BLOCK_N, # + 4 - STAGE, + offs_m, + offs_n, + SEQ_LEN_KV, + HEAD_DIM_V, + V.dtype.element_ty == tl.bfloat16, # + ) + # stage 2: on-band + if STAGE & 2: + # barrier makes it easier for compielr to schedule the + # two loops independently + acc, acc_A, acc_B, l_i, m_i, r_i = _attn_fwd_inner( + acc, + acc_A, + acc_B, + l_i, + m_i, + r_i, + q, + tq, # + K_block_ptr, + V_block_ptr, + tK_block_ptr, + tV_block_ptr, # + start_m, + sm_scale, # + BLOCK_M, + BLOCK_N, # + 2, + offs_m, + offs_n, + SEQ_LEN_KV, + HEAD_DIM_V, + V.dtype.element_ty == tl.bfloat16, # + ) + + # epilogue + # m_i += tl.math.log2(l_i) + empty_mask = l_i == 0.0 + # NOTE: This happens if the entire block is masked out. + l_i = tl.where(empty_mask, 1.0, l_i) + # NOTE: This is needed to compute the logsumexp for the backward pass. + m_i = m_i + tl.where( + empty_mask, + 0.0, + tl.math.log2(l_i), + ) + + acc = acc / l_i[:, None] + tO_i = (acc_A + acc_B - (r_i[:, None] * acc)) / l_i[:, None] + m_ptrs = M + off_hz * SEQ_LEN_Q + offs_m + O_block_ptr = Out + o_offset + offs_m[:, None] * stride_om + offs_d_v[None, :] * stride_od + tO_block_ptr = tOut + o_offset + offs_m[:, None] * stride_om + offs_d_v[None, :] * stride_od + # mask if SEQ_LEN_Q % BLOCK_M != 0 + mask_lse = offs_m < SEQ_LEN_Q + mask = offs_m[:, None] < SEQ_LEN_Q + tl.store(m_ptrs, m_i * 0.69314718, mask=mask_lse) + tl.store(O_block_ptr, acc.to(Out.type.element_ty), mask=mask) + tl.store(tO_block_ptr, tO_i.to(tOut.type.element_ty), mask=mask) + + +def generate_qkv(q, k, v): + """ + Arguments: + q: (batch_size, nheads, seqlen_q, d) + k: (batch_size, nheads_k, seqlen_k, d) + v: (batch_size, nheads_k, seqlen_k, d) + """ + batch_size, _, seqlen_q, d = q.shape + _, nheads_k, seqlen_k, _ = k.shape + assert k.shape == (batch_size, nheads_k, seqlen_k, d) + assert v.shape == (batch_size, nheads_k, seqlen_k, d) + + def unpad_fn(x): + return rearrange(x, "b h s d -> (b s) h d") + + def lse_unpad_fn(x): + return rearrange(x, "b h s -> (b s) h") + + def pad_fn(x): + return rearrange(x, "(b s) h d -> b h s d", b=batch_size) + + # unpad_fn = lambda x: rearrange(x, "b h s d -> (b s) h d") + # lse_unpad_fn = lambda x: rearrange(x, "b h s -> (b s) h") + # pad_fn = lambda x: rearrange(x, "(b s) h d -> b h s d", b=batch_size) + + cu_seqlens_q = torch.arange(0, (batch_size + 1) * seqlen_q, step=seqlen_q, dtype=torch.int32, device=q.device) + max_seqlen_q = seqlen_q + + cu_seqlens_k = torch.arange(0, (batch_size + 1) * seqlen_k, step=seqlen_k, dtype=torch.int32, device=q.device) + max_seqlen_k = seqlen_k + + return ( + unpad_fn, + lse_unpad_fn, + pad_fn, + cu_seqlens_q, + cu_seqlens_k, + max_seqlen_q, + max_seqlen_k, + ) + + +class _attention(torch.autograd.Function): + """ + Arguments: + q, tq: (batch_size, nheads, seqlen_q, d_qk) + k, tk: (batch_size, nheads, seqlen_kv, d_qk) + v, tv: (batch_size, nheads, seqlen_kv, d_v) + Returns: + o, to: (batch_size, nheads, seqlen_q, d_v) + + Backward is only supported when d_qk=d_v. + """ + + @staticmethod + def forward(ctx, q, k, v, tq, tk, tv, causal=False, sm_scale=None): + is_grad = any(x.requires_grad for x in [q, k, v]) + # shape constraints + assert q.shape[:-2] == k.shape[:-2] and k.shape[:-2] == v.shape[:-2] + assert k.shape[-2] == v.shape[-2] and q.shape[-1] == k.shape[-1] + Z, H = q.shape[:-2] + SEQ_LEN_Q, SEQ_LEN_KV = q.shape[-2], k.shape[-2] + HEAD_DIM_QK, HEAD_DIM_V = q.shape[-1], v.shape[-1] + assert HEAD_DIM_QK in {16, 32, 64, 128, 256} + assert HEAD_DIM_V in {16, 32, 64, 128, 256} + assert (SEQ_LEN_Q == SEQ_LEN_KV) or (not causal), "Causal cross-attention is currently not supported." + assert tq.shape == q.shape and tk.shape == k.shape and tv.shape == v.shape + assert tq.stride() == q.stride() and tk.stride() == k.stride() and tv.stride() == v.stride() + if sm_scale is None: + sm_scale = HEAD_DIM_QK ** (-0.5) + o = torch.empty((Z, H, SEQ_LEN_Q, HEAD_DIM_V), device=q.device, dtype=q.dtype) + to = torch.empty_like(o) + stage = 3 if causal else 1 + + M = torch.empty((Z, H, SEQ_LEN_Q), device=q.device, dtype=torch.float32) + + def grid(args): + return (triton.cdiv(SEQ_LEN_Q, args["BLOCK_M"]), Z * H, 1) + + # grid = lambda args: (triton.cdiv(SEQ_LEN_Q, args["BLOCK_M"]), Z * H, 1) + ctx.grid = grid + _attn_fwd[grid]( + q, + k, + v, + tq, + tk, + tv, + sm_scale, + M, + o, + to, # + q.stride(0), + q.stride(1), + q.stride(2), + q.stride(3), # + k.stride(0), + k.stride(1), + k.stride(2), + k.stride(3), # + v.stride(0), + v.stride(1), + v.stride(2), + v.stride(3), # + o.stride(0), + o.stride(1), + o.stride(2), + o.stride(3), # + Z, + H, # + SEQ_LEN_Q, + SEQ_LEN_KV, # + HEAD_DIM_QK, + HEAD_DIM_V, # + STAGE=stage, + ) + + if is_grad: + ctx.save_for_backward(q, k, v, o, M) + ctx.sm_scale = sm_scale + ctx.causal = causal + return o, to + + @staticmethod + def backward(ctx, dout, *args): + q, k, v, out, softmax_lse = ctx.saved_tensors + assert q.shape[-1] == k.shape[-1] and k.shape[-1] == v.shape[-1], ( + "Backward not supported with different headdim." + ) + # flash_attn uses the shape (batch_size, seqlen, nheads, headdim) + # torch.nn.functional.scaled_dot_product_attention and this implementation use (batch_size, nheads, seqlen, headdim) + if q.shape[-2] == k.shape[-2]: + dq, dk, dv = torch.empty_like(q), torch.empty_like(k), torch.empty_like(v) + _flash_attn_backward( + dout.transpose(1, 2), + q.transpose(1, 2), + k.transpose(1, 2), + v.transpose(1, 2), + out.transpose(1, 2), + softmax_lse, + dq.transpose(1, 2), + dk.transpose(1, 2), + dv.transpose(1, 2), + dropout_p=0.0, + softmax_scale=ctx.sm_scale, + causal=ctx.causal, + window_size=(-1, -1), + # softcap=0, + alibi_slopes=None, + deterministic=False, + ) + else: + unpad_fn, lse_unpad_fn, pad_fn, cu_seqlens_q, cu_seqlens_k, max_seqlen_q, max_seqlen_k = generate_qkv( + q, k, v + ) + q_unpad, k_unpad, v_unpad = unpad_fn(q), unpad_fn(k), unpad_fn(v) + dq, dk, dv = torch.empty_like(q_unpad), torch.empty_like(k_unpad), torch.empty_like(v_unpad) + _flash_attn_varlen_backward( + unpad_fn(dout), + q_unpad, + k_unpad, + v_unpad, + unpad_fn(out), + lse_unpad_fn(softmax_lse), + dq, + dk, + dv, + cu_seqlens_q, + cu_seqlens_k, + max_seqlen_q, + max_seqlen_k, + dropout_p=0.0, + softmax_scale=ctx.sm_scale, + causal=ctx.causal, + window_size=(-1, -1), + # softcap=0, + alibi_slopes=None, + deterministic=False, + ) + dq, dk, dv = pad_fn(dq), pad_fn(dk), pad_fn(dv) + return dq, dk, dv, None, None, None, None, None + + +attention = _attention.apply + + +# def _test_fwd_bwd(Z, H, SEQ_LEN, HEAD_DIM, causal, dtype=torch.float16): +# torch.manual_seed(20) +# q = torch.empty((Z, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5).requires_grad_() +# k = torch.empty((Z, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5).requires_grad_() +# v = torch.empty((Z, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5).requires_grad_() +# tq = torch.zeros_like(q) +# tk = torch.zeros_like(k) +# tv = torch.zeros_like(v) +# sm_scale = 0.5 +# dout = torch.randn_like(q) +# # reference implementation +# M = torch.tril(torch.ones((SEQ_LEN, SEQ_LEN), device=DEVICE)) +# p = torch.matmul(q, k.transpose(2, 3)) * sm_scale +# if causal: +# p[:, :, M == 0] = float("-inf") +# p = torch.softmax(p.float(), dim=-1).to(dtype) +# ref_out = torch.matmul(p, v) +# ref_out.backward(dout) +# ref_dv, v.grad = v.grad.clone(), None +# ref_dk, k.grad = k.grad.clone(), None +# ref_dq, q.grad = q.grad.clone(), None +# # triton implementation +# tri_out = attention(q, k, v, tq, tk, tv, causal, sm_scale)[0].to(dtype) +# tri_out.backward(dout) +# tri_dv, v.grad = v.grad.clone(), None +# tri_dk, k.grad = k.grad.clone(), None +# tri_dq, q.grad = q.grad.clone(), None +# # compare +# rtol = 2e-2 if dtype == torch.bfloat16 else 0 +# torch.testing.assert_close(ref_out, tri_out, atol=1e-2, rtol=0) +# torch.testing.assert_close(ref_dq, tri_dq, atol=1e-2, rtol=rtol / 2) +# torch.testing.assert_close(ref_dk, tri_dk, atol=1e-2, rtol=rtol / 2) +# torch.testing.assert_close(ref_dv, tri_dv, atol=1e-2, rtol=rtol) + + +# def test_fwd_bwd(): +# for shape in [(1, 2, 1024, 64), (1, 2, 999, 64)]: +# for causal in [True, False]: +# for dtype in [torch.float16, torch.bfloat16]: +# _test_fwd_bwd(*shape, causal, dtype) +# print(f"Shape={shape}, Causal={causal}, Dtype={dtype} Passed (SA fwd/bwd).") + + +# def _test_jvp(Z, H, SEQ_LEN, HEAD_DIM, causal, dtype=torch.float16): +# torch.manual_seed(20) +# q = torch.empty((Z, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5).requires_grad_() +# k = torch.empty((Z, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5).requires_grad_() +# v = torch.empty((Z, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5).requires_grad_() +# tq = torch.empty((Z, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5) +# tk = torch.empty((Z, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5) +# tv = torch.empty((Z, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5) +# sm_scale = 0.5 + +# def naive_attention(q, k, v): +# # reference implementation +# M = torch.tril(torch.ones((SEQ_LEN, SEQ_LEN), device=DEVICE)) +# p = torch.matmul(q, k.transpose(2, 3)) * sm_scale +# if causal: +# p[:, :, M == 0] = float("-inf") +# p = torch.softmax(p.float(), dim=-1).to(dtype) +# ref_out = torch.matmul(p, v) +# return ref_out + +# _, ref_tout = torch.func.jvp(naive_attention, (q, k, v), (tq, tk, tv)) +# # triton implementation +# tri_tout = attention(q, k, v, tq, tk, tv, causal, sm_scale)[1].to(dtype) +# # compare +# torch.testing.assert_close(ref_tout, tri_tout, atol=1e-2, rtol=1e-2) + + +# def test_jvp(): +# for shape in [(1, 2, 1024, 64), (1, 2, 999, 64)]: +# for causal in [True, False]: +# for dtype in [torch.float16, torch.bfloat16]: +# _test_jvp(*shape, causal, dtype) +# print(f"Shape={shape}, Causal={causal}, Dtype={dtype} Passed (SA JVP).") + + +# BATCH, N_HEADS, HEAD_DIM = 4, 32, 64 +# # vary seq length for fixed head and batch=4 +# configs = [] +# for mode in ["fwd", "bwd"]: +# for causal in [True, False]: +# if mode == "bwd" and not causal: +# continue +# configs.append( +# triton.testing.Benchmark( +# x_names=["SEQ_LEN"], +# x_vals=[2**i for i in range(10, 15)], +# line_arg="provider", +# line_vals=["triton-fp16", "flash"], +# line_names=["Triton [FP16]", "FlashAttn-2"], +# styles=[("red", "-"), ("blue", "-"), ("green", "-")], +# ylabel="TFLOPS", +# plot_name=f"fused-attention-batch{BATCH}-head{N_HEADS}-d{HEAD_DIM}-{mode}-causal={causal}", +# args={ +# "H": N_HEADS, +# "BATCH": BATCH, +# "HEAD_DIM": HEAD_DIM, +# "mode": mode, +# "causal": causal, +# }, +# ) +# ) + + +# @triton.testing.perf_report(configs) +# def bench_flash_attention(BATCH, H, SEQ_LEN, HEAD_DIM, causal, mode, provider, device=DEVICE): +# assert mode in ["fwd", "bwd"] +# dtype = torch.float16 +# if "triton" in provider: +# q = torch.randn((BATCH, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=device, requires_grad=True) +# k = torch.randn((BATCH, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=device, requires_grad=True) +# v = torch.randn((BATCH, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=device, requires_grad=True) +# tq = torch.zeros_like(q) +# tk = torch.zeros_like(k) +# tv = torch.zeros_like(v) +# sm_scale = 1.3 +# fn = lambda: attention(q, k, v, tq, tk, tv, causal, sm_scale)[0] +# if mode == "bwd": +# o = fn() +# do = torch.randn_like(o) +# fn = lambda: o.backward(do, retain_graph=True) +# ms = triton.testing.do_bench(fn) +# if provider == "flash": +# from flash_attn.flash_attn_interface import flash_attn_func + +# q = torch.randn((BATCH, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=device, requires_grad=True) +# k = torch.randn((BATCH, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=device, requires_grad=True) +# v = torch.randn((BATCH, H, SEQ_LEN, HEAD_DIM), dtype=dtype, device=device, requires_grad=True) +# fn = lambda: flash_attn_func(q.transpose(1, 2), k.transpose(1, 2), v.transpose(1, 2), causal=causal) +# if mode == "bwd": +# o = fn() +# do = torch.randn_like(o) +# fn = lambda: o.backward(do, retain_graph=True) +# ms = triton.testing.do_bench(fn) +# # there are 2 matmuls in the forward pass +# flops_per_matmul = 2.0 * BATCH * H * SEQ_LEN * SEQ_LEN * HEAD_DIM +# total_flops = 2 * flops_per_matmul +# if causal: +# total_flops *= 0.5 +# if mode == "bwd": +# # there are 5 matmuls in the backward pass +# total_flops *= 2.5 # 2.0(bwd) + 0.5(recompute) +# elif "triton" in provider: +# # there are 6 matmuls in the forward pass with JVP computation +# total_flops *= 3 +# return total_flops * 1e-12 / (ms * 1e-3) + + +# def _test_fwd_bwd_ca(Z, H, SEQ_LEN_Q, SEQ_LEN_KV, HEAD_DIM, dtype=torch.float16): +# torch.manual_seed(20) +# q = torch.empty((Z, H, SEQ_LEN_Q, HEAD_DIM), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5).requires_grad_() +# k = ( +# torch.empty((Z, H, SEQ_LEN_KV, HEAD_DIM), dtype=dtype, device=DEVICE) +# .normal_(mean=0.0, std=0.5) +# .requires_grad_() +# ) +# v = ( +# torch.empty((Z, H, SEQ_LEN_KV, HEAD_DIM), dtype=dtype, device=DEVICE) +# .normal_(mean=0.0, std=0.5) +# .requires_grad_() +# ) +# tq = torch.zeros_like(q) +# tk = torch.zeros_like(k) +# tv = torch.zeros_like(v) +# sm_scale = 0.5 +# dout = torch.randn((Z, H, SEQ_LEN_Q, HEAD_DIM), device=q.device, dtype=q.dtype) +# # reference implementation +# p = torch.matmul(q, k.transpose(2, 3)) * sm_scale +# p = torch.softmax(p.float(), dim=-1).to(dtype) +# ref_out = torch.matmul(p, v) +# ref_out.backward(dout) +# ref_dv, v.grad = v.grad.clone(), None +# ref_dk, k.grad = k.grad.clone(), None +# ref_dq, q.grad = q.grad.clone(), None +# # triton implementation +# tri_out = attention(q, k, v, tq, tk, tv, False, sm_scale)[0].to(dtype) +# tri_out.backward(dout) +# tri_dv, v.grad = v.grad.clone(), None +# tri_dk, k.grad = k.grad.clone(), None +# tri_dq, q.grad = q.grad.clone(), None +# # compare +# atol = 2e-2 if dtype == torch.bfloat16 else 1e-2 +# rtol = 2e-2 if dtype == torch.bfloat16 else 0 +# torch.testing.assert_close(ref_out, tri_out, atol=1e-2, rtol=0) +# torch.testing.assert_close(ref_dq, tri_dq, atol=atol, rtol=rtol / 2) +# torch.testing.assert_close(ref_dk, tri_dk, atol=atol, rtol=rtol / 2) +# torch.testing.assert_close(ref_dv, tri_dv, atol=atol, rtol=rtol) + + +# def test_fwd_bwd_ca(): +# for shape in [(1, 2, 256, 1024, 128), (1, 2, 1024, 256, 128), (1, 2, 1024, 512, 64), (1, 2, 1000, 515, 64)]: +# for dtype in [torch.float16, torch.bfloat16]: +# _test_fwd_bwd_ca(*shape, dtype) +# print(f"Shape={shape}, Dtype={dtype} Passed (CA fwd/bwd with the same headdim).") + + +# def _test_jvp_ca(Z, H, SEQ_LEN_Q, SEQ_LEN_KV, HEAD_DIM_QK, HEAD_DIM_V, dtype=torch.float16): +# torch.manual_seed(20) +# q = ( +# torch.empty((Z, H, SEQ_LEN_Q, HEAD_DIM_QK), dtype=dtype, device=DEVICE) +# .normal_(mean=0.0, std=0.5) +# .requires_grad_() +# ) +# k = ( +# torch.empty((Z, H, SEQ_LEN_KV, HEAD_DIM_QK), dtype=dtype, device=DEVICE) +# .normal_(mean=0.0, std=0.5) +# .requires_grad_() +# ) +# v = ( +# torch.empty((Z, H, SEQ_LEN_KV, HEAD_DIM_V), dtype=dtype, device=DEVICE) +# .normal_(mean=0.0, std=0.5) +# .requires_grad_() +# ) +# tq = torch.empty((Z, H, SEQ_LEN_Q, HEAD_DIM_QK), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5) +# tk = torch.empty((Z, H, SEQ_LEN_KV, HEAD_DIM_QK), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5) +# tv = torch.empty((Z, H, SEQ_LEN_KV, HEAD_DIM_V), dtype=dtype, device=DEVICE).normal_(mean=0.0, std=0.5) +# sm_scale = 0.5 + +# def naive_attention(q, k, v): +# # reference implementation +# p = torch.matmul(q, k.transpose(2, 3)) * sm_scale +# p = torch.softmax(p.float(), dim=-1).to(dtype) +# ref_out = torch.matmul(p, v) +# return ref_out + +# ref_out, ref_tout = torch.func.jvp(naive_attention, (q, k, v), (tq, tk, tv)) +# # triton implementation +# tri_out, tri_tout = attention(q, k, v, tq, tk, tv, False, sm_scale) +# # compare +# atol = 2e-2 if dtype == torch.bfloat16 else 1e-2 +# torch.testing.assert_close(ref_out, tri_out, atol=1e-2, rtol=0) +# torch.testing.assert_close(ref_tout, tri_tout, atol=atol, rtol=1e-2) + + +# def test_jvp_ca(): +# for shape in [ +# (1, 2, 256, 1024, 64, 128), +# (1, 2, 1000, 15, 128, 32), +# (1, 2, 512, 512, 16, 32), +# (1, 2, 515, 999, 16, 32), +# ]: +# for dtype in [torch.float16, torch.bfloat16]: +# _test_jvp_ca(*shape, dtype) +# print(f"Shape={shape}, Dtype={dtype} Passed (CA fwd/JVP with different headdim).") + + +# if __name__ == "__main__": +# test_fwd_bwd() +# test_jvp() +# # only works on post-Ampere GPUs right now +# bench_flash_attention.run(save_path=".", print_data=True) +# test_fwd_bwd_ca() +# test_jvp_ca() diff --git a/REGEN-main/cosmos_policy/_src/predict2/utils/fused_adam_dtensor.py b/REGEN-main/cosmos_policy/_src/predict2/utils/fused_adam_dtensor.py new file mode 100644 index 0000000000000000000000000000000000000000..32b9b15f868cb12065d610aeb3164559c2295fc1 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/utils/fused_adam_dtensor.py @@ -0,0 +1,393 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import torch +import transformer_engine as te +import transformer_engine_torch as tex + +from cosmos_policy._src.imaginaire.utils import distributed, log +from cosmos_policy._src.imaginaire.utils.misc import get_local_tensor_if_DTensor + + +class FusedAdam(torch.optim.Optimizer): + """Implements Adam algorithm. + + Currently GPU-only. Requires Apex to be installed via + ``pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./``. + + This version of fused Adam implements 2 fusions. + + * Fusion of the Adam update's elementwise operations + * A multi-tensor apply launch that batches the elementwise updates applied to all the model's parameters + into one or a few kernel launches. + + :class:`FusedAdam` may be used as a drop-in replacement for ``torch.optim.AdamW``, + or ``torch.optim.Adam`` with ``adam_w_mode=False``:: + + opt = FusedAdam(model.parameters(), lr = ....) + ... + opt.step() + + .. warning:: + A previous version of :class:`FusedAdam` allowed a number of additional arguments to ``step``. + These additional arguments are now deprecated and unnecessary. + + Adam was been proposed in `Adam: A Method for Stochastic Optimization`_. + + Arguments: + params (iterable): iterable of parameters to optimize or dicts defining + parameter groups. + lr (float, optional): learning rate. (default: 1e-3) + betas (Tuple[float, float], optional): coefficients used for computing + running averages of gradient and its square. (default: (0.9, 0.999)) + eps (float, optional): term added to the denominator to improve + numerical stability. (default: 1e-8) + weight_decay (float, optional): weight decay (L2 penalty) (default: 0) + amsgrad (boolean, optional): whether to use the AMSGrad variant of this + algorithm from the paper `On the Convergence of Adam and Beyond`_ + (default: False) NOT SUPPORTED in FusedAdam! + adam_w_mode (boolean, optional): Apply L2 regularization or weight decay + True for decoupled weight decay(also known as AdamW) (default: True) + capturable (bool, optional): whether to use the version of the optimizer + that can be used with CUDA Graphs. (default: False) + master_weights (bool, optional): whether to maintain FP32 master weights + in the optimizer with FP16 mixed precision training, currently can + only be used with capturable set to True. (default: False) + + .. _Adam - A Method for Stochastic Optimization: + https://arxiv.org/abs/1412.6980 + .. _On the Convergence of Adam and Beyond: + https://openreview.net/forum?id=ryQu7f-RZ + """ + + def __init__( + self, + params, + lr=1e-3, + bias_correction=True, + betas=(0.9, 0.999), + eps=1e-8, + adam_w_mode=True, + weight_decay=0.0, + amsgrad=False, + capturable=False, + master_weights=False, + ): + if amsgrad: + raise RuntimeError("FusedAdam does not support the AMSGrad variant.") + if master_weights and not capturable: + raise RuntimeError("Master weights is currently only supported with the capturable version.") + # If the optimizer is capturable then LR should be a tensor (on GPU) + log.warning(f"FusedAdam master_weights: {master_weights} capturable: {capturable}") + lr = torch.tensor(lr, dtype=torch.float32) if capturable else lr + defaults = dict(lr=lr, bias_correction=bias_correction, betas=betas, eps=eps, weight_decay=weight_decay) + super(FusedAdam, self).__init__(params, defaults) + self.adam_w_mode = 1 if adam_w_mode else 0 + + self.capturable = capturable + self.master_weights = master_weights + + self.param_groups_master = None + + if capturable: + for idx, group in enumerate(self.param_groups): + if len(group["params"]) == 0: + continue + device = group["params"][0].device + for item in ["lr"]: + if isinstance(group[item], float): + group[item] = torch.tensor(group[item], dtype=torch.float32) + self.param_groups[idx][item] = group[item].to(device=device) + + self._step_supports_amp_scaling = True + + # Skip buffer + self._dummy_overflow_buf = torch.tensor([0], dtype=torch.int, device="cuda") + self.multi_tensor_adam = tex.multi_tensor_adam + self.multi_tensor_adam_capturable = tex.multi_tensor_adam_capturable + self.multi_tensor_adam_capturable_master = tex.multi_tensor_adam_capturable_master + + def step(self, closure=None, grads=None, output_params=None, scale=None, grad_norms=None, grad_scaler=None): + """Performs a single optimization step. + + Arguments: + closure (callable, optional): A closure that reevaluates the model + and returns the loss. + + The remaining arguments are deprecated, and are only retained (for the moment) for error-checking purposes. + """ + if any(p is not None for p in [grads, output_params, scale, grad_norms]): + raise RuntimeError( + "FusedAdam has been updated. " + "Simply initialize it identically to torch.optim.Adam, and call step() with no arguments." + ) + loss = None + if closure is not None: + loss = closure() + + if self.param_groups_master is None: + # Create full precision master weights + self.param_groups_master = [] + for i, pg in enumerate(self.param_groups): + param_list = pg["params"] + self.param_groups_master.append( + { + # Change related to master weights + "params": [ + get_local_tensor_if_DTensor(p).clone().detach().float() if self.master_weights else None + for p in param_list + ], + } + ) + + for group, group_master in zip(self.param_groups, self.param_groups_master): + if len(group["params"]) == 0: + continue + device = group["params"][0].device + bias_correction = 1 if "bias_correction" in group and group["bias_correction"] else 0 + beta1, beta2 = group["betas"] + + # assume same step across group now to simplify things + # per parameter step can be easily support by making it tensor, or pass list into kernel + if "step" in group: + if self.capturable: + group["step"] = ( + group["step"].to(device=device) + if isinstance(group["step"], torch.Tensor) + else torch.tensor(group["step"], dtype=torch.int32, device=device) + ) + group["step"] += (self._dummy_overflow_buf != 1).to(torch.int) + else: + group["step"] += 1 + else: + group["step"] = 1 if not self.capturable else torch.tensor([1], dtype=torch.int, device=device) + + if self.capturable: + group["lr"] = ( + group["lr"].to(device=device) + if isinstance(group["lr"], torch.Tensor) + else torch.tensor(group["lr"], dtype=torch.float32, device=device) + ) + + # create lists for multi-tensor apply + g_16, p_16, m_16, v_16 = [], [], [], [] + g_bf, p_bf, m_bf, v_bf = [], [], [], [] + g_32, p_32, m_32, v_32 = [], [], [], [] + p_16_master = [] + p_32_master = [] + bf16_master = [] + + for p, p_master in zip(group["params"], group_master["params"]): + if p.grad is None: + continue + if p.grad.data.is_sparse: + raise RuntimeError( + "FusedAdam does not support sparse gradients, please consider SparseAdam instead" + ) + + state = self.state[p] + # State initialization + if len(state) == 0: + # Exponential moving average of gradient values + # Change that makes .step() not crash + state["exp_avg"] = torch.zeros_like(get_local_tensor_if_DTensor(p).data).float() + # Exponential moving average of squared gradient values + # Change that makes .step() not crash + state["exp_avg_sq"] = torch.zeros_like(get_local_tensor_if_DTensor(p).data).float() + + if p.dtype == torch.float16: + if self.master_weights: + p_16_master.append(get_local_tensor_if_DTensor(p_master).data) + g_16.append(get_local_tensor_if_DTensor(p.grad)) + p_16.append(get_local_tensor_if_DTensor(p)) + m_16.append(state["exp_avg"]) + v_16.append(state["exp_avg_sq"]) + elif p.dtype == torch.bfloat16: + if self.master_weights: + # Change that makes .step() not crash + bf16_master.append(get_local_tensor_if_DTensor(p_master).data) + # Change that makes .step() not crash + g_bf.append(get_local_tensor_if_DTensor(p.grad)) + # Change that makes .step() not crash + p_bf.append(get_local_tensor_if_DTensor(p)) + m_bf.append(state["exp_avg"]) + v_bf.append(state["exp_avg_sq"]) + elif p.dtype == torch.float32: + if self.master_weights: + p_32_master.append(p_master.data) + g_32.append(p.grad.data) + p_32.append(p.data) + m_32.append(state["exp_avg"]) + v_32.append(state["exp_avg_sq"]) + else: + raise RuntimeError("FusedAdam only support fp16 and fp32.") + + # If the optimizer is capturable, then if there's a grad scaler it works + # on the GPU + a different multi_tensor_applier should be called + if self.capturable: + # overflow check of gradients + found_inf = ( + grad_scaler._check_inf_per_device(self)[device] + if grad_scaler is not None + else torch.zeros((1,), device=device) + ) + self._dummy_overflow_buf.copy_(found_inf) + + # get unscale scale factor + scale, inv_scale = None, None + if grad_scaler: + scale = grad_scaler._get_scale_async() + inv_scale = scale.double().reciprocal().float() + else: + scale = torch.ones((1,), device=device, dtype=torch.float32) + inv_scale = torch.ones((1,), device=device, dtype=torch.float32) + + if len(g_16) > 0: + te.pytorch.optimizers.multi_tensor_applier( + ( + self.multi_tensor_adam_capturable_master + if self.master_weights + else self.multi_tensor_adam_capturable + ), + self._dummy_overflow_buf, + [g_16, p_16, m_16, v_16, p_16_master] if self.master_weights else [g_16, p_16, m_16, v_16], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + inv_scale, + ) + + if len(g_bf) > 0: + te.pytorch.optimizers.multi_tensor_applier( + ( + self.multi_tensor_adam_capturable_master + if self.master_weights + else self.multi_tensor_adam_capturable + ), + self._dummy_overflow_buf, + [g_bf, p_bf, m_bf, v_bf, bf16_master] if self.master_weights else [g_bf, p_bf, m_bf, v_bf], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + inv_scale, + ) + + if len(g_32) > 0: + te.pytorch.optimizers.multi_tensor_applier( + ( + self.multi_tensor_adam_capturable_master + if self.master_weights + else self.multi_tensor_adam_capturable + ), + self._dummy_overflow_buf, + [g_32, p_32, m_32, v_32, p_32_master] if self.master_weights else [g_32, p_32, m_32, v_32], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + inv_scale, + ) + else: + if len(g_16) > 0: + te.pytorch.optimizers.multi_tensor_applier( + self.multi_tensor_adam, + self._dummy_overflow_buf, + [g_16, p_16, m_16, v_16], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + ) + + if len(g_bf) > 0: + te.pytorch.optimizers.multi_tensor_applier( + self.multi_tensor_adam, + self._dummy_overflow_buf, + [g_bf, p_bf, m_bf, v_bf], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + ) + + if len(g_32) > 0: + te.pytorch.optimizers.multi_tensor_applier( + self.multi_tensor_adam, + self._dummy_overflow_buf, + [g_32, p_32, m_32, v_32], + group["lr"], + beta1, + beta2, + group["eps"], + group["step"], + self.adam_w_mode, + bias_correction, + group["weight_decay"], + ) + + return loss + + def load_state_dict(self, state_dict): + super().load_state_dict(state_dict) + for group in self.param_groups: + if self.capturable: + group["lr"] = ( + group["lr"].cuda() + if isinstance(group["lr"], torch.Tensor) + else torch.tensor(group["lr"], dtype=torch.float32).cuda() + ) + + if "step" in group: + if self.capturable: + if distributed.get_rank() == 0: + step = ( + group["step"].cuda() + if isinstance(group["step"], torch.Tensor) + else torch.tensor([group["step"]], dtype=torch.int32).cuda() + ) + else: + step = torch.zeros(1, dtype=torch.int32).cuda() + # make it compatible with FSDP optimizer + distributed.broadcast(step, 0) + group["step"] = step + elif isinstance(group["step"], torch.Tensor): + group["step"] = group["step"].item() + for p in group["params"]: + state = self.state[p] + if "exp_avg" in state: + state["exp_avg"] = state["exp_avg"].float() + state["exp_avg_sq"] = state["exp_avg_sq"].float() diff --git a/REGEN-main/cosmos_policy/_src/predict2/utils/kv_cache.py b/REGEN-main/cosmos_policy/_src/predict2/utils/kv_cache.py new file mode 100644 index 0000000000000000000000000000000000000000..daa9a5bcf315f939274eb1a549b2ceb981780ebe --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/utils/kv_cache.py @@ -0,0 +1,169 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from __future__ import annotations + +from dataclasses import dataclass +from typing import Any, Optional + +import torch +import torch.nn as nn + + +@dataclass +class KVCacheConfig: + run_with_kv: bool = False + store_kv: bool = False + current_idx: int = 0 + recompute_cross_attn_kv: bool = False + + +class AttentionOpWithKVCache(nn.Module): + """A thin wrapper that adds K/V caching to an existing attention op. + + This wrapper expects the wrapped op to accept (q, k, v, attn_mask=None) + and return attention outputs with heads already flattened on the last dim. + + Cache semantics: + - Cache entries are stored as per-chunk tensors, where each chunk corresponds + to one latent frame composed of HxW tokens (after patchify). + - The `max_cache_size` capacity therefore refers to the number of latent + frames (chunks), NOT the number of individual tokens. + - When `max_cache_size` is None, the cache grows without an automatic + rolling window; otherwise, it acts as a rolling window of at most + `max_cache_size` frames. Upon overflow, the oldest frames are dropped. + """ + + def __init__(self, attn_op: nn.Module | Any, max_cache_size: Optional[int] = None): + """Initialize the KV cache wrapper. + + Args: + attn_op: The underlying attention operation (q, k, v[, attn_mask]) -> out. + max_cache_size: Optional capacity measured in number of latent frames + (chunks). Each chunk is a single frame worth of HxW tokens. If None, + the cache does not enforce a rolling capacity. + """ + super().__init__() + self.attn_op = attn_op + self.reset_kv_cache(max_cache_size=max_cache_size) + self.pg: Optional[Any] = None + self.stream: Optional[Any] = None + + def reset_kv_cache(self, max_cache_size: Optional[int] = None) -> None: + """Reset/initialize the KV caches. + + Args: + max_cache_size: Optional capacity measured in number of latent frames + (chunks). Each chunk is a single frame worth of HxW tokens. If None, + the cache does not enforce a rolling capacity. + """ + # Initialize list-based caches and optionally set capacity in chunks + self.start_idx = 0 + self.k_cache: list[torch.Tensor | None] = [None] * (max_cache_size or 99999) + self.v_cache: list[torch.Tensor | None] = [None] * (max_cache_size or 99999) + self.max_cache_size = max_cache_size + + def forward( + self, + q: torch.Tensor, + k: torch.Tensor, + v: torch.Tensor, + *, + kv_cache_cfg: KVCacheConfig, + **kwargs, + ) -> torch.Tensor: + assert self.k_cache is not None and self.v_cache is not None, ( + "KV cache is not initialized. Call reset_kv_cache() first." + ) + + # Store into cache at start_idx location (list-based) + if kv_cache_cfg.store_kv: + index = int(kv_cache_cfg.current_idx) + self.k_cache[index] = k.detach() + self.v_cache[index] = v.detach() + + # Prepend cached prefix up to start_idx (list-based) + if kv_cache_cfg.run_with_kv and kv_cache_cfg.current_idx > 0: + history_k = self.k_cache[self.start_idx : kv_cache_cfg.current_idx] + history_v = self.v_cache[self.start_idx : kv_cache_cfg.current_idx] + assert not any(x is None for x in history_k) + assert not any(x is None for x in history_v) + k_out = torch.cat(history_k + [k], dim=1) # type: ignore + v_out = torch.cat(history_v + [v], dim=1) # type: ignore + else: + k_out = k + v_out = v + + # Enforce rolling capacity in number of cached chunks (frames) + if kv_cache_cfg.run_with_kv and self.max_cache_size is not None: + # Instead of deleting, just update start_idx for rolling window + self.start_idx = max(0, int(kv_cache_cfg.current_idx) - self.max_cache_size) + + return self.attn_op(q, k_out, v_out, **kwargs) + + def set_context_parallel_group(self, process_group, ranks, stream, cp_comm_type: str = "p2p"): + self.attn_op.set_context_parallel_group(process_group, ranks, stream, cp_comm_type=cp_comm_type) # type: ignore + + +class VideoSeqPos: + """Flattened 3D grid positions for a video clip. + + Stores flattened t/h/w indices of length L = T*H*W to enable constructing + RoPE frequencies aligned with global positions across sequential chunks. + """ + + def __init__(self, T: int, H: int, W: int, pos_h=None, pos_w=None, pos_t=None) -> None: + self.T = T + self.H = H + self.W = W + + if pos_h is not None and pos_w is not None and pos_t is not None: + self.pos_h = pos_h.to(dtype=torch.long) + self.pos_w = pos_w.to(dtype=torch.long) + self.pos_t = pos_t.to(dtype=torch.long) + return + + device = torch.device("cuda", torch.cuda.current_device()) if torch.cuda.is_available() else torch.device("cpu") + t = torch.arange(self.T, device=device, dtype=torch.long) + h = torch.arange(self.H, device=device, dtype=torch.long) + w = torch.arange(self.W, device=device, dtype=torch.long) + pos_t, pos_h, pos_w = torch.meshgrid(t, h, w, indexing="ij") + self.pos_t = pos_t.reshape(-1) + self.pos_h = pos_h.reshape(-1) + self.pos_w = pos_w.reshape(-1) + + def size(self) -> int: + return int(self.pos_h.numel()) + + def frame(self, t_idx: int) -> "VideoSeqPos": + """Return a `VideoSeqPos` view for a single frame at absolute index `t_idx`. + + This is useful for streaming / KV-cache inference where the model is run on + one frame at a time but RoPE positions must reflect global video indices. + """ + t_idx = int(t_idx) + if t_idx < 0 or t_idx >= int(self.T): + raise IndexError(f"t_idx out of range: {t_idx} (valid: [0, {self.T}))") + tokens_per_frame = int(self.H) * int(self.W) + start = t_idx * tokens_per_frame + end = start + tokens_per_frame + return VideoSeqPos( + T=1, + H=int(self.H), + W=int(self.W), + pos_h=self.pos_h[start:end], + pos_w=self.pos_w[start:end], + pos_t=self.pos_t[start:end], + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/utils/model_comp.py b/REGEN-main/cosmos_policy/_src/predict2/utils/model_comp.py new file mode 100644 index 0000000000000000000000000000000000000000..cbd3ff782d536d27978b51a91300334378e2b69c --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/utils/model_comp.py @@ -0,0 +1,133 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import torch + + +def compare_models_thoroughly(model1, model2, verbose=True): + """ + Thoroughly compare two models by checking all parameters and buffers. + + Args: + model1: First PyTorch model + model2: Second PyTorch model + verbose: If True, prints detailed comparison information + + Returns: + dict: Comparison results containing: + - mismatched_params: List of parameter names that don't match + - mismatched_buffers: List of buffer names that don't match + - is_identical: Boolean indicating if models are identical + """ + + def print_if_verbose(*args, **kwargs): + if verbose: + print(*args, **kwargs) + + mismatches = {"mismatched_params": [], "mismatched_buffers": [], "is_identical": True} + + # Compare parameters + print_if_verbose("\n=== Comparing Parameters ===") + params1 = dict(model1.named_parameters()) + params2 = dict(model2.named_parameters()) + + # Check parameter names + param_names1 = set(params1.keys()) + param_names2 = set(params2.keys()) + if param_names1 != param_names2: + mismatches["is_identical"] = False + extra_in_1 = param_names1 - param_names2 + extra_in_2 = param_names2 - param_names1 + if extra_in_1: + print_if_verbose(f"Parameters only in model1: {extra_in_1}") + mismatches["mismatched_params"].extend(list(extra_in_1)) + if extra_in_2: + print_if_verbose(f"Parameters only in model2: {extra_in_2}") + mismatches["mismatched_params"].extend(list(extra_in_2)) + + # Compare common parameters + common_params = param_names1 & param_names2 + for name in common_params: + param1 = params1[name] + param2 = params2[name] + + # Compare shapes + if param1.shape != param2.shape: + mismatches["is_identical"] = False + mismatches["mismatched_params"].append(name) + print_if_verbose(f"Shape mismatch for parameter {name}:") + print_if_verbose(f" Model1: {param1.shape}") + print_if_verbose(f" Model2: {param2.shape}") + continue + + # Compare values + if not torch.equal(param1.cpu(), param2.cpu()): + mismatches["is_identical"] = False + mismatches["mismatched_params"].append(name) + print_if_verbose(f"Value mismatch for parameter {name}") + + # Compare all buffers (both persistent and non-persistent) + print_if_verbose("\n=== Comparing Buffers ===") + buffers1 = dict(model1.named_buffers()) + buffers2 = dict(model2.named_buffers()) + + # Check buffer names + buffer_names1 = set(buffers1.keys()) + buffer_names2 = set(buffers2.keys()) + if buffer_names1 != buffer_names2: + mismatches["is_identical"] = False + extra_in_1 = buffer_names1 - buffer_names2 + extra_in_2 = buffer_names2 - buffer_names1 + if extra_in_1: + print_if_verbose(f"Buffers only in model1: {extra_in_1}") + mismatches["mismatched_buffers"].extend(list(extra_in_1)) + if extra_in_2: + print_if_verbose(f"Buffers only in model2: {extra_in_2}") + mismatches["mismatched_buffers"].extend(list(extra_in_2)) + + # Compare common buffers + common_buffers = buffer_names1 & buffer_names2 + for name in common_buffers: + buf1 = buffers1[name] + buf2 = buffers2[name] + + # Compare shapes + if buf1.shape != buf2.shape: + mismatches["is_identical"] = False + mismatches["mismatched_buffers"].append(name) + print_if_verbose(f"Shape mismatch for buffer {name}:") + print_if_verbose(f" Model1: {buf1.shape}") + print_if_verbose(f" Model2: {buf2.shape}") + continue + + # Compare values + try: + if not torch.equal(buf1.cpu(), buf2.cpu()): + mismatches["is_identical"] = False + mismatches["mismatched_buffers"].append(name) + print_if_verbose(f"Value mismatch for buffer {name}") + except RuntimeError as e: + print_if_verbose(f"Error comparing buffer {name}: {e}") + mismatches["mismatched_buffers"].append(name) + + # Print summary + print_if_verbose("\n=== Summary ===") + print_if_verbose(f"Total parameters checked: {len(common_params)}") + print_if_verbose(f"Total buffers checked: {len(common_buffers)}") + print_if_verbose(f"Mismatched parameters: {len(mismatches['mismatched_params'])}") + print_if_verbose(f"Mismatched buffers: {len(mismatches['mismatched_buffers'])}") + print_if_verbose(f"Models are {'identical' if mismatches['is_identical'] else 'different'}") + + return mismatches diff --git a/REGEN-main/cosmos_policy/_src/predict2/utils/model_loader.py b/REGEN-main/cosmos_policy/_src/predict2/utils/model_loader.py new file mode 100644 index 0000000000000000000000000000000000000000..f90c96b857f1f7bbbd49ec71a8b2f88798f48f68 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/utils/model_loader.py @@ -0,0 +1,405 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import importlib +import os +from typing import Optional + +import torch +from peft import LoraConfig, set_peft_model_state_dict + +from cosmos_policy._src.imaginaire.config import Config +from cosmos_policy._src.imaginaire.flags import INTERNAL, SMOKE +from cosmos_policy._src.imaginaire.lazy_config import instantiate +from cosmos_policy._src.imaginaire.model import ImaginaireModel +from cosmos_policy._src.imaginaire.utils import distributed, log, misc +from cosmos_policy._src.imaginaire.utils.config_helper import get_config_module, override +from cosmos_policy._src.imaginaire.utils.easy_io import easy_io +from cosmos_policy._src.imaginaire.utils.fsdp_helper import hsdp_device_mesh +from cosmos_policy._src.predict2.checkpointer.dcp import ( + DefaultLoadPlanner, + DistributedCheckpointer, + ModelWrapper, + dcp_load_state_dict, +) + + +def load_model_from_checkpoint( + experiment_name, + s3_checkpoint_dir, + config_file="cosmos_policy/_src/predict2/configs/video2world/config.py", + enable_fsdp=False, + load_ema_to_reg=False, + instantiate_ema=True, + seed=0, + local_cache_dir=None, + override_cache: bool = False, + experiment_opts: Optional[list[str]] = None, + skip_load_model: bool = False, + adapter_checkpoint_paths: Optional[list[str]] = None, + cache_text_encoder: bool = False, + to_device: Optional[str] = None, +): + """ + Load model from checkpoint with optional multi-adapter support. + + Args: + experiment_name: experiment name + s3_checkpoint_dir: s3 path to iteration_model + config_file: config file path + enable_fsdp: enable fsdp + load_ema_to_reg: load ema as regular model + instantiate_ema: whether to instantiate EMA + seed: random seed + local_cache_dir: local cache directory, if None, do not cache + override_cache: override cache, if True, override cache if local cache exists + experiment_opts: experiment options + skip_load_model: skip loading model weights + adapter_checkpoint_paths: list of checkpoint paths for loading multiple adapters + Supports both .pt and DCP checkpoint formats (auto-detected by file extension). + Example: + adapter_checkpoint_paths=[ + "s3://bucket/exp1/checkpoints/model.pt", # .pt format + "s3://bucket/exp2/checkpoints" # DCP format + ] + cache_text_encoder: cache text encoder, if True, cache text encoder. This is default to False to avoid race condition if multiple nodes are running inference concurrently (e.g., running inference pipeline). + + Returns: + model: loaded model + config: config object + """ + if experiment_opts is None: + experiment_opts = [] + config_module = get_config_module(config_file) + config = importlib.import_module(config_module).make_config() + config = override(config, ["--", f"experiment={experiment_name}"] + experiment_opts) + + # Override checkpoint path if provided + if s3_checkpoint_dir: + log.info(f"Overriding config checkpoint path with: {s3_checkpoint_dir}") + config.checkpoint.load_path = str(s3_checkpoint_dir) + + if load_ema_to_reg: + config.model.config.ema.enabled = False + + if instantiate_ema is False and config.model.config.ema.enabled: + config.model.config.ema.enabled = False + + # Check that the config is valid + config.validate() + # Freeze the config so developers don't change it during training. + config.freeze() # type: ignore + misc.set_random_seed(seed=seed, by_rank=True) + # Initialize cuDNN. + torch.backends.cudnn.deterministic = config.trainer.cudnn.deterministic + torch.backends.cudnn.benchmark = config.trainer.cudnn.benchmark + # Floating-point precision settings. + torch.backends.cudnn.allow_tf32 = torch.backends.cuda.matmul.allow_tf32 = True + + log.info(f"Loading model from {s3_checkpoint_dir}") + + if not enable_fsdp: + # disable fsdp + config.model.config.fsdp_shard_size = 1 + with misc.timer("instantiate model"): + model = instantiate(config.model) + if to_device is not None: + model.to(torch.device(to_device)) + # Convert the model parameters to bf16 + model.on_train_start() + + if not skip_load_model: + # Handle different adapter loading scenarios + if adapter_checkpoint_paths: + # First load base model + model = load_model_state_dict_from_checkpoint( + model, config, s3_checkpoint_dir, load_ema_to_reg, local_cache_dir, override_cache + ) + # Then load additional adapters from different checkpoints + log.info(f"Loading {len(adapter_checkpoint_paths)} adapters from different checkpoints") + adapter_names = [f"adapter_{i}" for i in range(len(adapter_checkpoint_paths))] + + for adapter_name, checkpoint_path in zip(adapter_names, adapter_checkpoint_paths): + log.info(f"Loading adapter '{adapter_name}' from {checkpoint_path}") + lora_config = LoraConfig( + r=model.config.lora_rank, + lora_alpha=model.config.lora_alpha, + init_lora_weights=model.config.init_lora_weights, + target_modules=[module.strip() for module in model.config.lora_target_modules.split(",")], + use_dora=model.config.use_dora, + ) + model.net.add_adapter(adapter_name, lora_config) + + if checkpoint_path.endswith(".pt"): + # adapter_state_dict = easy_io.load(checkpoint_path) + adapter_state_dict = torch.load(checkpoint_path, map_location="cpu") + old_keys = list(adapter_state_dict.keys()) + for key in old_keys: + if "lora_" in key: + net_prefix = "net." if load_ema_to_reg else "net_ema." + new_key = key.replace(net_prefix, "base_model.model.").replace("default.", "") + adapter_state_dict[new_key] = adapter_state_dict.pop(key) + load_result = set_peft_model_state_dict(model.net, adapter_state_dict, adapter_name=adapter_name) + # for key in load_result.missing_keys: + # log.warning(f"Missing key: {key}") + for key in load_result.unexpected_keys: + log.warning(f"Unexpected key: {key}") + assert False, "Unexpected key found" + else: + log.info(f"Loading adapter '{adapter_name}' from s3 {checkpoint_path}") + if checkpoint_path.rstrip("/").endswith("/model"): + cur_key_ckpt_full_path = checkpoint_path + else: + cur_key_ckpt_full_path = os.path.join(checkpoint_path, "model") + + checkpointer = DistributedCheckpointer( + config.checkpoint, config.job, callbacks=None, disable_async=True + ) + + _model_wrapper = ModelWrapper(model, load_ema_to_reg=load_ema_to_reg) + mapping_keys = { + adapter_name + ".": "default.", + } + _state_dict = _model_wrapper.state_dict(mapping_keys=mapping_keys) + storage_reader = checkpointer.get_storage_reader(cur_key_ckpt_full_path) + load_planner = DefaultLoadPlanner(allow_partial_load=True) + dcp_load_state_dict(_state_dict, storage_reader, load_planner) + _model_wrapper.load_state_dict(_state_dict) + + log.info(f"Loaded adapter '{adapter_name}'") + + # Activate first adapter + model.net.set_adapter(adapter_names[0]) + log.info(f"Activated adapter: {adapter_names[0]}") + else: + # Load normally (single checkpoint) + model = load_model_state_dict_from_checkpoint( + model, config, s3_checkpoint_dir, load_ema_to_reg, local_cache_dir, override_cache + ) + + return model, config + + +def load_model_state_dict_from_checkpoint( + model, + config, + s3_checkpoint_dir, + load_ema_to_reg=False, + local_cache_dir=None, + override_cache: bool = False, +): + if s3_checkpoint_dir is not None: + s3_checkpoint_dir = str(s3_checkpoint_dir) + checkpoint_format = "pt" if s3_checkpoint_dir.endswith(".pt") else "dcp" + if s3_checkpoint_dir.startswith("s3:"): + if checkpoint_format == "pt": + cur_key_ckpt_full_path = s3_checkpoint_dir + elif s3_checkpoint_dir.rstrip("/").endswith("/model"): + cur_key_ckpt_full_path = s3_checkpoint_dir + else: + cur_key_ckpt_full_path = os.path.join(s3_checkpoint_dir, "model") + else: + cur_key_ckpt_full_path = s3_checkpoint_dir + + from cosmos_policy._src.imaginaire.utils.checkpoint_db import get_checkpoint_path + + load_from_local = True + local_s3_ckpt_fp = get_checkpoint_path(cur_key_ckpt_full_path) + + if SMOKE: + return model + + if load_from_local: + # Load on rank0 only and broadcast + if distributed.is_rank0(): + log.info(f"Loading model cached locally from {local_s3_ckpt_fp}") + if checkpoint_format == "dcp": + # Local training checkpoints are DCP directories + # (``.metadata`` plus one or more ``*.distcp`` shards). + # ``easy_io.load`` only supports single-file checkpoints. + checkpointer = DistributedCheckpointer( + config.checkpoint, + config.job, + callbacks=None, + disable_async=True, + ) + model_wrapper = ModelWrapper(model, load_ema_to_reg=load_ema_to_reg) + dcp_state_dict = model_wrapper.state_dict() + storage_reader = checkpointer.get_storage_reader(local_s3_ckpt_fp) + load_planner = DefaultLoadPlanner(allow_partial_load=True) + dcp_load_state_dict(dcp_state_dict, storage_reader, load_planner) + model_wrapper.load_state_dict(dcp_state_dict) + else: + local_state_dict = easy_io.load(local_s3_ckpt_fp, weights_only=INTERNAL) + + # Handle LoRA key mapping if the model uses LoRA and checkpoint is in .pt format + if hasattr(model, "config") and hasattr(model.config, "use_lora") and model.config.use_lora: + log.info("Model uses LoRA, mapping checkpoint keys to model keys with base_layer...") + mapped_state_dict = {} + mapped_keys = [] + missing_keys = [] + + # Get current model state dict to understand what keys are expected + model_state_dict = model.state_dict() + + for model_key in model_state_dict.keys(): + if "base_layer." in model_key or "base_model.model." in model_key: + # This is a LoRA layer - map from checkpoint key (without base_layer) + checkpoint_key = model_key.replace("base_layer.", "").replace("base_model.model.", "") + if checkpoint_key in local_state_dict: + mapped_state_dict[model_key] = local_state_dict[checkpoint_key] + mapped_keys.append(f"{checkpoint_key} -> {model_key}") + else: + missing_keys.append(model_key) + elif model_key in local_state_dict: + # Direct mapping for non-LoRA keys + mapped_state_dict[model_key] = local_state_dict[model_key] + else: + missing_keys.append(model_key) + + if mapped_keys: + log.info(f"Mapped {len(mapped_keys)} LoRA keys from checkpoint to model (showing first 5):") + for mapped_key in mapped_keys[:5]: + log.info(f" {mapped_key}") + if missing_keys: + log.warning(f"Missing keys in checkpoint: {missing_keys[:10]}... (showing first 10)") + + local_state_dict = mapped_state_dict + + # `strict=False` is needed to avoid errors: `Skipping key ... introduced by TransformerEngine for FP8 in the checkpoint.` + model.load_state_dict(local_state_dict, strict=False) + + # Synchronize model states from rank 0 to all other ranks + # Skip EMA parameters and buffers to avoid OOM - they are on CPU now, and will be moved to CUDA and synced via copy from main model after FSDP + params_and_buffers_to_ignore = set() + if hasattr(model, "net_ema") and model.net_ema is not None: + # Add all parameters + for param_name, _ in model.net_ema.named_parameters(): + params_and_buffers_to_ignore.add(f"net_ema.{param_name}") + # Add all buffers (e.g., running_mean, running_var in BatchNorm) + for buffer_name, _ in model.net_ema.named_buffers(): + params_and_buffers_to_ignore.add(f"net_ema.{buffer_name}") + log.info( + f"Skipping sync for {len(params_and_buffers_to_ignore)} EMA parameters and buffers to avoid OOM during initialization" + ) + + distributed.sync_model_states(model, src=0, params_and_buffers_to_ignore=params_and_buffers_to_ignore) + else: + log.info(f"Loading model from s3 {s3_checkpoint_dir}") + + checkpointer = DistributedCheckpointer(config.checkpoint, config.job, callbacks=None, disable_async=True) + + _model_wrapper = ModelWrapper( + model, + load_ema_to_reg=load_ema_to_reg if checkpoint_format == "dcp" else False, + ) + _state_dict = _model_wrapper.state_dict() + if checkpoint_format == "dcp": + storage_reader = checkpointer.get_storage_reader(cur_key_ckpt_full_path) + load_planner = DefaultLoadPlanner(allow_partial_load=True) + dcp_load_state_dict(_state_dict, storage_reader, load_planner) + _model_wrapper.load_state_dict(_state_dict) + else: # pt format - load on rank0 only and broadcast + if distributed.is_rank0(): + if "s3://" in s3_checkpoint_dir: + pt_state_dict = easy_io.load( + s3_checkpoint_dir, + backend_args={ + "backend": "s3", + "s3_credential_path": "credentials/s3_training.secret", + }, + ) + else: + pt_state_dict = easy_io.load(s3_checkpoint_dir) + # Handle different .pt checkpoint formats + if "model" in pt_state_dict: + # Checkpoint contains multiple components (model, optimizer, etc.) + model_state = pt_state_dict["model"] + elif "state_dict" in pt_state_dict: + # Alternative format + model_state = pt_state_dict["state_dict"] + else: + # Assume the checkpoint is the state dict itself + model_state = pt_state_dict + # Update the state dict with loaded weights + # Handle potential key mismatches + missing_keys = [] + unexpected_keys = [] + for key in _state_dict.keys(): + if key in model_state: + _state_dict[key] = model_state[key] + else: + missing_keys.append(key) + + for key in model_state.keys(): + if key not in _state_dict: + unexpected_keys.append(key) + + if missing_keys: + log.warning(f"Missing keys in checkpoint: {missing_keys[:10]}... (showing first 10)") + if unexpected_keys: + log.warning(f"Unexpected keys in checkpoint: {unexpected_keys[:10]}... (showing first 10)") + + # only load on rank0 + _model_wrapper.load_state_dict(_state_dict) + + # Synchronize model states from rank 0 to all other ranks + distributed.sync_model_states(model, src=0) + + # Cache the model state dict only on rank0 to be consistent with loading + if local_cache_dir is not None and distributed.is_rank0(): + log.info(f"Caching model state dict to {local_s3_ckpt_fp}") + easy_io.dump(model.state_dict(), local_s3_ckpt_fp) + + # Clear unused reserved memory from fp32 + torch.cuda.empty_cache() + return model + + +def create_model_from_consolidated_checkpoint_with_fsdp(config: Config) -> ImaginaireModel: + """ + Instantiate a model, load weights from a consolidated checkpoint, and initialize FSDP if required. + + Args: + config: The configuration object for the experiment. + + Returns: + model: The loaded and (optionally) FSDP-wrapped model. + """ + # To avoid DTensor issues, load the model from a consolidated checkpoint in Tensor format before applying FSDP. + fsdp_shard_size = config.model.config.fsdp_shard_size + config.model.config.fsdp_shard_size = 1 # Set to 1 to disable FSDP during model instantiation. + model = instantiate(config.model).cuda() + # DCP checkpointer does not support loading from a consolidated checkpoint, so we support it here. + model = load_model_state_dict_from_checkpoint( + model=model, + config=config, + s3_checkpoint_dir=config.checkpoint.load_path, + load_ema_to_reg=config.checkpoint.load_ema_to_reg, + ) + # If FSDP is enabled, apply FSDP to the model. + if fsdp_shard_size > 1: + config.model.config.fsdp_shard_size = fsdp_shard_size + fsdp_device_mesh = hsdp_device_mesh( + sharding_group_size=fsdp_shard_size, + ) + if hasattr(model, "apply_fsdp") and callable(model.apply_fsdp): + model.apply_fsdp(fsdp_device_mesh) + else: + raise AttributeError( + "Model does not implement 'apply_fsdp'. Please implement this method to enable FSDP after consolidated checkpoint loading." + ) + + return model diff --git a/REGEN-main/cosmos_policy/_src/predict2/utils/optim_instantiate.py b/REGEN-main/cosmos_policy/_src/predict2/utils/optim_instantiate.py new file mode 100644 index 0000000000000000000000000000000000000000..5b3e530215ee15e0054cee631614fbee5c1464b7 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/utils/optim_instantiate.py @@ -0,0 +1,88 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +import hydra +import torch +from omegaconf import ListConfig +from torch import nn + +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.predict2.utils.fused_adam_dtensor import FusedAdam + + +def get_regular_param_group(net: nn.Module): + """ + seperate the parameters of the network into two groups: decay and no_decay. + based on nano_gpt codebase. + """ + param_dict = {pn: p for pn, p in net.named_parameters()} + param_dict = {pn: p for pn, p in param_dict.items() if p.requires_grad} + + decay_params = [p for n, p in param_dict.items() if p.dim() >= 2] + nodecay_params = [p for n, p in param_dict.items() if p.dim() < 2] + return decay_params, nodecay_params + + +def get_base_optimizer( + model: nn.Module, + lr: float, + weight_decay: float, + optim_type: str = "adamw", + **kwargs, +) -> torch.optim.Optimizer: + net_decay_param, net_nodecay_param = get_regular_param_group(model) + + num_decay_params = sum(p.numel() for p in net_decay_param) + num_nodecay_params = sum(p.numel() for p in net_nodecay_param) + net_param_total = num_decay_params + num_nodecay_params + log.critical(f"total num parameters : {net_param_total:,}") + + param_group = [ + { + "params": net_decay_param + net_nodecay_param, + "lr": lr, + "weight_decay": weight_decay, + }, + ] + + if optim_type == "adamw": + opt_cls = torch.optim.AdamW + elif optim_type == "fusedadam": + opt_cls = FusedAdam + else: + raise ValueError(f"Unknown optimizer type: {optim_type}") + + for k, v in kwargs.items(): + if isinstance(v, ListConfig): + kwargs[k] = list(v) + + return opt_cls(param_group, **kwargs) + + +def get_base_scheduler( + optimizer: torch.optim.Optimizer, + model: nn.Module, + scheduler_config: dict, +): + net_scheduler = hydra.utils.instantiate(scheduler_config) + net_scheduler.model = model + + return torch.optim.lr_scheduler.LambdaLR( + optimizer, + lr_lambda=[ + net_scheduler.schedule, + ], + ) diff --git a/REGEN-main/cosmos_policy/_src/predict2/utils/test_helper.py b/REGEN-main/cosmos_policy/_src/predict2/utils/test_helper.py new file mode 100644 index 0000000000000000000000000000000000000000..60328539b8dd3b60dcff73d3e77e6943a559bd59 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/utils/test_helper.py @@ -0,0 +1,169 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""Utilities for comparing PyTorch tensors with detailed difference reporting.""" + +from dataclasses import dataclass +from typing import List, Optional, Sequence, Tuple + +import torch +from torch import Tensor + + +@dataclass +class TensorDifference: + """Contains detailed information about differences between two tensors.""" + + name: str + max_absolute_diff: float + max_relative_diff: float + abs_diff_index: List[int] + rel_diff_index: List[int] + absolute_tolerance: float + relative_tolerance: float + tensor_shape: Tuple[int, ...] + error_message: str + + def __str__(self) -> str: + """Formats the difference information as a human-readable string.""" + return ( + f"{self.name}:\n" + f" Shape: {self.tensor_shape}\n" + f" Max absolute difference: {self.max_absolute_diff:.6f} at index {self.abs_diff_index}" + f" (tolerance: {self.absolute_tolerance})\n" + f" Max relative difference: {self.max_relative_diff:.6f} at index {self.rel_diff_index}" + f" (tolerance: {self.relative_tolerance})\n" + f" Details: {self.error_message}" + ) + + +def compute_tensor_differences(tensor_a: Tensor, tensor_b: Tensor, epsilon: float = 1e-12) -> Tuple[Tensor, Tensor]: + """Computes absolute and relative differences between two tensors. + + Args: + tensor_a: First tensor for comparison. + tensor_b: Second tensor for comparison. + epsilon: Small value to prevent division by zero in relative difference. + + Returns: + Tuple of (absolute_differences, relative_differences) tensors. + """ + abs_diff = torch.abs(tensor_a - tensor_b) + rel_diff = abs_diff / (torch.abs(tensor_b) + epsilon) + return abs_diff, rel_diff + + +def get_max_difference_info(differences: Tensor, flatten: bool = True) -> Tuple[float, List[int]]: + """Gets the maximum difference and its location in the tensor. + + Args: + differences: Tensor containing differences. + flatten: Whether to flatten the tensor before finding max. + + Returns: + Tuple of (max_difference, index_of_max). + """ + if flatten: + differences = differences.flatten() + max_diff = differences.max().item() + max_idx = differences.argmax().tolist() + return max_diff, [max_idx] if isinstance(max_idx, int) else max_idx + + +def compare_tensors( + names: Sequence[str], + tensors_a: Sequence[Tensor], + tensors_b: Sequence[Tensor], + atol: float = 0.5, + rtol: float = 0.05, + raise_on_mismatch: bool = True, + verbose: bool = True, +) -> List[Optional[TensorDifference]]: + """Compares two sets of tensors and provides detailed mismatch information. + + Args: + names: Sequence of tensor names or layer identifiers. + tensors_a: First sequence of tensors for comparison. + tensors_b: Second sequence of tensors for comparison. + atol: Absolute tolerance for comparison. + rtol: Relative tolerance for comparison. + raise_on_mismatch: Whether to raise ValueError on tolerance violations. + verbose: Whether to print detailed comparison information. + + Returns: + List of TensorDifference objects for mismatched tensors, None for matched ones. + + Raises: + ValueError: If tensors don't match and raise_on_mismatch is True. + RuntimeError: If input sequences have different lengths. + """ + if len(names) != len(tensors_a) or len(tensors_a) != len(tensors_b): + raise RuntimeError( + f"Input sequence lengths must match: " + f"names({len(names)}), tensors_a({len(tensors_a)}), " + f"tensors_b({len(tensors_b)})" + ) + + differences: List[Optional[TensorDifference]] = [] + mismatched_names: List[str] = [] + + for name, tensor_a, tensor_b in zip(names, tensors_a, tensors_b): + if tensor_a.shape != tensor_b.shape: + diff = TensorDifference( + name=name, + max_absolute_diff=float("inf"), + max_relative_diff=float("inf"), + abs_diff_index=[], + rel_diff_index=[], + absolute_tolerance=atol, + relative_tolerance=rtol, + tensor_shape=tensor_a.shape, + error_message=f"Shape mismatch: {tensor_a.shape} vs {tensor_b.shape}", + ) + differences.append(diff) + mismatched_names.append(name) + if verbose: + print(str(diff)) + continue + + try: + torch.testing.assert_close(tensor_a, tensor_b, atol=atol, rtol=rtol, check_device=False) + differences.append(None) + except AssertionError as e: + abs_diff, rel_diff = compute_tensor_differences(tensor_a, tensor_b) + max_abs_diff, max_abs_idx = get_max_difference_info(abs_diff) + max_rel_diff, max_rel_idx = get_max_difference_info(rel_diff) + + diff = TensorDifference( + name=name, + max_absolute_diff=max_abs_diff, + max_relative_diff=max_rel_diff, + abs_diff_index=max_abs_idx, + rel_diff_index=max_rel_idx, + absolute_tolerance=atol, + relative_tolerance=rtol, + tensor_shape=tensor_a.shape, + error_message=str(e), + ) + differences.append(diff) + mismatched_names.append(name) + + if verbose: + print(str(diff)) + + if mismatched_names and raise_on_mismatch: + raise ValueError(f"Tensors did not match within tolerances for: {', '.join(mismatched_names)}") + + return differences diff --git a/REGEN-main/cosmos_policy/_src/predict2/utils/tokenizer_benchmarking.py b/REGEN-main/cosmos_policy/_src/predict2/utils/tokenizer_benchmarking.py new file mode 100644 index 0000000000000000000000000000000000000000..5ea400b8adb3e4bcce224233ecb307af0fe381c6 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/predict2/utils/tokenizer_benchmarking.py @@ -0,0 +1,35 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from dataclasses import dataclass + + +@dataclass +class BenchmarkTimes: + """ + Class used to store times computed during tokenizer benchmarking. + All times are in seconds. + """ + + model_invocation: float = 0.0 + # Model's invocation time + overhead + total: float = 0.0 + + @property + def overhead(self) -> float: + return self.total - self.model_invocation + + def __repr__(self) -> str: + return f"BenchmarkTimes(model_invocation={self.model_invocation}, overhead={self.overhead}, total={self.total})" diff --git a/REGEN-main/cosmos_policy/_src/reason1/__init__.py b/REGEN-main/cosmos_policy/_src/reason1/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..7ab23eecabce51f7ab7edcf3835f22ebba880d81 --- /dev/null +++ b/REGEN-main/cosmos_policy/_src/reason1/__init__.py @@ -0,0 +1,15 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + diff --git a/REGEN-main/cosmos_policy/config/conditioner/__init__.py b/REGEN-main/cosmos_policy/config/conditioner/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/config/conditioner/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/config/conditioner/video2world_conditioner.py b/REGEN-main/cosmos_policy/config/conditioner/video2world_conditioner.py new file mode 100644 index 0000000000000000000000000000000000000000..b8a75679702a5ac23f7072d552711170137470e8 --- /dev/null +++ b/REGEN-main/cosmos_policy/config/conditioner/video2world_conditioner.py @@ -0,0 +1,365 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Video2World conditioner configurations for Cosmos Policy. + +Provides mutable versions of Video2WorldCondition and related classes. +These need to be mutable for Cosmos Policy since it modifies parts of the condition +objects during training. +""" + +import random +from dataclasses import dataclass +from typing import Dict, Optional + +import torch +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.utils.context_parallel import broadcast_split_tensor +from cosmos_policy._src.predict2.conditioner import ( + BooleanFlag, + ReMapkey, + TextAttr, + TextAttrEmptyStringDrop, +) +from cosmos_policy._src.predict2.models.video2world_wan2pt1_model import WAN2PT1_I2V_COND_LATENT_KEY +from cosmos_policy._src.predict2.networks.clip import Wan2pt1CLIPEmb +from cosmos_policy.conditioner import ( + GeneralConditioner, + Text2WorldCondition, +) + + +@dataclass(frozen=False) +class Video2WorldCondition(Text2WorldCondition): + """Mutable version of Video2WorldCondition for Cosmos Policy.""" + + use_video_condition: bool = False + # the following two attributes are used to set the video condition; during training, inference + gt_frames: Optional[torch.Tensor] = None + condition_video_input_mask_B_C_T_H_W: Optional[torch.Tensor] = None + + def set_video_condition( + self, + gt_frames: torch.Tensor, + random_min_num_conditional_frames: int, + random_max_num_conditional_frames: int, + num_conditional_frames: Optional[int] = None, + conditional_frames_probs: Optional[Dict[int, float]] = None, + ) -> "Video2WorldCondition": + """ + Sets the video conditioning frames for video-to-video generation. + + This method creates a conditioning mask for the input video frames that determines + which frames will be used as context frames for generating new frames. The method + handles both image batches (T=1) and video batches (T>1) differently. + + Args: + gt_frames: A tensor of ground truth frames with shape [B, C, T, H, W], where: + B = batch size + C = number of channels + T = number of frames + H = height + W = width + + random_min_num_conditional_frames: Minimum number of frames to use for conditioning + when randomly selecting a number of conditioning frames. + + random_max_num_conditional_frames: Maximum number of frames to use for conditioning + when randomly selecting a number of conditioning frames. + + num_conditional_frames: Optional; If provided, all examples in the batch will use + exactly this many frames for conditioning. If None, a random number of frames + between random_min_num_conditional_frames and random_max_num_conditional_frames + will be selected for each example in the batch. + + conditional_frames_probs: Optional; Dictionary mapping number of frames to probabilities. + If provided, overrides the random_min/max_num_conditional_frames with weighted sampling. + Example: {0: 0.5, 1: 0.25, 2: 0.25} for 50% chance of 0 frames, 25% for 1, 25% for 2. + + Returns: + A new Video2WorldCondition object with the gt_frames and conditioning mask set. + The conditioning mask (condition_video_input_mask_B_C_T_H_W) is a binary tensor + of shape [B, 1, T, H, W] where 1 indicates frames used for conditioning and 0 + indicates frames to be generated. + + Notes: + - For image batches (T=1), no conditioning frames are used (num_conditional_frames_B = 0). + - For video batches: + - If num_conditional_frames is provided, all examples use that fixed number of frames. + - Otherwise, each example randomly uses between random_min_num_conditional_frames and + random_max_num_conditional_frames frames. + - The mask marks the first N frames as conditioning frames (set to 1) for each example. + """ + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = gt_frames + + # condition_video_input_mask_B_C_T_H_W + B, _, T, H, W = gt_frames.shape + condition_video_input_mask_B_C_T_H_W = torch.zeros( + B, 1, T, H, W, dtype=gt_frames.dtype, device=gt_frames.device + ) + if T == 1: # handle image batch + num_conditional_frames_B = torch.zeros(B, dtype=torch.int32) + else: # handle video batch + if num_conditional_frames is not None: + if isinstance(num_conditional_frames, torch.Tensor): + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames.cpu() + else: + num_conditional_frames_B = torch.ones(B, dtype=torch.int32) * num_conditional_frames + elif conditional_frames_probs is not None: + # Use weighted sampling based on provided probabilities + frames_options = list(conditional_frames_probs.keys()) + weights = list(conditional_frames_probs.values()) + num_conditional_frames_B = torch.tensor( + random.choices(frames_options, weights=weights, k=B), dtype=torch.int32 + ) + else: + num_conditional_frames_B = torch.randint( + random_min_num_conditional_frames, random_max_num_conditional_frames + 1, size=(B,) + ) + for idx in range(B): + condition_video_input_mask_B_C_T_H_W[idx, :, : num_conditional_frames_B[idx], :, :] += 1 + + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + return type(self)(**kwargs) + + def edit_for_inference( + self, is_cfg_conditional: bool = True, num_conditional_frames: int = 1 + ) -> "Video2WorldCondition": + _condition = self.set_video_condition( + gt_frames=self.gt_frames, + random_min_num_conditional_frames=0, + random_max_num_conditional_frames=0, + num_conditional_frames=num_conditional_frames, + ) + if not is_cfg_conditional: + # Do not use classifier free guidance on conditional frames. + # YB found that it leads to worse results. + _condition.use_video_condition.fill_(True) + return _condition + + def broadcast(self, process_group: torch.distributed.ProcessGroup) -> "Video2WorldCondition": + if self.is_broadcasted: + return self + # extra efforts + gt_frames = self.gt_frames + condition_video_input_mask_B_C_T_H_W = self.condition_video_input_mask_B_C_T_H_W + kwargs = self.to_dict(skip_underscore=False) + kwargs["gt_frames"] = None + kwargs["condition_video_input_mask_B_C_T_H_W"] = None + new_condition = Text2WorldCondition.broadcast( + type(self)(**kwargs), + process_group, + ) + + kwargs = new_condition.to_dict(skip_underscore=False) + _, _, T, _, _ = gt_frames.shape + if process_group is not None: + if T > 1 and process_group.size() > 1: + gt_frames = broadcast_split_tensor(gt_frames, seq_dim=2, process_group=process_group) + condition_video_input_mask_B_C_T_H_W = broadcast_split_tensor( + condition_video_input_mask_B_C_T_H_W, seq_dim=2, process_group=process_group + ) + kwargs["gt_frames"] = gt_frames + kwargs["condition_video_input_mask_B_C_T_H_W"] = condition_video_input_mask_B_C_T_H_W + return type(self)(**kwargs) + + +class Video2WorldConditionV2(Video2WorldCondition): + """ + compared to Video2WorldCondition, this class apply zero frames when use_video_condition is False~(unconditional generation in cfg) + in the case, we do zero-out conditional frames in the video condition + """ + + def set_video_condition( + self, + gt_frames: torch.Tensor, + random_min_num_conditional_frames: int, + random_max_num_conditional_frames: int, + num_conditional_frames: Optional[int] = None, + conditional_frames_probs: Optional[Dict[int, float]] = None, + ) -> "Video2WorldConditionV2": + num_conditional_frames = 0 if not self.use_video_condition else num_conditional_frames + return super().set_video_condition( + gt_frames=gt_frames, + random_min_num_conditional_frames=random_min_num_conditional_frames, + random_max_num_conditional_frames=random_max_num_conditional_frames, + num_conditional_frames=num_conditional_frames, + conditional_frames_probs=conditional_frames_probs, + ) + + def edit_for_inference( + self, is_cfg_conditional: bool = True, num_conditional_frames: int = 1 + ) -> "Video2WorldConditionV2": + del is_cfg_conditional + _condition = super().set_video_condition( + gt_frames=self.gt_frames, + random_min_num_conditional_frames=0, + random_max_num_conditional_frames=0, + num_conditional_frames=num_conditional_frames, + ) + return _condition + + +class Video2WorldConditioner(GeneralConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> Video2WorldCondition: + output = super()._forward(batch, override_dropout_rate) + return Video2WorldCondition(**output) + + +class Video2WorldConditionerV2(GeneralConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> Video2WorldConditionV2: + output = super()._forward(batch, override_dropout_rate) + return Video2WorldConditionV2(**output) + + +_SHARED_CONFIG = dict( + fps=L(ReMapkey)( + input_key="fps", + output_key="fps", + dropout_rate=0.0, + dtype=None, + ), + padding_mask=L(ReMapkey)( + input_key="padding_mask", + output_key="padding_mask", + dropout_rate=0.0, + dtype=None, + ), + text=L(TextAttr)( + input_key=["t5_text_embeddings"], + dropout_rate=0.2, + use_empty_string=False, + ), + use_video_condition=L(BooleanFlag)( + input_key="fps", + output_key="use_video_condition", + dropout_rate=0.2, + ), +) + +VideoPredictionConditioner: LazyDict = L(Video2WorldConditioner)( + **_SHARED_CONFIG, +) + +VideoPredictionConditionerV2: LazyDict = L(Video2WorldConditionerV2)( + **_SHARED_CONFIG, +) + + +@dataclass(frozen=False) +class VideoPredictionWan2pt1Condition(Text2WorldCondition): + """Mutable version of VideoPredictionWan2pt1Condition for Cosmos Policy.""" + + frame_cond_crossattn_emb_B_L_D: Optional[torch.Tensor] = None + y_B_C_T_H_W: Optional[torch.Tensor] = None # image condition + # latent_condition: Optional[torch.Tensor] = None # latent condition + + def broadcast(self, process_group: torch.distributed.ProcessGroup) -> "Video2WorldCondition": + """Broadcasts and splits the condition across the checkpoint parallelism group. + For most condition, such asT2VCondition, we do not need split. + + Args: + process_group: The process group for broadcast and split + + Returns: + A new BaseCondition instance with the broadcasted and split condition. + """ + if self.is_broadcasted: + return self + + y_B_C_T_H_W = self.y_B_C_T_H_W + kwargs = self.to_dict(skip_underscore=False) + kwargs["y_B_C_T_H_W"] = None + new_condition = Text2WorldCondition.broadcast( + type(self)(**kwargs), + process_group, + ) + kwargs = new_condition.to_dict(skip_underscore=False) + if process_group is not None: + y_B_C_T_H_W = broadcast_split_tensor(y_B_C_T_H_W, seq_dim=2, process_group=process_group) + kwargs["y_B_C_T_H_W"] = y_B_C_T_H_W + return type(self)(**kwargs) + + +class VideoPredictionWan2pt1Conditioner(GeneralConditioner): + def forward( + self, + batch: Dict, + override_dropout_rate: Optional[Dict[str, float]] = None, + ) -> VideoPredictionWan2pt1Condition: + output = super()._forward(batch, override_dropout_rate) + return VideoPredictionWan2pt1Condition(**output) + + +VideoConditionerFpsPaddingEmptyStringDrppConfig: LazyDict = L(VideoPredictionWan2pt1Conditioner)( + text=L(TextAttrEmptyStringDrop)( + input_key=["t5_text_embeddings"], + dropout_rate=0.2, + ), + fps=L(ReMapkey)( + input_key="fps", + output_key="fps", + dropout_rate=0.0, + dtype=None, + ), + padding_mask=L(ReMapkey)( + input_key="padding_mask", + output_key="padding_mask", + dropout_rate=0.0, + dtype=None, + ), + wanclip=L(Wan2pt1CLIPEmb)( + input_key=["images", "video", WAN2PT1_I2V_COND_LATENT_KEY], + dropout_rate=0.0, + dtype="bfloat16", + ), +) + + +def register_conditioner(): + cs = ConfigStore.instance() + cs.store( + group="conditioner", + package="model.config.conditioner", + name="video_prediction_conditioner", + node=VideoPredictionConditioner, + ) + + cs.store( + group="conditioner", + package="model.config.conditioner", + name="video_prediction_conditioner_v2", + node=VideoPredictionConditionerV2, + ) + + cs.store( + group="conditioner", + package="model.config.conditioner", + name="wan2pt1_video_prediction_conditioner_empty_string_drop", + node=VideoConditionerFpsPaddingEmptyStringDrppConfig, + ) diff --git a/REGEN-main/cosmos_policy/config/defaults/__init__.py b/REGEN-main/cosmos_policy/config/defaults/__init__.py new file mode 100644 index 0000000000000000000000000000000000000000..3159bfe65645499015bd92609b99d476d69544e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/config/defaults/__init__.py @@ -0,0 +1,14 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/REGEN-main/cosmos_policy/config/defaults/model.py b/REGEN-main/cosmos_policy/config/defaults/model.py new file mode 100644 index 0000000000000000000000000000000000000000..a81ea639b48e0f5ecd432f40bee7e639b265601c --- /dev/null +++ b/REGEN-main/cosmos_policy/config/defaults/model.py @@ -0,0 +1,58 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Cosmos Policy model registration for Hydra ConfigStore. + +This registers policy-specific model classes that extend the base predict2 models. +""" + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy.models.policy_video2world_model import ( + CosmosPolicyVideo2WorldConfig, + CosmosPolicyVideo2WorldModel, +) + +# Use policy-specific models with the same config structure +POLICY_DDP_CONFIG = dict( + trainer=dict( + distributed_parallelism="ddp", + ), + model=L(CosmosPolicyVideo2WorldModel)( + config=CosmosPolicyVideo2WorldConfig(), + _recursive_=False, + ), +) + +POLICY_FSDP_CONFIG = dict( + trainer=dict( + distributed_parallelism="fsdp", + ), + model=L(CosmosPolicyVideo2WorldModel)( + config=CosmosPolicyVideo2WorldConfig( + fsdp_shard_size=8, + ), + _recursive_=False, + ), +) + + +def register_policy_model(): + """Register Cosmos Policy model configurations.""" + cs = ConfigStore.instance() + cs.store(group="model", package="_global_", name="policy_ddp", node=POLICY_DDP_CONFIG) + cs.store(group="model", package="_global_", name="policy_fsdp", node=POLICY_FSDP_CONFIG) diff --git a/REGEN-main/cosmos_policy/config/defaults/tokenizer.py b/REGEN-main/cosmos_policy/config/defaults/tokenizer.py new file mode 100644 index 0000000000000000000000000000000000000000..6b793a2a21a25e10d345ec5618aa11a016211e3f --- /dev/null +++ b/REGEN-main/cosmos_policy/config/defaults/tokenizer.py @@ -0,0 +1,59 @@ +# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. +# SPDX-License-Identifier: Apache-2.0 +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Cosmos Policy tokenizer registration with deterministic seeding support. +""" + +import os + +from hydra.core.config_store import ConfigStore + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy.tokenizers.wan2pt1 import Wan2pt1VAEInterface + +# Policy-specific wan2pt1 tokenizer with deterministic seeding. The official +# default is gated on Hugging Face; local rollout launchers set this environment +# variable so every worker reuses one verified local copy instead. +POLICY_WAN_VAE_PATH = os.environ.get( + "COSMOS_POLICY_WAN_VAE_PATH", + "hf://nvidia/Cosmos-Predict2-2B-Video2World/tokenizer/tokenizer.pth", +) + +PolicyWan2pt1VAEConfig = L(Wan2pt1VAEInterface)( + vae_pth=POLICY_WAN_VAE_PATH, + s3_credential_path="credentials/s3_training.secret", + load_mean_std=False, + temporal_window=4, + is_parallel=False, + cp_grid_shape=None, +) + + +def register_policy_tokenizer(): + """ + Register Cosmos Policy tokenizer configurations. + + This registers the wan2pt1 tokenizer with deterministic seeding support. + To enable deterministic encoding, set: DETERMINISTIC=true + """ + cs = ConfigStore.instance() + # Also register with explicit policy prefix + cs.store( + group="tokenizer", + package="model.config.tokenizer", + name="policy_wan2pt1_tokenizer", + node=PolicyWan2pt1VAEConfig, + ) diff --git a/REGEN-main/cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py b/REGEN-main/cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py new file mode 100644 index 0000000000000000000000000000000000000000..31d8811d458c3f4f1ca10d4a51eba28b0f2f63d3 --- /dev/null +++ b/REGEN-main/cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py @@ -0,0 +1,1821 @@ +# ----------------------------------------------------------------------------- +# Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES. +# All rights reserved. +# +# This codebase constitutes NVIDIA proprietary technology and is strictly +# confidential. Any unauthorized reproduction, distribution, or disclosure +# of this code, in whole or in part, outside NVIDIA is strictly prohibited +# without prior written consent. +# +# For inquiries regarding the use of this code in other NVIDIA proprietary +# projects, please contact the Deep Imagination Research Team at +# dir@exchange.nvidia.com. +# ----------------------------------------------------------------------------- + +import os + +from hydra.core.config_store import ConfigStore +from megatron.core import parallel_state +from torch.utils.data import DataLoader, DistributedSampler + +from cosmos_policy._src.imaginaire.lazy_config import LazyCall as L +from cosmos_policy._src.imaginaire.lazy_config import LazyDict +from cosmos_policy._src.imaginaire.utils import log +from cosmos_policy._src.imaginaire.utils.checkpoint_db import get_checkpoint_path # noqa: F401 +from cosmos_policy.config.experiment.libero_task_constants import get_replay_tasks +from cosmos_policy.datasets.aloha_dataset import ALOHADataset +from cosmos_policy.datasets.libero_dataset import LIBERODataset +from cosmos_policy.datasets.robocasa_dataset import RoboCasaDataset +from cosmos_policy.models.policy_video2world_model import CosmosPolicyVideo2WorldModel +from cosmos_policy.modules.hybrid_edm_sde import HybridEDMSDE + +cs = ConfigStore.instance() +val_sampling_size_override = dict( + video_length=121, + video_height=704, + video_width=1280, +) +BASE_DATASETS_DIR = os.environ.get("BASE_DATASETS_DIR", "/workspace/cosmos-policy") + + +def _env_int_list(name: str, default: list[int]) -> list[int]: + raw = os.environ.get(name, "").strip() + return [int(value.strip()) for value in raw.split(",") if value.strip()] if raw else default + + +def _env_bool(name: str, default: bool = False) -> bool: + raw = os.environ.get(name, "").strip().lower() + if not raw: + return default + if raw in {"1", "true", "yes", "on"}: + return True + if raw in {"0", "false", "no", "off"}: + return False + raise ValueError(f"{name} must be a boolean, got {raw!r}") + + +COSMOS_PREDICT2_BASE_CHECKPOINT_URI = ( + "hf://nvidia/Cosmos-Predict2-2B-Video2World/model-480p-16fps.pt" +) + + +def _default_base_checkpoint_path() -> str: + # Evaluation always supplies --ckpt_path, which model_loader applies after + # composing this config. Avoid eagerly downloading the unrelated default + # while the experiment registry is imported; training keeps the original + # eager resolution unless it explicitly opts into the deferred behavior. + if _env_bool("COSMOS_POLICY_DEFER_BASE_CHECKPOINT_RESOLUTION"): + return COSMOS_PREDICT2_BASE_CHECKPOINT_URI + return get_checkpoint_path(COSMOS_PREDICT2_BASE_CHECKPOINT_URI) + + +def _env_fraction_schedule(name: str, default: list[tuple[int, float]]) -> list[tuple[int, float]]: + raw = os.environ.get(name, "").strip() + if not raw: + return default + schedule = [] + for item in raw.split(","): + step, fraction = item.split(":", 1) + schedule.append((int(step.strip()), float(fraction.strip()))) + return schedule + + +# *** Main checkpoint *** +libero_all_4_suites_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only"), # Successful demos + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + rollout_data_dir=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "all_episodes" + ), # All demo rollouts (successes + failures) + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) + + +libero_goal_suites_base_stage_task_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), # Successful demos + current_tasks_ids = get_replay_tasks("libero_goal", [0,1,2,3,4,5]), + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + rollout_data_dir=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "all_episodes" + ), # All demo rollouts (successes + failures) + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) + +libero_goal_suites_cl_stage_task_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids = get_replay_tasks("libero_goal", [2]), + # WARM-UP (student, CL step-4): task2 REAL demonstrations ONLY. + # replay_data_dir / rollout_data_dir left unset (default "") -> no old-task replay, no rollout mixing. + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + demonstration_sampling_prob=1.0, + return_value_function_returns=True, + gamma=0.99, +) + +libero_goal_suites_task6_ft_task_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids = get_replay_tasks("libero_goal", [6]), + # CL fine-tune: libero_goal task6 (put_the_cream_cheese_in_the_bowl) REAL demonstrations ONLY. + # This task is NOT in the base-stage training set ([0,1,2,3,4,5]) -> genuine new task. + # replay_data_dir / rollout_data_dir left unset (default "") -> no old-task replay, no rollout mixing. + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + demonstration_sampling_prob=1.0, + return_value_function_returns=True, + gamma=0.99, +) + +# Pure fine-tune on the 8 MERGED task3 segment trajectories (seg1+seg2 concatenated per demo, +# from new_task6/task3/replay_merged; 8 trajs / 1367 steps, 155-188 steps each). These are WAM +# DREAM rollouts, not real demos: seg1 is dreamt from t=0, seg2 is dreamt from a real anchor frame +# at t_grasp, and the two are joined at that anchor (arm state continuous, sharpness jumps). +# NOTE: replay_merged/ carries SYMLINKED dataset_statistics{,_post_norm}.json pointing at +# success_only/libero_goal_regen, so normalization matches base/eval instead of being recomputed +# from these 8 trajectories. No replay_data_dir / rollout_data_dir -> task3 only, no mixing. +libero_goal_suites_task3_merged_ft_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "new_task6", "task3", "replay_merged"), + current_tasks_ids=get_replay_tasks("libero_goal", [3]), + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + demonstration_sampling_prob=1.0, + return_value_function_returns=True, + gamma=0.99, +) + +# task7 CL with FORGETTING-CURATED replay: new task7 real demos + the 10 most-forgotten generated +# trajectories per old task (0-6). Forgetting was measured by forgetting_harness with +# Teacher = fcv2 iter_350 and Student = that same model after 100 steps of task7-only fine-tuning, +# i.e. exactly the drift this run is meant to undo. Candidate pool was the IDM-filtered new_task7 +# generations (20/task); task3's 20 are seg1+seg2 merges. replay dir holds 70 symlinks + symlinked +# dataset_statistics so normalization matches base/eval instead of being recomputed. +# max_replay_demos=16 > 10/task makes the per-task cap inert - the directory itself is the selection. +libero_goal_suites_task7_forget_cl_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [7]), + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "new_task7", "forget_replay_top10"), + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5, 6]), + max_replay_demos=16, + current_task_step_fraction=0.5, # 1:1 new:replay by STEP count (project default for CL runs) + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + demonstration_sampling_prob=1.0, + return_value_function_returns=True, + gamma=0.99, +) + +# Round 2 of forgetting-curated task7 CL. Same recipe as round 1, but the replay set was re-picked +# by re-running forgetting_harness with Student = the ROUND-1 model (task7_forget_cl iter_350) +# instead of the pure fine-tune, so it targets what is still drifting after round 1 rather than +# what drifted right after learning task7. Overlap with round 1's picks is only 5-8 per task. +libero_goal_suites_task7_forget_cl_v2_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [7]), + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "new_task7", "forget_replay_top10_v2"), + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5, 6]), + max_replay_demos=16, + current_task_step_fraction=0.5, + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + demonstration_sampling_prob=1.0, + return_value_function_returns=True, + gamma=0.99, +) + +# REGEN continual-learning dataset: NEW task6 real demos (data_dir + current_tasks_ids) mixed with +# OLD tasks 0-5 REAL demos replayed from the SAME success_only/libero_goal_regen pool (replay_data_dir +# + replay_tasks + max_replay_demos). Switched from WAM rollouts (old/rollouts_old_tasks_wv) to real +# demos because the regenerated replay was unfaithful for contact-rich grasp tasks (task1 collapsed to +# 0% at iter800). Real old demos load into the DEMONSTRATION pool (data/demo_N format), so there is NO +# rollout channel: the 0.5/0.5 demo/rollout split is inert and sampling is step-proportional over +# task6 (39) + tasks 0-5 (<=50/task = 260) = 299 real demos. +libero_goal_suites_task6_regen_cl_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [6]), # NEW task6 real demos (39) + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), # OLD tasks 0-5 REAL demos + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5]), # OLD tasks 0-5 real replay + max_replay_demos=50, # <=50 per old task -> 260 total + current_task_step_fraction=0.5, # 50/50 new(task6) vs replay by sampled steps (user: all CL runs 1:1) + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) + +# FORGETTING-CURATED continual-learning dataset: same recipe as regen_cl (NEW task6 real demos + +# OLD tasks 0-5 REAL demo replay), but the replay set is HAND-PICKED by Teacher-vs-Student forgetting +# score instead of "first N per task". replay_data_dir points at new/task_forget_replay_curated/, a +# directory of per-demo HDF5 symlinks: tasks 0,1,2,4,5 -> their 10 most-forgotten trajectories each +# (forgetting_harness top-10), task3 -> all 16 relaxed-reliable segments (replay10 seg1/seg2, renamed +# so the filename encodes the task3 instruction). max_replay_demos=16 only needs to be >= the largest +# per-task count (16, task3); the curated dir IS the selection, so the cap is inert (10<16 for the +# others). dataset_statistics{,_post_norm}.json are symlinked from libero_goal_regen so normalization +# matches the base model / eval. Total = task6 (39) + 5x10 + 16 = 105 real demos. +libero_goal_suites_task6_forget_cl_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [6]), # NEW task6 real demos (39) + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "new", "task_forget_replay_curated"), # curated forgetting-picked replay + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5]), # OLD tasks 0-5 real replay + max_replay_demos=16, # >= max per-task count; cap inert (dir is the selection) + current_task_step_fraction=0.5, # 50/50 new(task6) vs replay by sampled steps + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) + +# CONTROLLED EXPERIMENT vs regen_cl: the SAME 60 WAM old-task rollouts (10/task, tasks 0-5) that +# regen_cl fed through the ROLLOUT channel (rollout_data_dir=old/rollouts_old_tasks_wv, flat format) +# are here routed through the DEMO channel (replay_data_dir) instead — using their demo-format twins +# old/task*/trajectories (data/demo_N/agentview_rgb_jpeg..., md5-identical content to the wv rollouts), +# symlinked into old/regen_replay_demofmt. Only the channel changes vs regen_cl; warm-start (iter_40000), +# 50/50 new:replay (current_task_step_fraction), 800 steps all match regen_cl. Isolates "does the +# demo(BC) vs rollout channel routing change task1's fate?". +libero_goal_suites_task6_regen_cl_demochan_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [6]), # NEW task6 real demos (39) + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "old", "regen_replay_demofmt"), # regen WAM rollouts (demo-format), demo channel + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5]), # OLD tasks 0-5 replay + max_replay_demos=10, # 10 per old task -> 60 total + current_task_step_fraction=0.5, # 50/50 new(task6) vs replay by sampled steps + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) + +cosmos_predict2_2b_480p_libero = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-102-Size-2B-Res-480-Fps-16-Note-HQ_V5_from_26", + {"override /data_train": "mock"}, + {"override /model": "policy_fsdp"}, + {"override /tokenizer": "policy_wan2pt1_tokenizer"}, + { + "override /callbacks": [ + "basic", + "long", + "cluster_speed", + "wandb", + "wandb_callback_actions", + ] + }, + "_self_", + ], + trainer=dict( + callbacks=dict( + every_n_sample_reg=dict( + every_n=100000, + save_s3=False, + use_negative_prompt=False, + guidance=[0], + num_sampling_step=9, + ), + ), + run_validation=False, + logging_iter=5, + max_iter=1000000, + straggler_detection=dict( + enabled=False, + ), + ), + optimizer=dict( + lr=1e-4, + ), + scheduler=dict( + # LR decay for 30K steps in cycle #1, then decay by 5x and stay constant forever in cycle #2 + cycle_lengths=[30000, 100000000000000], + warm_up_steps=[1000, 0], + f_start=[1e-6, 0.06], + f_max=[1.0, 0.06], + f_min=[0.3, 0.06], + ), + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + conditioner=dict( + text=dict( + # IMPORTANT: We don't want any text dropout; otherwise, the model may fail to follow language + dropout_rate=0.0, + ), + ), + state_t=9, # Latent temporal dim (blank, proprio, wrist, primary, action, future proprio, future wrist, future primary, value) + min_num_conditional_frames=4, # 1 blank, 3 conditioning (proprio, wrist, primary) + max_num_conditional_frames=4, # 1 blank, 3 conditioning (proprio, wrist, primary) + sigma_conditional=0.0, # No noise on conditional latents + conditioning_strategy="frame_replace", + denoise_replace_gt_frames=True, + tokenizer=dict( + chunk_duration=33, # 1 blank + 32 images (4 proprio, 4 wrist image, 4 primary image, 4 action, 4 future proprio, 4 future wrist, 4 future primary, 4 value) + ), + ema=dict( + enabled=False, + ), + input_data_key="video", + sde=L(HybridEDMSDE)( + hybrid_sigma_distribution=True, + p_mean=1.3862943611198906, # Copied from base model config + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + uniform_lower=1.0, + uniform_upper=85.0, + ), + adjust_video_noise=True, + resize_online=True, + resolution="224", + high_sigma_strategy="none", + ), + ), + model_parallel=dict( + context_parallel_size=1, + ), + checkpoint=dict( + load_path=_default_base_checkpoint_path(), + load_training_state=False, # This means do not load train state from the base checkpoint above (load_path); but when resuming this job, will load train state + strict_resume=False, + save_iter=1000, + load_ema_to_reg=True, + load_from_object_store=dict( + enabled=False, + ), + save_to_object_store=dict( + enabled=False, + ), + ), + dataloader_train=L(DataLoader)( + num_workers=12, + persistent_workers=True, + pin_memory=True, + dataset=libero_all_4_suites_dataset, + sampler=L(DistributedSampler)( + dataset=libero_all_4_suites_dataset, + num_replicas=L(parallel_state.get_data_parallel_world_size)(), + rank=L(parallel_state.get_data_parallel_rank)(), + shuffle=True, + seed=0, + ), + batch_size=30, + drop_last=True, + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero", + ), + upload_reproducible_setup=False, + ) +) +cosmos_predict2_2b_480p_libero_base_stage = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-102-Size-2B-Res-480-Fps-16-Note-HQ_V5_from_26", + {"override /data_train": "mock"}, + {"override /model": "policy_fsdp"}, + {"override /tokenizer": "policy_wan2pt1_tokenizer"}, + { + "override /callbacks": [ + "basic", + "long", + "cluster_speed", + "wandb", + "wandb_callback_actions", + ] + }, + "_self_", + ], + trainer=dict( + callbacks=dict( + every_n_sample_reg=dict( + every_n=100000, + save_s3=False, + use_negative_prompt=False, + guidance=[0], + num_sampling_step=9, + ), + ), + run_validation=False, + logging_iter=5, + max_iter=5000, + straggler_detection=dict( + enabled=False, + ), + ), + optimizer=dict( + lr=1e-4, + ), + scheduler=dict( + # LR decay for 30K steps in cycle #1, then decay by 5x and stay constant forever in cycle #2 + cycle_lengths=[30000, 100000000000000], + warm_up_steps=[1000, 0], + f_start=[1e-6, 0.06], + f_max=[1.0, 0.06], + f_min=[0.3, 0.06], + ), + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + conditioner=dict( + text=dict( + # IMPORTANT: We don't want any text dropout; otherwise, the model may fail to follow language + dropout_rate=0.0, + ), + ), + state_t=9, # Latent temporal dim (blank, proprio, wrist, primary, action, future proprio, future wrist, future primary, value) + min_num_conditional_frames=4, # 1 blank, 3 conditioning (proprio, wrist, primary) + max_num_conditional_frames=4, # 1 blank, 3 conditioning (proprio, wrist, primary) + sigma_conditional=0.0, # No noise on conditional latents + conditioning_strategy="frame_replace", + denoise_replace_gt_frames=True, + tokenizer=dict( + chunk_duration=33, # 1 blank + 32 images (4 proprio, 4 wrist image, 4 primary image, 4 action, 4 future proprio, 4 future wrist, 4 future primary, 4 value) + ), + ema=dict( + enabled=False, + ), + input_data_key="video", + sde=L(HybridEDMSDE)( + hybrid_sigma_distribution=True, + p_mean=1.3862943611198906, # Copied from base model config + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + uniform_lower=1.0, + uniform_upper=85.0, + ), + adjust_video_noise=True, + resize_online=True, + resolution="224", + high_sigma_strategy="none", + ), + ), + model_parallel=dict( + context_parallel_size=1, + ), + checkpoint=dict( + load_path=_default_base_checkpoint_path(), + load_training_state=False, # This means do not load train state from the base checkpoint above (load_path); but when resuming this job, will load train state + strict_resume=False, + save_iter=1000, + load_ema_to_reg=True, + load_from_object_store=dict( + enabled=False, + ), + save_to_object_store=dict( + enabled=False, + ), + ), + dataloader_train=L(DataLoader)( + num_workers=12, + persistent_workers=True, + pin_memory=True, + dataset=libero_goal_suites_base_stage_task_dataset, + sampler=L(DistributedSampler)( + dataset=libero_goal_suites_base_stage_task_dataset, + num_replicas=L(parallel_state.get_data_parallel_world_size)(), + rank=L(parallel_state.get_data_parallel_rank)(), + shuffle=True, + seed=0, + ), + batch_size=30, + drop_last=True, + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_base_stage", + ), + upload_reproducible_setup=False, + ) +) + +cosmos_predict2_2b_480p_libero_cl_stage = LazyDict( + dict( + defaults=[ + "/experiment/Stage-c_pt_4-Index-102-Size-2B-Res-480-Fps-16-Note-HQ_V5_from_26", + {"override /data_train": "mock"}, + {"override /model": "policy_fsdp"}, + {"override /tokenizer": "policy_wan2pt1_tokenizer"}, + { + "override /callbacks": [ + "basic", + "long", + "cluster_speed", + "wandb", + "wandb_callback_actions", + ] + }, + "_self_", + ], + trainer=dict( + callbacks=dict( + every_n_sample_reg=dict( + every_n=100000, + save_s3=False, + use_negative_prompt=False, + guidance=[0], + num_sampling_step=9, + ), + ), + run_validation=False, + logging_iter=5, + max_iter=800, + straggler_detection=dict( + enabled=False, + ), + ), + optimizer=dict( + lr=5e-5, + ), + scheduler=dict( + # 800-step schedule: 50-step warm-up (LR 0 -> peak 5e-5), then LINEAR decay over the + # remaining 750 steps down to f_min*lr = 0.3*5e-5 = 1.5e-5 by step 800. + # multiplier f x optimizer.lr(5e-5): warmup 0->1.0 @step50, decay 1.0->0.3 @step800. + cycle_lengths=[800, 100000000000000], + warm_up_steps=[50, 0], + f_start=[0.0, 0.06], + f_max=[1.0, 0.06], + f_min=[0.3, 0.06], + ), + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + conditioner=dict( + text=dict( + # IMPORTANT: We don't want any text dropout; otherwise, the model may fail to follow language + dropout_rate=0.0, + ), + ), + state_t=9, # Latent temporal dim (blank, proprio, wrist, primary, action, future proprio, future wrist, future primary, value) + min_num_conditional_frames=4, # 1 blank, 3 conditioning (proprio, wrist, primary) + max_num_conditional_frames=4, # 1 blank, 3 conditioning (proprio, wrist, primary) + sigma_conditional=0.0, # No noise on conditional latents + conditioning_strategy="frame_replace", + denoise_replace_gt_frames=True, + tokenizer=dict( + chunk_duration=33, # 1 blank + 32 images (4 proprio, 4 wrist image, 4 primary image, 4 action, 4 future proprio, 4 future wrist, 4 future primary, 4 value) + ), + ema=dict( + enabled=False, + ), + input_data_key="video", + sde=L(HybridEDMSDE)( + hybrid_sigma_distribution=True, + p_mean=1.3862943611198906, # Copied from base model config + p_std=1.2, + sigma_max=200, + sigma_min=0.01, + uniform_lower=1.0, + uniform_upper=85.0, + ), + adjust_video_noise=True, + resize_online=True, + resolution="224", + high_sigma_strategy="none", + + ), + ), + model_parallel=dict( + context_parallel_size=1, + ), + checkpoint=dict( + # Keep the previous-stage checkpoint lazy: this module is imported + # even for base-stage runs, and eagerly resolving a container-only + # local path here makes every experiment fail before CLI overrides + # can be applied. + load_path="/workspace/cosmos-policy/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage/iter_000007000", + load_training_state=False, # This means do not load train state from the base checkpoint above (load_path); but when resuming this job, will load train state + strict_resume=False, + save_iter=1000, + load_ema_to_reg=True, + load_from_object_store=dict( + enabled=False, + ), + save_to_object_store=dict( + enabled=False, + ), + ), + dataloader_train=L(DataLoader)( + num_workers=16, + persistent_workers=True, + pin_memory=True, + dataset=libero_goal_suites_cl_stage_task_dataset, + sampler=L(DistributedSampler)( + dataset=libero_goal_suites_cl_stage_task_dataset, + num_replicas=L(parallel_state.get_data_parallel_world_size)(), + rank=L(parallel_state.get_data_parallel_rank)(), + shuffle=True, + seed=0, + ), + batch_size=30, + drop_last=True, + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_cl_stage", + ), + upload_reproducible_setup=False, + ) +) + + +# CL fine-tune on libero_goal task6 (put_the_cream_cheese_in_the_bowl). +# Inherits the entire cl_stage recipe (lr, scheduler, model, callbacks incl. wandb), +# and ONLY swaps the training dataset to the task6 REAL-demos dataset + a distinct job.name +# so its checkpoints land in a fresh output dir (does NOT touch the base-stage iter_000040000). +# NOTE: the dataset OBJECT is bound here in the config (not via CLI string override) -- the +# CLI `dataloader_train.dataset=` override stays a bare string and breaks the DataLoader. +cosmos_predict2_2b_480p_libero_goal_task6_ft = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task6_ft_task_dataset, + sampler=dict( + dataset=libero_goal_suites_task6_ft_task_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task6_ft", + ), + upload_reproducible_setup=False, + ) +) + + +# Pure fine-tune on the 8 MERGED task3 dream trajectories (no replay, no rollout mixing). +# Inherits the entire cl_stage recipe INCLUDING its 800-step LR schedule (50-step warmup to 5e-5, +# then decay toward 1.5e-5 over 800) -- deliberately left as-is per user, so a 200-step run ends +# around ~4.3e-5 without finishing the decay. Only the dataset OBJECT and job.name are swapped. +cosmos_predict2_2b_480p_libero_goal_task3_merged_ft = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task3_merged_ft_dataset, + sampler=dict( + dataset=libero_goal_suites_task3_merged_ft_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task3_merged_ft", + ), + upload_reproducible_setup=False, + ) +) + + +# task7 CL on forgetting-curated replay (see libero_goal_suites_task7_forget_cl_dataset). +# Warm-starts from the 100-step task7-only fine-tune, so this run's job is to claw back the old +# tasks that fine-tune just damaged. Inherits cl_stage wholesale, including its 800-step LR +# schedule (left unchanged per user); a 350-step run therefore stops mid-decay. +cosmos_predict2_2b_480p_libero_goal_task7_forget_cl = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task7_forget_cl_dataset, + sampler=dict( + dataset=libero_goal_suites_task7_forget_cl_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task7_forget_cl", + ), + upload_reproducible_setup=False, + ) +) + + +# Round 2 of the forgetting-curated task7 CL, warm-started from round 1's iter_350. +cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2 = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task7_forget_cl_v2_dataset, + sampler=dict( + dataset=libero_goal_suites_task7_forget_cl_v2_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2", + ), + upload_reproducible_setup=False, + ) +) + + +# REGEN continual-learning on task6: NEW task6 real demos + OLD tasks 0-5 rollout replay. +# Inherits the entire cl_stage recipe (lr, scheduler, model, callbacks incl. wandb) and ONLY swaps +# the training dataset OBJECT + job.name. Warm-start from base_stage iter_40000 is set on the CLI +# via checkpoint.load_path. NOTE: the dataset OBJECT is bound here (NOT via CLI +# `dataloader_train.dataset=`, which stays a bare string and breaks the DataLoader at step 0). +cosmos_predict2_2b_480p_libero_goal_task6_regen_cl = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task6_regen_cl_dataset, + sampler=dict( + dataset=libero_goal_suites_task6_regen_cl_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task6_regen_cl", + ), + ) +) + + +# FORGETTING-CURATED continual-learning on task6: same cl_stage recipe as regen_cl, only the training +# dataset OBJECT + job.name differ. Replay = forgetting-picked demos (task_forget_replay_curated), +# new:replay sampling forced to 50/50 via current_task_step_fraction. Warm-start (typically the 100-step +# task6 ft) is passed on the CLI via checkpoint.load_path. Dataset object bound here, NOT via CLI. +cosmos_predict2_2b_480p_libero_goal_task6_forget_cl = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task6_forget_cl_dataset, + sampler=dict( + dataset=libero_goal_suites_task6_forget_cl_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task6_forget_cl", + ), + ) +) + + +# V2 forgetting-curated CL dataset: Student=forget_cl iter_350 selection (run_forget_v2.sh manifests) +# via task_forget_replay_curated_v2 -- different task0/1/2/4/5 picks vs v1, SAME task3 16 segments. +# Everything else identical to the v1 forget_cl dataset (50/50 new:replay). +libero_goal_suites_task6_forget_cl_v2_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [6]), + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "new", "task_forget_replay_curated_v2"), + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5]), + max_replay_demos=16, + current_task_step_fraction=0.5, + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, + use_stronger_image_aug=True, + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) + + +# V2 forgetting CL experiment: v2 selection + CONSTANT LR 3.5e-5 (no warm-up), continuing from the +# converged forget_cl iter_350 (passed via CLI checkpoint.load_path). Inherits cl_stage, overrides +# dataset + scheduler + optimizer.lr + job.name only. +cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2 = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task6_forget_cl_v2_dataset, + sampler=dict( + dataset=libero_goal_suites_task6_forget_cl_v2_dataset, + ), + ), + # Constant LR 3.5e-5, no warm-up: flat multiplier 1.0 (warm_up_steps=0, f_start=f_max=f_min=1.0, + # huge cycle so we never leave cycle 0) x optimizer.lr=3.5e-5. Overrides cl_stage's 1000-step warmup. + optimizer=dict(lr=3.5e-5), + scheduler=dict( + warm_up_steps=[0, 0], + cycle_lengths=[100000000, 100000000000000], + f_start=[1.0, 1.0], + f_max=[1.0, 1.0], + f_min=[1.0, 1.0], + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2", + ), + ) +) + + +# TASK7 (turn_on_the_stove) continual-learning dataset: NEW task7 real demos (data_dir + current_tasks_ids=[7], +# 50 demos) + REPLAY of task0-6 from demochan-iter800-generated rollouts (old_task7/replay_curated, 70 traj = +# task7 initial states + task0-6 language). 50/50 new:replay. Warm-start = demochan iter_800 (via CLI load_path). +libero_goal_suites_task7_cl_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [7]), # NEW task7 real demos (50) + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "old_task7", "replay_curated"), # demochan-gen task0-6 replay (70) + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5, 6]), # OLD tasks 0-6 replay + max_replay_demos=20, # >= 10/task; cap inert (dir is selection) + current_task_step_fraction=0.5, # 50/50 new(task7) vs replay + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, + use_stronger_image_aug=True, + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) + + +# TASK7 CL experiment: inherits cl_stage (now 800-step schedule: 50 warmup 0->5e-5, 750 decay ->1.5e-5), +# only swaps dataset + job.name. Warm-start (demochan iter_800) passed via CLI checkpoint.load_path. +cosmos_predict2_2b_480p_libero_goal_task7_cl = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task7_cl_dataset, + sampler=dict( + dataset=libero_goal_suites_task7_cl_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task7_cl", + ), + ) +) + + +# TASK7 REAL-replay continual-learning dataset: NEW task7 real demos (50) + REPLAY of task0-6 from the +# ORIGINAL human demos in success_only/libero_goal_regen (cap 50/task -> 42+50+46+38+47+37+39 = 299). +# Same recipe as the task6 real_cl run (libero_goal_suites_task6_regen_cl_dataset), shifted one task: +# the task7_cl run above used demochan-generated replay and forgot task0-6 down to ~33%, while real +# human-demo replay held task0-5 at 93% on task6 -> this run isolates replay quality at task7. +# Warm-start = real_cl iter_800 (task6 real_cl, 326/350 on task0-6), passed via CLI checkpoint.load_path. +libero_goal_suites_task7_real_cl_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [7]), # NEW task7 real demos (50) + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), # OLD tasks 0-6 REAL demos + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5, 6]), # OLD tasks 0-6 real replay + max_replay_demos=50, # <=50 per old task -> 299 total + current_task_step_fraction=0.5, # 50/50 new(task7) vs replay (user: all CL runs 1:1) + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, + use_stronger_image_aug=True, + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) + + +# TASK7 REAL-replay CL experiment: inherits cl_stage (800-step schedule: 50 warmup 0->5e-5, 750 decay +# ->1.5e-5), only swaps dataset + job.name. Warm-start (real_cl iter_800) via CLI checkpoint.load_path. +cosmos_predict2_2b_480p_libero_goal_task7_real_cl = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task7_real_cl_dataset, + sampler=dict( + dataset=libero_goal_suites_task7_real_cl_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task7_real_cl", + ), + ) +) + + +# TASK8 REAL-replay continual-learning dataset: NEW task8 (put_the_bowl_on_the_plate) real demos (50) + +# REPLAY of task0-7 from the ORIGINAL human demos in success_only/libero_goal_regen +# (cap 50/task -> 42+50+46+38+47+37+39+50 = 349). +# Same recipe as libero_goal_suites_task7_real_cl_dataset, shifted one task. +# Warm-start = real_cl7 iter_800 (task7 real_cl), passed via CLI checkpoint.load_path. +libero_goal_suites_task8_real_cl_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [8]), # NEW task8 real demos (50) + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), # OLD tasks 0-7 REAL demos + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5, 6, 7]), # OLD tasks 0-7 real replay + max_replay_demos=50, # <=50 per old task -> 349 total + current_task_step_fraction=0.5, # 50/50 new(task8) vs replay (user: all CL runs 1:1) + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, + use_stronger_image_aug=True, + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) + + +# TASK8 REAL-replay CL experiment: inherits cl_stage (800-step schedule: 50 warmup 0->5e-5, 750 decay +# ->1.5e-5), only swaps dataset + job.name. Warm-start (real_cl7 iter_800) via CLI checkpoint.load_path. +cosmos_predict2_2b_480p_libero_goal_task8_real_cl = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task8_real_cl_dataset, + sampler=dict( + dataset=libero_goal_suites_task8_real_cl_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task8_real_cl", + ), + ) +) + + +# TASK8 DEMOCHAN7 continual-learning dataset: NEW task8 real demos (50) + task0-7 replay generated +# from task8 initial states by the demochan7 policy (old_task8/replay_curated, 10 demos/task = 80). +# Replay uses the demonstration channel only; trajectories_wv is intentionally excluded. +libero_goal_suites_task8_demochan7_cl_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [8]), + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "old_task8", "replay_curated"), + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5, 6, 7]), + max_replay_demos=10, + current_task_step_fraction=0.5, + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, + use_stronger_image_aug=True, + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) + + +# TASK8 DEMOCHAN7 CL experiment: inherits the native 800-step cl_stage LR schedule. Warm-start from +# task7_cl_from_demochan800 iter_800 is supplied by the launcher with load_training_state=False. +cosmos_predict2_2b_480p_libero_goal_task8_demochan7_cl = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task8_demochan7_cl_dataset, + sampler=dict( + dataset=libero_goal_suites_task8_demochan7_cl_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task8_demochan7_cl", + ), + ) +) + + +# TASK8 forgetting-curated repair: task8 real demos + the 10 highest-forgetting, IDM-filtered +# generated trajectories from each old task (task0-7, 80 total). The replay directory contains only +# canonical demo-format HDF5 symlinks; task/seed subdirectories prevent basename collisions. +libero_goal_suites_task8_forget_cl_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [8]), + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "new_task8", "forget_replay_top10"), + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5, 6, 7]), + max_replay_demos=10, + current_task_step_fraction=0.5, + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, + use_stronger_image_aug=True, + demonstration_sampling_prob=1.0, + return_value_function_returns=True, + gamma=0.99, +) + + +# TASK8 forgetting-curated CL experiment. The launcher warm-starts from the 100-step task8-only FT +# and keeps cl_stage's original 800-step LR schedule (50-step warmup), matching Task7 Round1. +cosmos_predict2_2b_480p_libero_goal_task8_forget_cl = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task8_forget_cl_dataset, + sampler=dict( + dataset=libero_goal_suites_task8_forget_cl_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task8_forget_cl", + ), + upload_reproducible_setup=False, + ) +) + + +# TASK8 language-aware forgetting replay: the replay directory is produced by +# select_language_aware_replay.py. It keeps the same total budget as uniform top-10 replay +# (8 old tasks * 10 = 80), guarantees at least 6 trajectories per old task, and assigns the +# remainder from Task8-vs-old-task T5 language similarity. The directory itself is the exact +# selection, so max_replay_demos only needs to be above the largest dynamic quota. +libero_goal_suites_task8_forget_lang80_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [8]), + replay_data_dir=os.path.join(BASE_DATASETS_DIR, "new_task8", "forget_replay_language80"), + replay_tasks=get_replay_tasks("libero_goal", [0, 1, 2, 3, 4, 5, 6, 7]), + max_replay_demos=20, + current_task_step_fraction=0.5, + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, + use_stronger_image_aug=True, + demonstration_sampling_prob=1.0, + return_value_function_returns=True, + gamma=0.99, +) + + +cosmos_predict2_2b_480p_libero_goal_task8_forget_lang80 = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task8_forget_lang80_dataset, + sampler=dict( + dataset=libero_goal_suites_task8_forget_lang80_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task8_forget_lang80", + ), + upload_reproducible_setup=False, + ) +) + + +# User-editable DEMOCHAN trainer dataset. train_demochan.py supplies these paths and task IDs through +# environment variables before this module is imported, avoiding fragile CLI replacement of the dataset object. +_demochan_user_current_dir = os.environ.get( + "DEMOCHAN_CURRENT_DATA_DIR", + os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), +) +_demochan_user_current_ids = _env_int_list("DEMOCHAN_CURRENT_TASK_IDS", [8]) +_demochan_user_replay_dir = os.environ.get("DEMOCHAN_REPLAY_DATA_DIR", "").strip() +_demochan_user_replay_ids = _env_int_list("DEMOCHAN_REPLAY_TASK_IDS", list(range(8))) +_demochan_hierarchical_sampling = _env_bool("DEMOCHAN_HIERARCHICAL_SAMPLING") +_demochan_fraction_schedule = _env_fraction_schedule( + "DEMOCHAN_CURRENT_TASK_FRACTION_SCHEDULE", [(0, 0.5)] +) +_demochan_sampling_steps_raw = os.environ.get("DEMOCHAN_SAMPLING_OPTIMIZER_STEPS", "").strip() +_demochan_sampling_steps = ( + int(_demochan_sampling_steps_raw) if _demochan_sampling_steps_raw else None +) +_demochan_user_dataset_kwargs = dict( + data_dir=_demochan_user_current_dir, + # Continual stages must share the base policy's normalization coordinates even + # when replay_data_dir is refreshed between launches. + dataset_stats_dir=_demochan_user_current_dir, + current_tasks_ids=get_replay_tasks("libero_goal", _demochan_user_current_ids), + t5_text_embeddings_path=os.environ.get( + "DEMOCHAN_T5_PATH", + os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl"), + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, + use_stronger_image_aug=True, + demonstration_sampling_prob=1.0, + return_value_function_returns=True, + gamma=0.99, +) +if _demochan_user_replay_dir: + _demochan_user_dataset_kwargs.update( + replay_data_dir=_demochan_user_replay_dir, + replay_tasks=get_replay_tasks("libero_goal", _demochan_user_replay_ids), + max_replay_demos=int(os.environ.get("DEMOCHAN_MAX_REPLAY_DEMOS", "10")), + current_task_step_fraction=( + None + if _demochan_hierarchical_sampling + else float(os.environ.get("DEMOCHAN_CURRENT_TASK_STEP_FRACTION", "0.5")) + ), + hierarchical_sampling=_demochan_hierarchical_sampling, + current_task_fraction_schedule=_demochan_fraction_schedule, + coverage_window_steps=int(os.environ.get("DEMOCHAN_COVERAGE_WINDOW_STEPS", "10")), + sampling_optimizer_steps=_demochan_sampling_steps, + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + ) +libero_goal_suites_demochan_user_dataset = L(LIBERODataset)(**_demochan_user_dataset_kwargs) + + +cosmos_predict2_2b_480p_libero_goal_demochan_user = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_demochan_user_dataset, + sampler=dict( + dataset=libero_goal_suites_demochan_user_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_demochan_user", + ), + ) +) + + +# TASK7 pure fine-tune (NO replay): task7 (turn_on_the_stove) REAL demos ONLY. Baseline for how much a +# replay-free task7 ft forgets task0-6 (task7 analog of the 100-step ft6). No replay_data_dir -> no replay. +libero_goal_suites_task7_ft_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [7]), + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, + use_stronger_image_aug=True, + demonstration_sampling_prob=1.0, + return_value_function_returns=True, + gamma=0.99, +) + + +# TASK7 pure-ft experiment: inherits cl_stage, dataset = task7 only (no replay), job.name. +# Warm-start (fcv2 = forget_cl_v2 iter_350) passed via CLI checkpoint.load_path. +cosmos_predict2_2b_480p_libero_goal_task7_ft = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task7_ft_dataset, + sampler=dict( + dataset=libero_goal_suites_task7_ft_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task7_ft", + ), + ) +) + + +# TASK8 pure fine-tune (NO replay): task8 (put_the_bowl_on_the_plate) REAL demos ONLY. +libero_goal_suites_task8_ft_dataset = L(LIBERODataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "libero_goal_regen"), + current_tasks_ids=get_replay_tasks("libero_goal", [8]), + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "LIBERO-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=16, + use_image_aug=True, + use_wrist_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, + use_stronger_image_aug=True, + demonstration_sampling_prob=1.0, + return_value_function_returns=True, + gamma=0.99, +) + + +# TASK8 pure-ft experiment: inherits cl_stage, dataset = task8 only (no replay), job.name. +# Warm-start (fcv2 iter_350) is passed via CLI checkpoint.load_path. +cosmos_predict2_2b_480p_libero_goal_task8_ft = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task8_ft_dataset, + sampler=dict( + dataset=libero_goal_suites_task8_ft_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task8_ft", + ), + ) +) + + +# CONTROLLED regen_cl channel-swap: same cl_stage recipe, dataset = regen WAM rollouts via DEMO channel. +cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + dataloader_train=dict( + dataset=libero_goal_suites_task6_regen_cl_demochan_dataset, + sampler=dict( + dataset=libero_goal_suites_task6_regen_cl_demochan_dataset, + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan", + ), + ) +) + + + + + + + + +# Inference version +cosmos_predict2_2b_480p_libero__inference_only = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero", + "_self_", + ], + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + sde=L(HybridEDMSDE)( + sigma_max=80, + sigma_min=4, + ) + ) + ), + job=dict( + group="cosmos_v2_inference", + name="cosmos_predict2_2b_480p_libero__inference_only", + ), + ) +) + +cosmos_predict2_2b_480p_libero_base_stage_inference_only = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_base_stage", + "_self_", + ], + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + sde=L(HybridEDMSDE)( + sigma_max=80, + sigma_min=4, + ) + ) + ), + job=dict( + group="cosmos_v2_inference", + name="cosmos_predict2_2b_480p_libero_base_stage_inference_only", + ), + ) +) + +cosmos_predict2_2b_480p_libero_cl_stage_inference_only = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero_goal_cl_stage", + "_self_", + ], + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + sde=L(HybridEDMSDE)( + sigma_max=80, + sigma_min=4, + ) + ) + ), + job=dict( + group="cosmos_v2_inference", + name="cosmos_predict2_2b_480p_libero_cl_stage_inference_only", + ), + ) +) + + +# *** Main checkpoint *** +robocasa_50_demos_per_task_dataset = L(RoboCasaDataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "RoboCasa-Cosmos-Policy", "success_only"), # Successful demos + t5_text_embeddings_path=os.path.join( + BASE_DATASETS_DIR, "RoboCasa-Cosmos-Policy", "success_only", "t5_embeddings.pkl" + ), + chunk_size=32, + use_image_aug=True, + use_wrist_images=True, + use_third_person_images=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + use_stronger_image_aug=True, + rollout_data_dir=os.path.join( + BASE_DATASETS_DIR, "RoboCasa-Cosmos-Policy", "all_episodes" + ), # All demo rollouts (successes + failures) + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.99, +) +cosmos_predict2_2b_480p_robocasa_50_demos_per_task = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero", + "_self_", + ], + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + state_t=11, # Latent temporal dim (blank, proprio, wrist image, primary image, secondary image, action, future proprio, future wrist image, future primary image, future secondary image, value) + min_num_conditional_frames=5, # 1 blank, 4 conditioning (proprio, wrist image, primary image, secondary image) + max_num_conditional_frames=5, # 1 blank, 4 conditioning (proprio, wrist image, primary image, secondary image) + tokenizer=dict( + chunk_duration=41, # 1 blank + 40 images (4 proprio, 4 wrist image, 4 primary image, 4 secondary image, 4 action, 4 future proprio, 4 future wrist, 4 future primary, 4 future secondary, 4 value) + ), + ), + ), + dataloader_train=L(DataLoader)( + num_workers=8, + persistent_workers=True, + pin_memory=True, + dataset=robocasa_50_demos_per_task_dataset, + sampler=L(DistributedSampler)( + dataset=robocasa_50_demos_per_task_dataset, + num_replicas=L(parallel_state.get_data_parallel_world_size)(), + rank=L(parallel_state.get_data_parallel_rank)(), + shuffle=True, + seed=0, + ), + batch_size=25, + drop_last=True, + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_robocasa_50_demos_per_task", + ), + ) +) +# Inference version +cosmos_predict2_2b_480p_robocasa_50_demos_per_task__inference = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_robocasa_50_demos_per_task", + "_self_", + ], + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + sde=L(HybridEDMSDE)( + sigma_max=80, + sigma_min=4, + ) + ) + ), + job=dict( + group="cosmos_v2_inference", + name="cosmos_predict2_2b_480p_robocasa_50_demos_per_task__inference", + ), + ) +) + + +# *** Main checkpoint *** +aloha_cosmos_policy_dataset_185_demos = L(ALOHADataset)( + data_dir=os.path.join(BASE_DATASETS_DIR, "ALOHA-Cosmos-Policy", "preprocessed"), + t5_text_embeddings_path=os.path.join(BASE_DATASETS_DIR, "ALOHA-Cosmos-Policy", "preprocessed", "t5_embeddings.pkl"), + chunk_size=50, + use_image_aug=True, + use_stronger_image_aug=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + treat_demos_as_success_rollouts=True, # Include demos as success rollouts + demonstration_sampling_prob=0.5, + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.998, # Higher gamma for ALOHA because episodes can have up to 1.5-2.0K steps # (s, a, s', v) +) +cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80 = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_libero", + "_self_", + ], + scheduler=dict( + # LR decay for 20K steps in cycle #1, then decay by 5x and stay constant forever in cycle #2 + cycle_lengths=[20000, 100000000000000], + warm_up_steps=[2000, 0], + f_start=[1e-6, 0.06], + f_max=[1.0, 0.06], + f_min=[0.3, 0.06], + ), + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + state_t=11, # Latent temporal dim (blank, proprio, left wrist, right wrist, primary, action, future proprio, future left wrist, future right wrist, future primary, value) + min_num_conditional_frames=5, # 1 blank, 4 conditioning (proprio, left wrist, right wrist, primary) + max_num_conditional_frames=5, # 1 blank, 4 conditioning (proprio, left wrist, right wrist, primary) + tokenizer=dict( + chunk_duration=41, # 1 blank + 40 images (4 proprio, 4 left wrist image, 4 right wrist image, 4 primary image, 4 action, 4 future proprio, 4 future left wrist, 4 future right wrist, 4 future primary, 4 value) + ), + ), + ), + dataloader_train=L(DataLoader)( + num_workers=12, + persistent_workers=True, + pin_memory=True, + dataset=aloha_cosmos_policy_dataset_185_demos, + sampler=L(DistributedSampler)( + dataset=aloha_cosmos_policy_dataset_185_demos, + num_replicas=L(parallel_state.get_data_parallel_world_size)(), + rank=L(parallel_state.get_data_parallel_rank)(), + shuffle=True, + seed=0, + ), + batch_size=25, + drop_last=True, + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80", + ), + ) +) + + +# Inference version +cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__inference_only = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80", + "_self_", + ], + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + sde=L(HybridEDMSDE)( + sigma_max=80, + sigma_min=4, + ) + ) + ), + job=dict( + group="cosmos_v2_inference", + name="cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__inference_only", + ), + ) +) + + +# ALOHA planning model +# Dataset: 648 rollouts from evaluations with Cosmos Policy, pi05, pi0, OpenVLA-OFT+, Diffusion Policy +# NOTE: This rollouts dataset is not released; you will need to replace `rollout_data_dir` below with your own rollouts dataset +aloha_2025_09_18__648_rollouts__cosmos_policy__pi05__pi0__openvla_oft__diffusion_policy__dataset = L( + ALOHADataset +)( + data_dir=os.path.join(BASE_DATASETS_DIR, "ALOHA-Cosmos-Policy", "preprocessed"), + t5_text_embeddings_path=os.path.join(BASE_DATASETS_DIR, "ALOHA-Cosmos-Policy", "preprocessed", "t5_embeddings.pkl"), + chunk_size=50, + use_image_aug=True, + use_stronger_image_aug=True, + use_proprio=True, + normalize_proprio=True, + normalize_actions=True, + num_duplicates_per_image=4, # WAN 2.1 tokenizer: 4 images per latent frame + treat_demos_as_success_rollouts=False, # Don't include demos as success rollouts because they have a fixed episode length + we want to focus on real policy rollouts + demonstration_sampling_prob=0.1, # Smaller demonstration sampling prob - more emphasis on rollouts + success_rollout_sampling_prob=0.5, + return_value_function_returns=True, + gamma=0.998, # Higher gamma for ALOHA because episodes can have up to 1.5-2.0K steps # (s, a, s', v) + rollout_data_dir=os.path.join(BASE_DATASETS_DIR, "PATH/TO/YOUR/ROLLOUTS/DATASET"), # JPEG images + use_jpeg_for_rollouts=True, # JPEG images +) +cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80", + "_self_", + ], + checkpoint=dict( + # Resume from 50K checkpoint of base Cosmos Policy run + # Do not download the unrelated ALOHA checkpoint while importing + # LIBERO experiment configs. Resolve or override this URI only + # when launching the ALOHA resume experiment. + load_path="hf://nvidia/Cosmos-Policy-ALOHA-Predict2-2B/Cosmos-Policy-ALOHA-Predict2-2B.pt", + ), + scheduler=dict( + # LR decay for 15K steps in cycle #1, then decay by 5x and stay constant forever in cycle #2 + cycle_lengths=[15000, 100000000000000], + warm_up_steps=[1500, 0], + f_start=[1e-6, 0.06], + f_max=[1.0, 0.06], + f_min=[0.3, 0.06], + ), + dataloader_train=L(DataLoader)( + num_workers=12, + persistent_workers=True, + pin_memory=True, + dataset=aloha_2025_09_18__648_rollouts__cosmos_policy__pi05__pi0__openvla_oft__diffusion_policy__dataset, + sampler=L(DistributedSampler)( + dataset=aloha_2025_09_18__648_rollouts__cosmos_policy__pi05__pi0__openvla_oft__diffusion_policy__dataset, + num_replicas=L(parallel_state.get_data_parallel_world_size)(), + rank=L(parallel_state.get_data_parallel_rank)(), + shuffle=True, + seed=0, + ), + batch_size=25, + drop_last=True, + ), + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + mask_current_state_action_for_value_prediction=True, # Use input masking to mask out irrelevant inputs (current state and action) during value prediction + ), + ), + job=dict( + group="cosmos_v2_finetune", + name="cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func", + ), + ) +) +# Inference version +cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func__inference_only = LazyDict( + dict( + defaults=[ + "/experiment/cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func", + "_self_", + ], + model=L(CosmosPolicyVideo2WorldModel)( + config=dict( + sde=L(HybridEDMSDE)( + sigma_max=80, + sigma_min=4, + ) + ) + ), + job=dict( + group="cosmos_v2_inference", + name="cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func__inference_only", + ), + ) +) + + +def register_configs(): + cs = ConfigStore.instance() + # Register the experiments + for _item in [ + # LIBERO + cosmos_predict2_2b_480p_libero, # *** Main checkpoint *** + # LIBERO Base Stage + cosmos_predict2_2b_480p_libero_base_stage, + cosmos_predict2_2b_480p_libero_base_stage_inference_only, + # LIBERO CL Stage + cosmos_predict2_2b_480p_libero_cl_stage, + cosmos_predict2_2b_480p_libero_cl_stage_inference_only, + cosmos_predict2_2b_480p_libero__inference_only, + # LIBERO CL fine-tune on task6 (put_the_cream_cheese_in_the_bowl) + cosmos_predict2_2b_480p_libero_goal_task6_ft, + # LIBERO task7 CL on forgetting-curated replay (top-10 most-forgotten per old task) + cosmos_predict2_2b_480p_libero_goal_task7_forget_cl, + # round 2: replay re-picked against the round-1 model + cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2, + # LIBERO pure fine-tune on the 8 merged task3 dream trajectories (seg1+seg2 joined) + cosmos_predict2_2b_480p_libero_goal_task3_merged_ft, + # LIBERO REGEN CL on task6 (task6 real demos + task0-5 rollout replay) + cosmos_predict2_2b_480p_libero_goal_task6_regen_cl, + # LIBERO FORGETTING-CURATED CL on task6 (forgetting-picked replay, 50/50 new:replay) + cosmos_predict2_2b_480p_libero_goal_task6_forget_cl, + # LIBERO regen_cl channel-swap control (regen WAM rollouts via DEMO channel, 50/50) + cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan, + # LIBERO V2 forgetting CL (iter_350-student selection + constant LR 3.5e-5) + cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2, + # LIBERO task7 (turn_on_the_stove) CL: new task7 + demochan-gen task0-6 replay (old_task7) + cosmos_predict2_2b_480p_libero_goal_task7_cl, + # LIBERO task7 CL with REAL human-demo replay of task0-6 (cap 50/task = 299) + cosmos_predict2_2b_480p_libero_goal_task7_real_cl, + # LIBERO task7 pure fine-tune (no replay) baseline + cosmos_predict2_2b_480p_libero_goal_task7_ft, + # LIBERO task8 pure fine-tune (no replay) baseline + cosmos_predict2_2b_480p_libero_goal_task8_ft, + # LIBERO task8 (put_the_bowl_on_the_plate) CL with REAL human-demo replay of task0-7 (cap 50/task = 349) + cosmos_predict2_2b_480p_libero_goal_task8_real_cl, + # LIBERO task8 CL from demochan7 with old_task8 generated demo replay (10/task = 80) + cosmos_predict2_2b_480p_libero_goal_task8_demochan7_cl, + # LIBERO task8 repair from pure-FT iter100 with forgetting-curated replay (10/task = 80) + cosmos_predict2_2b_480p_libero_goal_task8_forget_cl, + # LIBERO task8 repair with language-aware dynamic old-task quotas (80 total, >=6/task) + cosmos_predict2_2b_480p_libero_goal_task8_forget_lang80, + # User-configurable demochan trainer (paths/task IDs supplied by train_demochan.py env) + cosmos_predict2_2b_480p_libero_goal_demochan_user, + + + # RoboCasa + cosmos_predict2_2b_480p_robocasa_50_demos_per_task, # *** Main checkpoint *** + cosmos_predict2_2b_480p_robocasa_50_demos_per_task__inference, + # ALOHA + cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80, # *** Main checkpoint *** + cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__inference_only, + cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func, # ALOHA planning model + cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func__inference_only, + ]: + experiment_name = _item["job"]["name"] + log.info(f"Registering experiment: {experiment_name}") + cs.store( + group="experiment", + package="_global_", + name=experiment_name, + node=_item, + ) diff --git a/REGEN-main/cosmos_policy/config/experiment/libero_task_constants.py b/REGEN-main/cosmos_policy/config/experiment/libero_task_constants.py new file mode 100644 index 0000000000000000000000000000000000000000..a78175c9b74a3111c867f508901fb4fa9346d0c3 --- /dev/null +++ b/REGEN-main/cosmos_policy/config/experiment/libero_task_constants.py @@ -0,0 +1,70 @@ +"""Task-id mappings for LIBERO-Cosmos-Policy suites. + +Supports both naming styles: +- success_only folders: ``libero_goal_regen`` (etc.) +- all_episodes filenames: ``libero_goal`` (etc.) +""" + +from __future__ import annotations + + +LIBERO_SUITE_TASK_ID_TO_DESCRIPTION: dict[str, dict[int, str]] = { + "libero_spatial": { + 0: "pick_up_the_black_bowl_between_the_plate_and_the_ramekin_and_place_it_on_the_plate", + 1: "pick_up_the_black_bowl_next_to_the_ramekin_and_place_it_on_the_plate", + 2: "pick_up_the_black_bowl_from_table_center_and_place_it_on_the_plate", + 3: "pick_up_the_black_bowl_on_the_cookie_box_and_place_it_on_the_plate", + 4: "pick_up_the_black_bowl_in_the_top_drawer_of_the_wooden_cabinet_and_place_it_on_the_plate", + 5: "pick_up_the_black_bowl_on_the_ramekin_and_place_it_on_the_plate", + 6: "pick_up_the_black_bowl_next_to_the_cookie_box_and_place_it_on_the_plate", + 7: "pick_up_the_black_bowl_on_the_stove_and_place_it_on_the_plate", + 8: "pick_up_the_black_bowl_next_to_the_plate_and_place_it_on_the_plate", + 9: "pick_up_the_black_bowl_on_the_wooden_cabinet_and_place_it_on_the_plate", + }, + "libero_object": { + 0: "pick_up_the_alphabet_soup_and_place_it_in_the_basket", + 1: "pick_up_the_cream_cheese_and_place_it_in_the_basket", + 2: "pick_up_the_salad_dressing_and_place_it_in_the_basket", + 3: "pick_up_the_bbq_sauce_and_place_it_in_the_basket", + 4: "pick_up_the_ketchup_and_place_it_in_the_basket", + 5: "pick_up_the_tomato_sauce_and_place_it_in_the_basket", + 6: "pick_up_the_butter_and_place_it_in_the_basket", + 7: "pick_up_the_milk_and_place_it_in_the_basket", + 8: "pick_up_the_chocolate_pudding_and_place_it_in_the_basket", + 9: "pick_up_the_orange_juice_and_place_it_in_the_basket", + }, + "libero_goal": { + 0: "open_the_middle_drawer_of_the_cabinet", + 1: "put_the_bowl_on_the_stove", + 2: "put_the_wine_bottle_on_top_of_the_cabinet", + 3: "open_the_top_drawer_and_put_the_bowl_inside", + 4: "put_the_bowl_on_top_of_the_cabinet", + 5: "push_the_plate_to_the_front_of_the_stove", + 6: "put_the_cream_cheese_in_the_bowl", + 7: "turn_on_the_stove", + 8: "put_the_bowl_on_the_plate", + 9: "put_the_wine_bottle_on_the_rack", + }, + "libero_10": { + 0: "put_both_the_alphabet_soup_and_the_tomato_sauce_in_the_basket", + 1: "put_both_the_cream_cheese_box_and_the_butter_in_the_basket", + 2: "turn_on_the_stove_and_put_the_moka_pot_on_it", + 3: "put_the_black_bowl_in_the_bottom_drawer_of_the_cabinet_and_close_it", + 4: "put_the_white_mug_on_the_left_plate_and_put_the_yellow_and_white_mug_on_the_right_plate", + 5: "pick_up_the_book_and_place_it_in_the_back_compartment_of_the_caddy", + 6: "put_the_white_mug_on_the_plate_and_put_the_chocolate_pudding_to_the_right_of_the_plate", + 7: "put_both_the_alphabet_soup_and_the_cream_cheese_box_in_the_basket", + 8: "put_both_moka_pots_on_the_stove", + 9: "put_the_yellow_and_white_mug_in_the_microwave_and_close_it", + }, +} + + +def task_description_to_replay_name(task_description: str) -> str: + return task_description.replace("_", " ") + + +def get_replay_tasks(suite_name: str, task_ids: list[int]) -> list[str]: + id_to_task = LIBERO_SUITE_TASK_ID_TO_DESCRIPTION[suite_name] + return [task_description_to_replay_name(id_to_task[task_id]) for task_id in task_ids] + diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_07_31-14_22_22--ft6_iter200_t1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_07_31-14_22_22--ft6_iter200_t1.txt new file mode 100644 index 0000000000000000000000000000000000000000..92c9ad46e5e19020208eaa44a55ba2501f3c6383 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_07_31-14_22_22--ft6_iter200_t1.txt @@ -0,0 +1,3365 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_1000step/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6_iter200_t1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=True, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 1.505 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4935 +t=10: Selected seed 195 with value = 0.4935 +Query 1/1: Action query time = 1.071 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5779 +t=26: Selected seed 195 with value = 0.5779 +Query 1/1: Action query time = 1.135 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6548 +t=42: Selected seed 195 with value = 0.6548 +Query 1/1: Action query time = 1.181 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7433 +t=58: Selected seed 195 with value = 0.7433 +Query 1/1: Action query time = 1.140 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8334 +t=74: Selected seed 195 with value = 0.8334 +Query 1/1: Action query time = 0.962 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9478 +t=90: Selected seed 195 with value = 0.9478 +Query 1/1: Action query time = 0.979 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.957 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.949 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=218: Selected seed 195 with value = 0.9968 +Query 1/1: Action query time = 0.959 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9881 +t=234: Selected seed 195 with value = 0.9881 +Query 1/1: Action query time = 0.953 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9959 +t=250: Selected seed 195 with value = 0.9959 +Query 1/1: Action query time = 0.949 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9913 +t=266: Selected seed 195 with value = 0.9913 +Query 1/1: Action query time = 0.952 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9186 +t=282: Selected seed 195 with value = 0.9186 +Query 1/1: Action query time = 0.952 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9335 +t=298: Selected seed 195 with value = 0.9335 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 0.962 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4790 +t=10: Selected seed 195 with value = 0.4790 +Query 1/1: Action query time = 0.959 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5499 +t=26: Selected seed 195 with value = 0.5499 +Query 1/1: Action query time = 0.966 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6561 +t=42: Selected seed 195 with value = 0.6561 +Query 1/1: Action query time = 0.960 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7555 +t=58: Selected seed 195 with value = 0.7555 +Query 1/1: Action query time = 0.968 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6628 +t=74: Selected seed 195 with value = 0.6628 +Query 1/1: Action query time = 0.975 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6780 +t=90: Selected seed 195 with value = 0.6780 +Query 1/1: Action query time = 0.977 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7259 +t=106: Selected seed 195 with value = 0.7259 +Query 1/1: Action query time = 0.958 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8382 +t=122: Selected seed 195 with value = 0.8382 +Query 1/1: Action query time = 0.976 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8573 +t=138: Selected seed 195 with value = 0.8573 +Query 1/1: Action query time = 0.960 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7821 +t=154: Selected seed 195 with value = 0.7821 +Query 1/1: Action query time = 0.960 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9359 +t=170: Selected seed 195 with value = 0.9359 +Query 1/1: Action query time = 0.968 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9947 +t=186: Selected seed 195 with value = 0.9947 +Query 1/1: Action query time = 0.966 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9540 +t=202: Selected seed 195 with value = 0.9540 +Query 1/1: Action query time = 0.959 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9760 +t=218: Selected seed 195 with value = 0.9760 +Query 1/1: Action query time = 0.959 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9781 +t=234: Selected seed 195 with value = 0.9781 +Query 1/1: Action query time = 0.960 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9768 +t=250: Selected seed 195 with value = 0.9768 +Query 1/1: Action query time = 0.966 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9718 +t=266: Selected seed 195 with value = 0.9718 +Query 1/1: Action query time = 1.186 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9504 +t=282: Selected seed 195 with value = 0.9504 +Query 1/1: Action query time = 1.149 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7955 +t=298: Selected seed 195 with value = 0.7955 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 0.981 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4742 +t=10: Selected seed 195 with value = 0.4742 +Query 1/1: Action query time = 0.976 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5625 +t=26: Selected seed 195 with value = 0.5625 +Query 1/1: Action query time = 0.959 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6550 +t=42: Selected seed 195 with value = 0.6550 +Query 1/1: Action query time = 0.960 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8003 +t=58: Selected seed 195 with value = 0.8003 +Query 1/1: Action query time = 0.977 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6526 +t=74: Selected seed 195 with value = 0.6526 +Query 1/1: Action query time = 0.952 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9173 +t=90: Selected seed 195 with value = 0.9173 +Query 1/1: Action query time = 0.975 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9115 +t=106: Selected seed 195 with value = 0.9115 +Query 1/1: Action query time = 0.961 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=138: Selected seed 195 with value = 0.9968 +Query 1/1: Action query time = 1.009 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9795 +t=154: Selected seed 195 with value = 0.9795 +Query 1/1: Action query time = 0.962 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8901 +t=170: Selected seed 195 with value = 0.8901 +Query 1/1: Action query time = 1.000 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8235 +t=186: Selected seed 195 with value = 0.8235 +Query 1/1: Action query time = 0.979 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8611 +t=202: Selected seed 195 with value = 0.8611 +Query 1/1: Action query time = 0.965 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9416 +t=218: Selected seed 195 with value = 0.9416 +Query 1/1: Action query time = 0.974 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9791 +t=234: Selected seed 195 with value = 0.9791 +Query 1/1: Action query time = 0.978 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 0.982 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4782 +t=10: Selected seed 195 with value = 0.4782 +Query 1/1: Action query time = 0.966 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5717 +t=26: Selected seed 195 with value = 0.5717 +Query 1/1: Action query time = 0.964 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6638 +t=42: Selected seed 195 with value = 0.6638 +Query 1/1: Action query time = 0.953 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8394 +t=58: Selected seed 195 with value = 0.8394 +Query 1/1: Action query time = 0.964 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8304 +t=74: Selected seed 195 with value = 0.8304 +Query 1/1: Action query time = 0.968 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9048 +t=90: Selected seed 195 with value = 0.9048 +Query 1/1: Action query time = 0.971 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9364 +t=106: Selected seed 195 with value = 0.9364 +Query 1/1: Action query time = 0.960 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.953 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=170: Selected seed 195 with value = 0.9968 +Query 1/1: Action query time = 0.953 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9935 +t=186: Selected seed 195 with value = 0.9935 +Query 1/1: Action query time = 0.965 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9592 +t=218: Selected seed 195 with value = 0.9592 +Query 1/1: Action query time = 0.965 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9822 +t=234: Selected seed 195 with value = 0.9822 +Query 1/1: Action query time = 0.970 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9854 +t=250: Selected seed 195 with value = 0.9854 +Query 1/1: Action query time = 0.957 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9892 +t=266: Selected seed 195 with value = 0.9892 +Query 1/1: Action query time = 0.968 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=282: Selected seed 195 with value = 0.9912 +Query 1/1: Action query time = 0.960 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=298: Selected seed 195 with value = 0.9912 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 0.976 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4952 +t=10: Selected seed 195 with value = 0.4952 +Query 1/1: Action query time = 0.972 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5434 +t=26: Selected seed 195 with value = 0.5434 +Query 1/1: Action query time = 0.990 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6510 +t=42: Selected seed 195 with value = 0.6510 +Query 1/1: Action query time = 1.043 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7910 +t=58: Selected seed 195 with value = 0.7910 +Query 1/1: Action query time = 1.037 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8348 +t=74: Selected seed 195 with value = 0.8348 +Query 1/1: Action query time = 1.059 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9552 +t=90: Selected seed 195 with value = 0.9552 +Query 1/1: Action query time = 1.065 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.119 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.189 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.198 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.152 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.135 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=186: Selected seed 195 with value = 0.9928 +Query 1/1: Action query time = 1.101 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9911 +t=202: Selected seed 195 with value = 0.9911 +Query 1/1: Action query time = 1.057 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.045 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.039 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=250: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 0.995 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=266: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 1.005 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=282: Selected seed 195 with value = 0.9928 +Query 1/1: Action query time = 0.983 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9947 +t=298: Selected seed 195 with value = 0.9947 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=5--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=5--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 5 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 0.966 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4700 +t=10: Selected seed 195 with value = 0.4700 +Query 1/1: Action query time = 0.963 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5496 +t=26: Selected seed 195 with value = 0.5496 +Query 1/1: Action query time = 0.956 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6720 +t=42: Selected seed 195 with value = 0.6720 +Query 1/1: Action query time = 0.960 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8888 +t=58: Selected seed 195 with value = 0.8888 +Query 1/1: Action query time = 0.973 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8882 +t=74: Selected seed 195 with value = 0.8882 +Query 1/1: Action query time = 0.957 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9866 +t=90: Selected seed 195 with value = 0.9866 +Query 1/1: Action query time = 0.959 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.959 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.961 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.991 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.959 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.952 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9805 +t=202: Selected seed 195 with value = 0.9805 +Query 1/1: Action query time = 0.961 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9695 +t=218: Selected seed 195 with value = 0.9695 +Query 1/1: Action query time = 0.976 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9933 +t=234: Selected seed 195 with value = 0.9933 +Query 1/1: Action query time = 0.959 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9935 +t=250: Selected seed 195 with value = 0.9935 +Query 1/1: Action query time = 0.966 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=266: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 0.952 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=282: Selected seed 195 with value = 0.9962 +Query 1/1: Action query time = 0.956 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=6--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=6--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 6 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 0.974 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4215 +t=10: Selected seed 195 with value = 0.4215 +Query 1/1: Action query time = 0.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5547 +t=26: Selected seed 195 with value = 0.5547 +Query 1/1: Action query time = 0.958 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6449 +t=42: Selected seed 195 with value = 0.6449 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8077 +t=58: Selected seed 195 with value = 0.8077 +Query 1/1: Action query time = 0.962 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8504 +t=74: Selected seed 195 with value = 0.8504 +Query 1/1: Action query time = 0.958 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9544 +t=90: Selected seed 195 with value = 0.9544 +Query 1/1: Action query time = 0.959 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=138: Selected seed 195 with value = 0.9982 +Query 1/1: Action query time = 0.969 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9874 +t=154: Selected seed 195 with value = 0.9874 +Query 1/1: Action query time = 0.962 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9774 +t=170: Selected seed 195 with value = 0.9774 +Query 1/1: Action query time = 0.968 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9796 +t=186: Selected seed 195 with value = 0.9796 +Query 1/1: Action query time = 0.964 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9905 +t=202: Selected seed 195 with value = 0.9905 +Query 1/1: Action query time = 0.968 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=234: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 0.969 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9444 +t=250: Selected seed 195 with value = 0.9444 +Query 1/1: Action query time = 1.191 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8268 +t=266: Selected seed 195 with value = 0.8268 +Query 1/1: Action query time = 1.199 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9221 +t=282: Selected seed 195 with value = 0.9221 +Query 1/1: Action query time = 1.182 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=7--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=7--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 7 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 0.965 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4594 +t=10: Selected seed 195 with value = 0.4594 +Query 1/1: Action query time = 0.959 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5835 +t=26: Selected seed 195 with value = 0.5835 +Query 1/1: Action query time = 0.991 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6459 +t=42: Selected seed 195 with value = 0.6459 +Query 1/1: Action query time = 0.974 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7424 +t=58: Selected seed 195 with value = 0.7424 +Query 1/1: Action query time = 0.961 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8344 +t=74: Selected seed 195 with value = 0.8344 +Query 1/1: Action query time = 0.968 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9505 +t=90: Selected seed 195 with value = 0.9505 +Query 1/1: Action query time = 0.971 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.961 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.970 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.960 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=202: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 0.959 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9434 +t=218: Selected seed 195 with value = 0.9434 +Query 1/1: Action query time = 0.964 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8402 +t=234: Selected seed 195 with value = 0.8402 +Query 1/1: Action query time = 0.972 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9299 +t=250: Selected seed 195 with value = 0.9299 +Query 1/1: Action query time = 0.967 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.960 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.959 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=8--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=8--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 8 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 1.162 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4728 +t=10: Selected seed 195 with value = 0.4728 +Query 1/1: Action query time = 1.124 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5716 +t=26: Selected seed 195 with value = 0.5716 +Query 1/1: Action query time = 1.100 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6728 +t=42: Selected seed 195 with value = 0.6728 +Query 1/1: Action query time = 1.109 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7975 +t=58: Selected seed 195 with value = 0.7975 +Query 1/1: Action query time = 1.082 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8748 +t=74: Selected seed 195 with value = 0.8748 +Query 1/1: Action query time = 1.100 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9820 +t=90: Selected seed 195 with value = 0.9820 +Query 1/1: Action query time = 0.978 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=218: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 0.979 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9831 +t=234: Selected seed 195 with value = 0.9831 +Query 1/1: Action query time = 0.978 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9453 +t=250: Selected seed 195 with value = 0.9453 +Query 1/1: Action query time = 0.958 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8331 +t=266: Selected seed 195 with value = 0.8331 +Query 1/1: Action query time = 0.971 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9257 +t=282: Selected seed 195 with value = 0.9257 +Query 1/1: Action query time = 0.964 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 9 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 0.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4902 +t=10: Selected seed 195 with value = 0.4902 +Query 1/1: Action query time = 0.958 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5495 +t=26: Selected seed 195 with value = 0.5495 +Query 1/1: Action query time = 1.111 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6644 +t=42: Selected seed 195 with value = 0.6644 +Query 1/1: Action query time = 1.074 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8868 +t=58: Selected seed 195 with value = 0.8868 +Query 1/1: Action query time = 1.079 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8928 +t=74: Selected seed 195 with value = 0.8928 +Query 1/1: Action query time = 1.110 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9841 +t=90: Selected seed 195 with value = 0.9841 +Query 1/1: Action query time = 1.060 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.147 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.173 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.198 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.153 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.165 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.197 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.149 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9878 +t=218: Selected seed 195 with value = 0.9878 +Query 1/1: Action query time = 1.178 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9466 +t=234: Selected seed 195 with value = 0.9466 +Query 1/1: Action query time = 1.177 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9893 +t=250: Selected seed 195 with value = 0.9893 +Query 1/1: Action query time = 1.174 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9900 +t=266: Selected seed 195 with value = 0.9900 +Query 1/1: Action query time = 1.187 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9930 +t=282: Selected seed 195 with value = 0.9930 +Query 1/1: Action query time = 1.226 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=298: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=10--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=10--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 10 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 0.973 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4942 +t=10: Selected seed 195 with value = 0.4942 +Query 1/1: Action query time = 0.976 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5735 +t=26: Selected seed 195 with value = 0.5735 +Query 1/1: Action query time = 0.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6747 +t=42: Selected seed 195 with value = 0.6747 +Query 1/1: Action query time = 0.981 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8802 +t=58: Selected seed 195 with value = 0.8802 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8368 +t=74: Selected seed 195 with value = 0.8368 +Query 1/1: Action query time = 1.027 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9191 +t=90: Selected seed 195 with value = 0.9191 +Query 1/1: Action query time = 1.040 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.055 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.098 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.150 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.137 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.051 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=218: Selected seed 195 with value = 0.9999 +Query 1/1: Action query time = 0.971 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.953 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=11--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=11--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 11 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 0.968 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4972 +t=10: Selected seed 195 with value = 0.4972 +Query 1/1: Action query time = 0.976 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5469 +t=26: Selected seed 195 with value = 0.5469 +Query 1/1: Action query time = 0.963 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6675 +t=42: Selected seed 195 with value = 0.6675 +Query 1/1: Action query time = 0.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7492 +t=58: Selected seed 195 with value = 0.7492 +Query 1/1: Action query time = 0.959 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8319 +t=74: Selected seed 195 with value = 0.8319 +Query 1/1: Action query time = 0.958 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9568 +t=90: Selected seed 195 with value = 0.9568 +Query 1/1: Action query time = 0.960 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.951 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=170: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 0.983 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9816 +t=202: Selected seed 195 with value = 0.9816 +Query 1/1: Action query time = 0.968 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9339 +t=218: Selected seed 195 with value = 0.9339 +Query 1/1: Action query time = 0.982 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8498 +t=234: Selected seed 195 with value = 0.8498 +Query 1/1: Action query time = 0.970 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8518 +t=250: Selected seed 195 with value = 0.8518 +Query 1/1: Action query time = 0.977 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.960 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9762 +t=298: Selected seed 195 with value = 0.9762 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=12--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=12--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 12 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 0.961 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4961 +t=10: Selected seed 195 with value = 0.4961 +Query 1/1: Action query time = 0.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5470 +t=26: Selected seed 195 with value = 0.5470 +Query 1/1: Action query time = 0.968 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6722 +t=42: Selected seed 195 with value = 0.6722 +Query 1/1: Action query time = 0.971 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9285 +t=58: Selected seed 195 with value = 0.9285 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8312 +t=74: Selected seed 195 with value = 0.8312 +Query 1/1: Action query time = 0.969 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9315 +t=90: Selected seed 195 with value = 0.9315 +Query 1/1: Action query time = 0.971 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9854 +t=106: Selected seed 195 with value = 0.9854 +Query 1/1: Action query time = 0.967 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=170: Selected seed 195 with value = 0.9928 +Query 1/1: Action query time = 0.964 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=186: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 0.971 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9740 +t=202: Selected seed 195 with value = 0.9740 +Query 1/1: Action query time = 0.970 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9088 +t=218: Selected seed 195 with value = 0.9088 +Query 1/1: Action query time = 0.975 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8277 +t=234: Selected seed 195 with value = 0.8277 +Query 1/1: Action query time = 0.979 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8831 +t=250: Selected seed 195 with value = 0.8831 +Query 1/1: Action query time = 0.966 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.986 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=13--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=13--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 13 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 14... +Query 1/1: Action query time = 0.990 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4767 +t=10: Selected seed 195 with value = 0.4767 +Query 1/1: Action query time = 0.960 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5427 +t=26: Selected seed 195 with value = 0.5427 +Query 1/1: Action query time = 0.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6527 +t=42: Selected seed 195 with value = 0.6527 +Query 1/1: Action query time = 0.984 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8114 +t=58: Selected seed 195 with value = 0.8114 +Query 1/1: Action query time = 1.035 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8370 +t=74: Selected seed 195 with value = 0.8370 +Query 1/1: Action query time = 1.235 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9416 +t=90: Selected seed 195 with value = 0.9416 +Query 1/1: Action query time = 1.150 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.086 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.030 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.047 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=154: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 0.983 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9818 +t=170: Selected seed 195 with value = 0.9818 +Query 1/1: Action query time = 0.981 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9895 +t=186: Selected seed 195 with value = 0.9895 +Query 1/1: Action query time = 0.973 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=218: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 0.973 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=234: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 0.985 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.984 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=298: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=14--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=14--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 14 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 15... +Query 1/1: Action query time = 0.997 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4773 +t=10: Selected seed 195 with value = 0.4773 +Query 1/1: Action query time = 0.974 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5717 +t=26: Selected seed 195 with value = 0.5717 +Query 1/1: Action query time = 0.981 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6928 +t=42: Selected seed 195 with value = 0.6928 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7530 +t=58: Selected seed 195 with value = 0.7530 +Query 1/1: Action query time = 0.987 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8232 +t=74: Selected seed 195 with value = 0.8232 +Query 1/1: Action query time = 0.978 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9412 +t=90: Selected seed 195 with value = 0.9412 +Query 1/1: Action query time = 0.995 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.001 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.011 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.990 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9831 +t=218: Selected seed 195 with value = 0.9831 +Query 1/1: Action query time = 1.001 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9195 +t=234: Selected seed 195 with value = 0.9195 +Query 1/1: Action query time = 0.979 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8525 +t=250: Selected seed 195 with value = 0.8525 +Query 1/1: Action query time = 1.002 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8443 +t=266: Selected seed 195 with value = 0.8443 +Query 1/1: Action query time = 0.975 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8983 +t=282: Selected seed 195 with value = 0.8983 +Query 1/1: Action query time = 1.205 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=15--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=15--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 15 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 16... +Query 1/1: Action query time = 1.010 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4123 +t=10: Selected seed 195 with value = 0.4123 +Query 1/1: Action query time = 0.982 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5504 +t=26: Selected seed 195 with value = 0.5504 +Query 1/1: Action query time = 1.167 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6534 +t=42: Selected seed 195 with value = 0.6534 +Query 1/1: Action query time = 1.154 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8941 +t=58: Selected seed 195 with value = 0.8941 +Query 1/1: Action query time = 1.118 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8361 +t=74: Selected seed 195 with value = 0.8361 +Query 1/1: Action query time = 1.133 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9331 +t=90: Selected seed 195 with value = 0.9331 +Query 1/1: Action query time = 1.138 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.113 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.138 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.116 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.150 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.144 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.114 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.121 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.092 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9201 +t=234: Selected seed 195 with value = 0.9201 +Query 1/1: Action query time = 1.117 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8936 +t=250: Selected seed 195 with value = 0.8936 +Query 1/1: Action query time = 1.130 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8153 +t=266: Selected seed 195 with value = 0.8153 +Query 1/1: Action query time = 1.096 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9026 +t=282: Selected seed 195 with value = 0.9026 +Query 1/1: Action query time = 1.002 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=16--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=16--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 16 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 17... +Query 1/1: Action query time = 0.994 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5049 +t=10: Selected seed 195 with value = 0.5049 +Query 1/1: Action query time = 0.962 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5944 +t=26: Selected seed 195 with value = 0.5944 +Query 1/1: Action query time = 0.979 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6725 +t=42: Selected seed 195 with value = 0.6725 +Query 1/1: Action query time = 0.974 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7388 +t=58: Selected seed 195 with value = 0.7388 +Query 1/1: Action query time = 0.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8325 +t=74: Selected seed 195 with value = 0.8325 +Query 1/1: Action query time = 0.981 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9554 +t=90: Selected seed 195 with value = 0.9554 +Query 1/1: Action query time = 0.995 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=170: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 1.003 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9811 +t=186: Selected seed 195 with value = 0.9811 +Query 1/1: Action query time = 0.999 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9840 +t=202: Selected seed 195 with value = 0.9840 +Query 1/1: Action query time = 1.000 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.021 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.037 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.065 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=266: Selected seed 195 with value = 0.9864 +Query 1/1: Action query time = 1.029 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9948 +t=282: Selected seed 195 with value = 0.9948 +Query 1/1: Action query time = 0.972 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=298: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=17--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=17--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 17 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 18... +Query 1/1: Action query time = 0.983 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4599 +t=10: Selected seed 195 with value = 0.4599 +Query 1/1: Action query time = 0.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5454 +t=26: Selected seed 195 with value = 0.5454 +Query 1/1: Action query time = 0.978 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6514 +t=42: Selected seed 195 with value = 0.6514 +Query 1/1: Action query time = 0.979 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7889 +t=58: Selected seed 195 with value = 0.7889 +Query 1/1: Action query time = 0.964 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8449 +t=74: Selected seed 195 with value = 0.8449 +Query 1/1: Action query time = 0.961 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9391 +t=90: Selected seed 195 with value = 0.9391 +Query 1/1: Action query time = 0.965 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=138: Selected seed 195 with value = 0.9962 +Query 1/1: Action query time = 0.967 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9773 +t=154: Selected seed 195 with value = 0.9773 +Query 1/1: Action query time = 0.968 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9882 +t=170: Selected seed 195 with value = 0.9882 +Query 1/1: Action query time = 1.059 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=186: Selected seed 195 with value = 0.9980 +Query 1/1: Action query time = 1.115 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.092 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=218: Selected seed 195 with value = 0.9983 +Query 1/1: Action query time = 1.195 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9575 +t=234: Selected seed 195 with value = 0.9575 +Query 1/1: Action query time = 1.118 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8991 +t=250: Selected seed 195 with value = 0.8991 +Query 1/1: Action query time = 1.056 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8370 +t=266: Selected seed 195 with value = 0.8370 +Query 1/1: Action query time = 0.970 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9293 +t=282: Selected seed 195 with value = 0.9293 +Query 1/1: Action query time = 0.993 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=18--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=18--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 18 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 19... +Query 1/1: Action query time = 0.989 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4811 +t=10: Selected seed 195 with value = 0.4811 +Query 1/1: Action query time = 0.985 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5485 +t=26: Selected seed 195 with value = 0.5485 +Query 1/1: Action query time = 0.993 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6687 +t=42: Selected seed 195 with value = 0.6687 +Query 1/1: Action query time = 0.974 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8509 +t=58: Selected seed 195 with value = 0.8509 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8936 +t=74: Selected seed 195 with value = 0.8936 +Query 1/1: Action query time = 0.976 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9713 +t=90: Selected seed 195 with value = 0.9713 +Query 1/1: Action query time = 0.971 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.984 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.990 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9709 +t=218: Selected seed 195 with value = 0.9709 +Query 1/1: Action query time = 0.977 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8206 +t=234: Selected seed 195 with value = 0.8206 +Query 1/1: Action query time = 0.977 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8134 +t=250: Selected seed 195 with value = 0.8134 +Query 1/1: Action query time = 0.969 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9098 +t=266: Selected seed 195 with value = 0.9098 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=19--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=19--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 19 +# successes: 1 (5.3%) + +Task: put the bowl on the stove +Starting episode 20... +Query 1/1: Action query time = 0.977 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4979 +t=10: Selected seed 195 with value = 0.4979 +Query 1/1: Action query time = 0.981 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5985 +t=26: Selected seed 195 with value = 0.5985 +Query 1/1: Action query time = 0.979 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6734 +t=42: Selected seed 195 with value = 0.6734 +Query 1/1: Action query time = 0.969 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7679 +t=58: Selected seed 195 with value = 0.7679 +Query 1/1: Action query time = 0.966 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8426 +t=74: Selected seed 195 with value = 0.8426 +Query 1/1: Action query time = 0.978 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9590 +t=90: Selected seed 195 with value = 0.9590 +Query 1/1: Action query time = 1.107 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.084 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.034 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.037 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.013 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.013 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9894 +t=202: Selected seed 195 with value = 0.9894 +Query 1/1: Action query time = 1.006 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9805 +t=218: Selected seed 195 with value = 0.9805 +Query 1/1: Action query time = 1.015 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9873 +t=234: Selected seed 195 with value = 0.9873 +Query 1/1: Action query time = 1.020 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=250: Selected seed 195 with value = 0.9871 +Query 1/1: Action query time = 1.010 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=266: Selected seed 195 with value = 0.9864 +Query 1/1: Action query time = 0.999 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9836 +t=282: Selected seed 195 with value = 0.9836 +Query 1/1: Action query time = 0.998 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9827 +t=298: Selected seed 195 with value = 0.9827 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=20--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=20--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 20 +# successes: 1 (5.0%) + +Task: put the bowl on the stove +Starting episode 21... +Query 1/1: Action query time = 0.986 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4679 +t=10: Selected seed 195 with value = 0.4679 +Query 1/1: Action query time = 0.959 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5616 +t=26: Selected seed 195 with value = 0.5616 +Query 1/1: Action query time = 0.978 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6565 +t=42: Selected seed 195 with value = 0.6565 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7550 +t=58: Selected seed 195 with value = 0.7550 +Query 1/1: Action query time = 0.964 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8359 +t=74: Selected seed 195 with value = 0.8359 +Query 1/1: Action query time = 0.971 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9557 +t=90: Selected seed 195 with value = 0.9557 +Query 1/1: Action query time = 0.976 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.984 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.970 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9892 +t=170: Selected seed 195 with value = 0.9892 +Query 1/1: Action query time = 0.966 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9696 +t=186: Selected seed 195 with value = 0.9696 +Query 1/1: Action query time = 0.986 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9855 +t=202: Selected seed 195 with value = 0.9855 +Query 1/1: Action query time = 0.985 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9932 +t=250: Selected seed 195 with value = 0.9932 +Query 1/1: Action query time = 0.967 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9587 +t=266: Selected seed 195 with value = 0.9587 +Query 1/1: Action query time = 0.970 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9834 +t=282: Selected seed 195 with value = 0.9834 +Query 1/1: Action query time = 0.960 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=21--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=21--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 21 +# successes: 1 (4.8%) + +Task: put the bowl on the stove +Starting episode 22... +Query 1/1: Action query time = 1.085 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4421 +t=10: Selected seed 195 with value = 0.4421 +Query 1/1: Action query time = 0.975 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5575 +t=26: Selected seed 195 with value = 0.5575 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6637 +t=42: Selected seed 195 with value = 0.6637 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9336 +t=58: Selected seed 195 with value = 0.9336 +Query 1/1: Action query time = 0.990 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8750 +t=74: Selected seed 195 with value = 0.8750 +Query 1/1: Action query time = 0.975 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9831 +t=90: Selected seed 195 with value = 0.9831 +Query 1/1: Action query time = 0.996 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.989 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.959 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9891 +t=250: Selected seed 195 with value = 0.9891 +Query 1/1: Action query time = 0.971 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9421 +t=266: Selected seed 195 with value = 0.9421 +Query 1/1: Action query time = 0.962 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9637 +t=282: Selected seed 195 with value = 0.9637 +Query 1/1: Action query time = 0.962 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8572 +t=298: Selected seed 195 with value = 0.8572 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=22--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=22--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 22 +# successes: 1 (4.5%) + +Task: put the bowl on the stove +Starting episode 23... +Query 1/1: Action query time = 0.973 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4623 +t=10: Selected seed 195 with value = 0.4623 +Query 1/1: Action query time = 0.966 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5451 +t=26: Selected seed 195 with value = 0.5451 +Query 1/1: Action query time = 0.969 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6586 +t=42: Selected seed 195 with value = 0.6586 +Query 1/1: Action query time = 0.969 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7853 +t=58: Selected seed 195 with value = 0.7853 +Query 1/1: Action query time = 0.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8498 +t=74: Selected seed 195 with value = 0.8498 +Query 1/1: Action query time = 0.983 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9717 +t=90: Selected seed 195 with value = 0.9717 +Query 1/1: Action query time = 0.960 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.000 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.984 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9947 +t=218: Selected seed 195 with value = 0.9947 +Query 1/1: Action query time = 0.973 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=266: Selected seed 195 with value = 0.9999 +Query 1/1: Action query time = 0.969 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9868 +t=282: Selected seed 195 with value = 0.9868 +Query 1/1: Action query time = 1.002 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=298: Selected seed 195 with value = 0.9871 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=23--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=23--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 23 +# successes: 1 (4.3%) + +Task: put the bowl on the stove +Starting episode 24... +Query 1/1: Action query time = 1.033 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4789 +t=10: Selected seed 195 with value = 0.4789 +Query 1/1: Action query time = 1.134 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5484 +t=26: Selected seed 195 with value = 0.5484 +Query 1/1: Action query time = 1.172 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6697 +t=42: Selected seed 195 with value = 0.6697 +Query 1/1: Action query time = 1.165 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8542 +t=58: Selected seed 195 with value = 0.8542 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8952 +t=74: Selected seed 195 with value = 0.8952 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9839 +t=90: Selected seed 195 with value = 0.9839 +Query 1/1: Action query time = 0.968 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.961 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.983 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.990 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.003 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=186: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 1.030 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=202: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 0.974 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9918 +t=250: Selected seed 195 with value = 0.9918 +Query 1/1: Action query time = 0.973 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9878 +t=266: Selected seed 195 with value = 0.9878 +Query 1/1: Action query time = 0.964 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9428 +t=282: Selected seed 195 with value = 0.9428 +Query 1/1: Action query time = 0.970 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8406 +t=298: Selected seed 195 with value = 0.8406 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=24--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=24--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 24 +# successes: 1 (4.2%) + +Task: put the bowl on the stove +Starting episode 25... +Query 1/1: Action query time = 0.979 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4630 +t=10: Selected seed 195 with value = 0.4630 +Query 1/1: Action query time = 0.980 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5406 +t=26: Selected seed 195 with value = 0.5406 +Query 1/1: Action query time = 0.963 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6443 +t=42: Selected seed 195 with value = 0.6443 +Query 1/1: Action query time = 0.961 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8289 +t=58: Selected seed 195 with value = 0.8289 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8470 +t=74: Selected seed 195 with value = 0.8470 +Query 1/1: Action query time = 0.983 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9327 +t=90: Selected seed 195 with value = 0.9327 +Query 1/1: Action query time = 0.971 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=106: Selected seed 195 with value = 0.9999 +Query 1/1: Action query time = 0.970 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.960 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9937 +t=154: Selected seed 195 with value = 0.9937 +Query 1/1: Action query time = 0.970 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9820 +t=170: Selected seed 195 with value = 0.9820 +Query 1/1: Action query time = 0.964 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9698 +t=186: Selected seed 195 with value = 0.9698 +Query 1/1: Action query time = 0.978 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=202: Selected seed 195 with value = 0.9886 +Query 1/1: Action query time = 0.966 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9907 +t=218: Selected seed 195 with value = 0.9907 +Query 1/1: Action query time = 0.973 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=234: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 0.992 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=250: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 1.600 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9792 +t=266: Selected seed 195 with value = 0.9792 +Query 1/1: Action query time = 1.212 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8701 +t=282: Selected seed 195 with value = 0.8701 +Query 1/1: Action query time = 1.174 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8366 +t=298: Selected seed 195 with value = 0.8366 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=25--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=25--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 25 +# successes: 1 (4.0%) + +Task: put the bowl on the stove +Starting episode 26... +Query 1/1: Action query time = 1.518 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4826 +t=10: Selected seed 195 with value = 0.4826 +Query 1/1: Action query time = 1.198 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5475 +t=26: Selected seed 195 with value = 0.5475 +Query 1/1: Action query time = 1.181 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6575 +t=42: Selected seed 195 with value = 0.6575 +Query 1/1: Action query time = 1.182 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8499 +t=58: Selected seed 195 with value = 0.8499 +Query 1/1: Action query time = 1.237 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8670 +t=74: Selected seed 195 with value = 0.8670 +Query 1/1: Action query time = 1.258 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9686 +t=90: Selected seed 195 with value = 0.9686 +Query 1/1: Action query time = 0.962 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=138: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 0.976 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=154: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 0.964 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=170: Selected seed 195 with value = 0.9871 +Query 1/1: Action query time = 0.977 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9829 +t=186: Selected seed 195 with value = 0.9829 +Query 1/1: Action query time = 2.387 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9825 +t=202: Selected seed 195 with value = 0.9825 +Query 1/1: Action query time = 0.966 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9838 +t=218: Selected seed 195 with value = 0.9838 +Query 1/1: Action query time = 0.969 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9769 +t=234: Selected seed 195 with value = 0.9769 +Query 1/1: Action query time = 0.975 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9074 +t=250: Selected seed 195 with value = 0.9074 +Query 1/1: Action query time = 0.967 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9375 +t=266: Selected seed 195 with value = 0.9375 +Query 1/1: Action query time = 1.113 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.496 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=26--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=26--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 26 +# successes: 1 (3.8%) + +Task: put the bowl on the stove +Starting episode 27... +Query 1/1: Action query time = 0.976 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4850 +t=10: Selected seed 195 with value = 0.4850 +Query 1/1: Action query time = 0.975 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6230 +t=26: Selected seed 195 with value = 0.6230 +Query 1/1: Action query time = 0.968 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7132 +t=42: Selected seed 195 with value = 0.7132 +Query 1/1: Action query time = 0.984 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7135 +t=58: Selected seed 195 with value = 0.7135 +Query 1/1: Action query time = 1.135 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8308 +t=74: Selected seed 195 with value = 0.8308 +Query 1/1: Action query time = 1.088 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9562 +t=90: Selected seed 195 with value = 0.9562 +Query 1/1: Action query time = 1.074 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.084 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.154 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.134 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9925 +t=154: Selected seed 195 with value = 0.9925 +Query 1/1: Action query time = 1.115 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9780 +t=170: Selected seed 195 with value = 0.9780 +Query 1/1: Action query time = 1.099 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9931 +t=234: Selected seed 195 with value = 0.9931 +Query 1/1: Action query time = 0.977 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9929 +t=250: Selected seed 195 with value = 0.9929 +Query 1/1: Action query time = 0.995 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=266: Selected seed 195 with value = 0.9989 +Query 1/1: Action query time = 0.962 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=282: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 0.963 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9879 +t=298: Selected seed 195 with value = 0.9879 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=27--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=27--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 27 +# successes: 1 (3.7%) + +Task: put the bowl on the stove +Starting episode 28... +Query 1/1: Action query time = 0.976 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4778 +t=10: Selected seed 195 with value = 0.4778 +Query 1/1: Action query time = 0.969 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5768 +t=26: Selected seed 195 with value = 0.5768 +Query 1/1: Action query time = 0.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6643 +t=42: Selected seed 195 with value = 0.6643 +Query 1/1: Action query time = 0.961 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8419 +t=58: Selected seed 195 with value = 0.8419 +Query 1/1: Action query time = 0.967 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8350 +t=74: Selected seed 195 with value = 0.8350 +Query 1/1: Action query time = 1.004 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9256 +t=90: Selected seed 195 with value = 0.9256 +Query 1/1: Action query time = 0.981 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=138: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 0.974 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=154: Selected seed 195 with value = 0.9809 +Query 1/1: Action query time = 0.973 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9736 +t=170: Selected seed 195 with value = 0.9736 +Query 1/1: Action query time = 0.976 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9802 +t=186: Selected seed 195 with value = 0.9802 +Query 1/1: Action query time = 0.972 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9857 +t=202: Selected seed 195 with value = 0.9857 +Query 1/1: Action query time = 0.980 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9833 +t=218: Selected seed 195 with value = 0.9833 +Query 1/1: Action query time = 0.965 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9828 +t=234: Selected seed 195 with value = 0.9828 +Query 1/1: Action query time = 0.952 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9875 +t=250: Selected seed 195 with value = 0.9875 +Query 1/1: Action query time = 0.973 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9897 +t=266: Selected seed 195 with value = 0.9897 +Query 1/1: Action query time = 0.964 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9885 +t=282: Selected seed 195 with value = 0.9885 +Query 1/1: Action query time = 0.960 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9932 +t=298: Selected seed 195 with value = 0.9932 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=28--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=28--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 28 +# successes: 1 (3.6%) + +Task: put the bowl on the stove +Starting episode 29... +Query 1/1: Action query time = 0.992 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4606 +t=10: Selected seed 195 with value = 0.4606 +Query 1/1: Action query time = 0.955 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5406 +t=26: Selected seed 195 with value = 0.5406 +Query 1/1: Action query time = 0.963 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6550 +t=42: Selected seed 195 with value = 0.6550 +Query 1/1: Action query time = 0.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8187 +t=58: Selected seed 195 with value = 0.8187 +Query 1/1: Action query time = 0.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8408 +t=74: Selected seed 195 with value = 0.8408 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9558 +t=90: Selected seed 195 with value = 0.9558 +Query 1/1: Action query time = 0.956 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.986 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.001 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=154: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 0.978 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9942 +t=170: Selected seed 195 with value = 0.9942 +Query 1/1: Action query time = 0.973 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=202: Selected seed 195 with value = 0.9968 +Query 1/1: Action query time = 0.970 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9758 +t=218: Selected seed 195 with value = 0.9758 +Query 1/1: Action query time = 0.963 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9585 +t=234: Selected seed 195 with value = 0.9585 +Query 1/1: Action query time = 0.958 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8435 +t=250: Selected seed 195 with value = 0.8435 +Query 1/1: Action query time = 0.976 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9015 +t=266: Selected seed 195 with value = 0.9015 +Query 1/1: Action query time = 0.972 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=29--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=29--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 29 +# successes: 1 (3.4%) + +Task: put the bowl on the stove +Starting episode 30... +Query 1/1: Action query time = 0.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5272 +t=10: Selected seed 195 with value = 0.5272 +Query 1/1: Action query time = 0.960 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6022 +t=26: Selected seed 195 with value = 0.6022 +Query 1/1: Action query time = 0.970 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6694 +t=42: Selected seed 195 with value = 0.6694 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7492 +t=58: Selected seed 195 with value = 0.7492 +Query 1/1: Action query time = 0.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8273 +t=74: Selected seed 195 with value = 0.8273 +Query 1/1: Action query time = 0.969 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9583 +t=90: Selected seed 195 with value = 0.9583 +Query 1/1: Action query time = 0.964 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=154: Selected seed 195 with value = 0.9986 +Query 1/1: Action query time = 0.970 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=170: Selected seed 195 with value = 0.9946 +Query 1/1: Action query time = 0.978 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9824 +t=186: Selected seed 195 with value = 0.9824 +Query 1/1: Action query time = 0.971 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9761 +t=202: Selected seed 195 with value = 0.9761 +Query 1/1: Action query time = 0.973 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9833 +t=218: Selected seed 195 with value = 0.9833 +Query 1/1: Action query time = 0.976 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9943 +t=234: Selected seed 195 with value = 0.9943 +Query 1/1: Action query time = 0.980 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9893 +t=282: Selected seed 195 with value = 0.9893 +Query 1/1: Action query time = 0.974 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8970 +t=298: Selected seed 195 with value = 0.8970 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=30--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=30--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 30 +# successes: 1 (3.3%) + +Task: put the bowl on the stove +Starting episode 31... +Query 1/1: Action query time = 0.976 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4847 +t=10: Selected seed 195 with value = 0.4847 +Query 1/1: Action query time = 0.966 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5333 +t=26: Selected seed 195 with value = 0.5333 +Query 1/1: Action query time = 0.963 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6436 +t=42: Selected seed 195 with value = 0.6436 +Query 1/1: Action query time = 0.954 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8294 +t=58: Selected seed 195 with value = 0.8294 +Query 1/1: Action query time = 0.969 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8143 +t=74: Selected seed 195 with value = 0.8143 +Query 1/1: Action query time = 0.964 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9132 +t=90: Selected seed 195 with value = 0.9132 +Query 1/1: Action query time = 0.962 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9684 +t=106: Selected seed 195 with value = 0.9684 +Query 1/1: Action query time = 0.965 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.986 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.983 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.950 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.960 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=298: Selected seed 195 with value = 0.9977 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=31--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=31--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 31 +# successes: 1 (3.2%) + +Task: put the bowl on the stove +Starting episode 32... +Query 1/1: Action query time = 0.983 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4838 +t=10: Selected seed 195 with value = 0.4838 +Query 1/1: Action query time = 0.976 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5584 +t=26: Selected seed 195 with value = 0.5584 +Query 1/1: Action query time = 0.967 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6623 +t=42: Selected seed 195 with value = 0.6623 +Query 1/1: Action query time = 0.976 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8600 +t=58: Selected seed 195 with value = 0.8600 +Query 1/1: Action query time = 0.987 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8842 +t=74: Selected seed 195 with value = 0.8842 +Query 1/1: Action query time = 1.009 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9866 +t=90: Selected seed 195 with value = 0.9866 +Query 1/1: Action query time = 0.992 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.983 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.992 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=202: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 0.972 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.989 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9959 +t=266: Selected seed 195 with value = 0.9959 +Query 1/1: Action query time = 0.974 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=32--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=32--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 32 +# successes: 1 (3.1%) + +Task: put the bowl on the stove +Starting episode 33... +Query 1/1: Action query time = 0.972 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5043 +t=10: Selected seed 195 with value = 0.5043 +Query 1/1: Action query time = 0.970 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5479 +t=26: Selected seed 195 with value = 0.5479 +Query 1/1: Action query time = 0.969 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6489 +t=42: Selected seed 195 with value = 0.6489 +Query 1/1: Action query time = 0.989 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8714 +t=58: Selected seed 195 with value = 0.8714 +Query 1/1: Action query time = 0.973 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8199 +t=74: Selected seed 195 with value = 0.8199 +Query 1/1: Action query time = 0.958 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9121 +t=90: Selected seed 195 with value = 0.9121 +Query 1/1: Action query time = 0.978 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9939 +t=106: Selected seed 195 with value = 0.9939 +Query 1/1: Action query time = 0.983 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.994 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.988 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.970 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=298: Selected seed 195 with value = 0.9946 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=33--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=33--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 33 +# successes: 1 (3.0%) + +Task: put the bowl on the stove +Starting episode 34... +Query 1/1: Action query time = 0.992 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4837 +t=10: Selected seed 195 with value = 0.4837 +Query 1/1: Action query time = 0.986 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5682 +t=26: Selected seed 195 with value = 0.5682 +Query 1/1: Action query time = 0.968 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6853 +t=42: Selected seed 195 with value = 0.6853 +Query 1/1: Action query time = 0.979 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8037 +t=58: Selected seed 195 with value = 0.8037 +Query 1/1: Action query time = 0.967 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8491 +t=74: Selected seed 195 with value = 0.8491 +Query 1/1: Action query time = 0.979 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9361 +t=90: Selected seed 195 with value = 0.9361 +Query 1/1: Action query time = 0.968 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=186: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 0.963 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9952 +t=202: Selected seed 195 with value = 0.9952 +Query 1/1: Action query time = 0.972 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=218: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 0.965 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.961 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=250: Selected seed 195 with value = 0.9909 +Query 1/1: Action query time = 0.981 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9420 +t=266: Selected seed 195 with value = 0.9420 +Query 1/1: Action query time = 0.967 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9040 +t=282: Selected seed 195 with value = 0.9040 +Query 1/1: Action query time = 0.987 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8390 +t=298: Selected seed 195 with value = 0.8390 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=34--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=34--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 34 +# successes: 1 (2.9%) + +Task: put the bowl on the stove +Starting episode 35... +Query 1/1: Action query time = 0.969 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4651 +t=10: Selected seed 195 with value = 0.4651 +Query 1/1: Action query time = 0.982 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5365 +t=26: Selected seed 195 with value = 0.5365 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6469 +t=42: Selected seed 195 with value = 0.6469 +Query 1/1: Action query time = 0.979 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8328 +t=58: Selected seed 195 with value = 0.8328 +Query 1/1: Action query time = 0.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8156 +t=74: Selected seed 195 with value = 0.8156 +Query 1/1: Action query time = 0.968 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9204 +t=90: Selected seed 195 with value = 0.9204 +Query 1/1: Action query time = 0.977 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9906 +t=106: Selected seed 195 with value = 0.9906 +Query 1/1: Action query time = 0.992 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=35--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=35--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 35 +# successes: 1 (2.9%) + +Task: put the bowl on the stove +Starting episode 36... +Query 1/1: Action query time = 0.982 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4848 +t=10: Selected seed 195 with value = 0.4848 +Query 1/1: Action query time = 0.970 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6214 +t=26: Selected seed 195 with value = 0.6214 +Query 1/1: Action query time = 0.971 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6520 +t=42: Selected seed 195 with value = 0.6520 +Query 1/1: Action query time = 0.984 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8085 +t=58: Selected seed 195 with value = 0.8085 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6868 +t=74: Selected seed 195 with value = 0.6868 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7750 +t=90: Selected seed 195 with value = 0.7750 +Query 1/1: Action query time = 0.969 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8469 +t=106: Selected seed 195 with value = 0.8469 +Query 1/1: Action query time = 0.974 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8637 +t=122: Selected seed 195 with value = 0.8637 +Query 1/1: Action query time = 0.981 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9646 +t=138: Selected seed 195 with value = 0.9646 +Query 1/1: Action query time = 0.981 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9747 +t=154: Selected seed 195 with value = 0.9747 +Query 1/1: Action query time = 0.970 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9868 +t=170: Selected seed 195 with value = 0.9868 +Query 1/1: Action query time = 0.973 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9795 +t=186: Selected seed 195 with value = 0.9795 +Query 1/1: Action query time = 0.970 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9740 +t=202: Selected seed 195 with value = 0.9740 +Query 1/1: Action query time = 0.964 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9638 +t=218: Selected seed 195 with value = 0.9638 +Query 1/1: Action query time = 0.989 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9416 +t=234: Selected seed 195 with value = 0.9416 +Query 1/1: Action query time = 0.965 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8564 +t=250: Selected seed 195 with value = 0.8564 +Query 1/1: Action query time = 0.975 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7001 +t=266: Selected seed 195 with value = 0.7001 +Query 1/1: Action query time = 1.019 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8003 +t=282: Selected seed 195 with value = 0.8003 +Query 1/1: Action query time = 0.970 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8530 +t=298: Selected seed 195 with value = 0.8530 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=36--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=36--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 36 +# successes: 1 (2.8%) + +Task: put the bowl on the stove +Starting episode 37... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4866 +t=10: Selected seed 195 with value = 0.4866 +Query 1/1: Action query time = 0.956 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5843 +t=26: Selected seed 195 with value = 0.5843 +Query 1/1: Action query time = 0.982 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6584 +t=42: Selected seed 195 with value = 0.6584 +Query 1/1: Action query time = 0.985 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9117 +t=58: Selected seed 195 with value = 0.9117 +Query 1/1: Action query time = 0.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8554 +t=74: Selected seed 195 with value = 0.8554 +Query 1/1: Action query time = 0.966 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9301 +t=90: Selected seed 195 with value = 0.9301 +Query 1/1: Action query time = 0.970 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9856 +t=106: Selected seed 195 with value = 0.9856 +Query 1/1: Action query time = 0.960 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.970 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9032 +t=266: Selected seed 195 with value = 0.9032 +Query 1/1: Action query time = 0.962 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8997 +t=282: Selected seed 195 with value = 0.8997 +Query 1/1: Action query time = 0.976 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8317 +t=298: Selected seed 195 with value = 0.8317 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=37--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=37--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 37 +# successes: 1 (2.7%) + +Task: put the bowl on the stove +Starting episode 38... +Query 1/1: Action query time = 0.968 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5009 +t=10: Selected seed 195 with value = 0.5009 +Query 1/1: Action query time = 0.962 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5866 +t=26: Selected seed 195 with value = 0.5866 +Query 1/1: Action query time = 0.966 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6396 +t=42: Selected seed 195 with value = 0.6396 +Query 1/1: Action query time = 0.960 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7340 +t=58: Selected seed 195 with value = 0.7340 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8182 +t=74: Selected seed 195 with value = 0.8182 +Query 1/1: Action query time = 0.971 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9601 +t=90: Selected seed 195 with value = 0.9601 +Query 1/1: Action query time = 0.980 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.959 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=218: Selected seed 195 with value = 0.9965 +Query 1/1: Action query time = 0.969 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=234: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 0.969 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=250: Selected seed 195 with value = 0.9968 +Query 1/1: Action query time = 0.969 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9804 +t=266: Selected seed 195 with value = 0.9804 +Query 1/1: Action query time = 0.970 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9657 +t=282: Selected seed 195 with value = 0.9657 +Query 1/1: Action query time = 0.965 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9724 +t=298: Selected seed 195 with value = 0.9724 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=38--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=38--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 38 +# successes: 1 (2.6%) + +Task: put the bowl on the stove +Starting episode 39... +Query 1/1: Action query time = 0.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4834 +t=10: Selected seed 195 with value = 0.4834 +Query 1/1: Action query time = 0.960 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5373 +t=26: Selected seed 195 with value = 0.5373 +Query 1/1: Action query time = 0.962 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6513 +t=42: Selected seed 195 with value = 0.6513 +Query 1/1: Action query time = 0.983 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8421 +t=58: Selected seed 195 with value = 0.8421 +Query 1/1: Action query time = 0.970 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8518 +t=74: Selected seed 195 with value = 0.8518 +Query 1/1: Action query time = 0.969 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9484 +t=90: Selected seed 195 with value = 0.9484 +Query 1/1: Action query time = 0.963 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.960 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.959 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=250: Selected seed 195 with value = 0.9999 +Query 1/1: Action query time = 0.988 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9312 +t=266: Selected seed 195 with value = 0.9312 +Query 1/1: Action query time = 0.963 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8858 +t=282: Selected seed 195 with value = 0.8858 +Query 1/1: Action query time = 0.963 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8430 +t=298: Selected seed 195 with value = 0.8430 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=39--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=39--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 39 +# successes: 1 (2.6%) + +Task: put the bowl on the stove +Starting episode 40... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5106 +t=10: Selected seed 195 with value = 0.5106 +Query 1/1: Action query time = 0.972 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5500 +t=26: Selected seed 195 with value = 0.5500 +Query 1/1: Action query time = 0.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6841 +t=42: Selected seed 195 with value = 0.6841 +Query 1/1: Action query time = 0.976 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8116 +t=58: Selected seed 195 with value = 0.8116 +Query 1/1: Action query time = 0.969 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8886 +t=74: Selected seed 195 with value = 0.8886 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=90: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 0.964 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9970 +t=202: Selected seed 195 with value = 0.9970 +Query 1/1: Action query time = 0.977 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9940 +t=218: Selected seed 195 with value = 0.9940 +Query 1/1: Action query time = 0.967 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9718 +t=234: Selected seed 195 with value = 0.9718 +Query 1/1: Action query time = 0.969 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9767 +t=250: Selected seed 195 with value = 0.9767 +Query 1/1: Action query time = 0.957 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9640 +t=266: Selected seed 195 with value = 0.9640 +Query 1/1: Action query time = 0.990 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9600 +t=282: Selected seed 195 with value = 0.9600 +Query 1/1: Action query time = 0.981 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9764 +t=298: Selected seed 195 with value = 0.9764 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--episode=40--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/ft6_iter200_t1/2026_07_31-14_22_22--with_future_img--episode=40--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 40 +# successes: 1 (2.5%) + +Task: put the bowl on the stove +Starting episode 41... +Query 1/1: Action query time = 0.995 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4674 +t=10: Selected seed 195 with value = 0.4674 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_07_31-14_46_45--CLi800_t6.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_07_31-14_46_45--CLi800_t6.txt new file mode 100644 index 0000000000000000000000000000000000000000..5cbff20b3b402133e25f6178feb26760e3e0108b --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_07_31-14_46_45--CLi800_t6.txt @@ -0,0 +1,2012 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='CLi800_t6', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=True, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 1.512 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4468 +t=10: Selected seed 195 with value = 0.4468 +Query 1/1: Action query time = 0.975 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5382 +t=26: Selected seed 195 with value = 0.5382 +Query 1/1: Action query time = 0.962 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6396 +t=42: Selected seed 195 with value = 0.6396 +Query 1/1: Action query time = 0.970 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4866 +t=58: Selected seed 195 with value = 0.4866 +Query 1/1: Action query time = 0.973 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6016 +t=74: Selected seed 195 with value = 0.6016 +Query 1/1: Action query time = 0.962 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6904 +t=90: Selected seed 195 with value = 0.6904 +Query 1/1: Action query time = 1.136 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8101 +t=106: Selected seed 195 with value = 0.8101 +Query 1/1: Action query time = 0.963 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9004 +t=122: Selected seed 195 with value = 0.9004 +Query 1/1: Action query time = 0.959 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=138: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 0.970 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.953 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=170: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 0.974 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=186: Selected seed 195 with value = 0.9968 +Query 1/1: Action query time = 0.977 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9967 +t=234: Selected seed 195 with value = 0.9967 +Query 1/1: Action query time = 0.969 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.085 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=266: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 1.056 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9875 +t=282: Selected seed 195 with value = 0.9875 +Query 1/1: Action query time = 0.968 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 0.986 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4393 +t=10: Selected seed 195 with value = 0.4393 +Query 1/1: Action query time = 0.978 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4662 +t=26: Selected seed 195 with value = 0.4662 +Query 1/1: Action query time = 0.984 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5736 +t=42: Selected seed 195 with value = 0.5736 +Query 1/1: Action query time = 1.045 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6849 +t=58: Selected seed 195 with value = 0.6849 +Query 1/1: Action query time = 1.129 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8401 +t=74: Selected seed 195 with value = 0.8401 +Query 1/1: Action query time = 1.095 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9636 +t=90: Selected seed 195 with value = 0.9636 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4538 +t=10: Selected seed 195 with value = 0.4538 +Query 1/1: Action query time = 0.979 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3904 +t=26: Selected seed 195 with value = 0.3904 +Query 1/1: Action query time = 1.078 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6191 +t=42: Selected seed 195 with value = 0.6191 +Query 1/1: Action query time = 1.176 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7252 +t=58: Selected seed 195 with value = 0.7252 +Query 1/1: Action query time = 1.172 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8555 +t=74: Selected seed 195 with value = 0.8555 +Query 1/1: Action query time = 1.109 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: put the cream cheese in the bowl +Starting episode 4... +Query 1/1: Action query time = 0.986 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4075 +t=10: Selected seed 195 with value = 0.4075 +Query 1/1: Action query time = 0.974 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5459 +t=26: Selected seed 195 with value = 0.5459 +Query 1/1: Action query time = 0.961 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6128 +t=42: Selected seed 195 with value = 0.6128 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7160 +t=58: Selected seed 195 with value = 0.7160 +Query 1/1: Action query time = 1.067 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8293 +t=74: Selected seed 195 with value = 0.8293 +Query 1/1: Action query time = 0.998 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9188 +t=90: Selected seed 195 with value = 0.9188 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=4--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=4--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 4 +# successes: 3 (75.0%) + +Task: put the cream cheese in the bowl +Starting episode 5... +Query 1/1: Action query time = 1.050 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4197 +t=10: Selected seed 195 with value = 0.4197 +Query 1/1: Action query time = 1.057 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4797 +t=26: Selected seed 195 with value = 0.4797 +Query 1/1: Action query time = 0.979 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5955 +t=42: Selected seed 195 with value = 0.5955 +Query 1/1: Action query time = 0.993 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6903 +t=58: Selected seed 195 with value = 0.6903 +Query 1/1: Action query time = 0.997 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7715 +t=74: Selected seed 195 with value = 0.7715 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8969 +t=90: Selected seed 195 with value = 0.8969 +Query 1/1: Action query time = 0.970 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9605 +t=106: Selected seed 195 with value = 0.9605 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=5--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=5--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 5 +# successes: 4 (80.0%) + +Task: put the cream cheese in the bowl +Starting episode 6... +Query 1/1: Action query time = 1.129 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4413 +t=10: Selected seed 195 with value = 0.4413 +Query 1/1: Action query time = 1.087 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5030 +t=26: Selected seed 195 with value = 0.5030 +Query 1/1: Action query time = 0.984 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6293 +t=42: Selected seed 195 with value = 0.6293 +Query 1/1: Action query time = 0.979 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7160 +t=58: Selected seed 195 with value = 0.7160 +Query 1/1: Action query time = 0.986 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8680 +t=74: Selected seed 195 with value = 0.8680 +Query 1/1: Action query time = 0.961 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9869 +t=90: Selected seed 195 with value = 0.9869 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=6--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=6--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 6 +# successes: 5 (83.3%) + +Task: put the cream cheese in the bowl +Starting episode 7... +Query 1/1: Action query time = 1.001 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4769 +t=10: Selected seed 195 with value = 0.4769 +Query 1/1: Action query time = 0.987 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5420 +t=26: Selected seed 195 with value = 0.5420 +Query 1/1: Action query time = 0.986 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6309 +t=42: Selected seed 195 with value = 0.6309 +Query 1/1: Action query time = 1.011 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7299 +t=58: Selected seed 195 with value = 0.7299 +Query 1/1: Action query time = 0.981 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8263 +t=74: Selected seed 195 with value = 0.8263 +Query 1/1: Action query time = 0.975 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9916 +t=90: Selected seed 195 with value = 0.9916 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=7--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=7--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 7 +# successes: 6 (85.7%) + +Task: put the cream cheese in the bowl +Starting episode 8... +Query 1/1: Action query time = 0.973 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4158 +t=10: Selected seed 195 with value = 0.4158 +Query 1/1: Action query time = 0.953 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5389 +t=26: Selected seed 195 with value = 0.5389 +Query 1/1: Action query time = 0.949 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6300 +t=42: Selected seed 195 with value = 0.6300 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7250 +t=58: Selected seed 195 with value = 0.7250 +Query 1/1: Action query time = 0.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8764 +t=74: Selected seed 195 with value = 0.8764 +Query 1/1: Action query time = 0.970 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=8--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=8--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 8 +# successes: 7 (87.5%) + +Task: put the cream cheese in the bowl +Starting episode 9... +Query 1/1: Action query time = 0.991 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4389 +t=10: Selected seed 195 with value = 0.4389 +Query 1/1: Action query time = 0.981 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5314 +t=26: Selected seed 195 with value = 0.5314 +Query 1/1: Action query time = 0.984 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6191 +t=42: Selected seed 195 with value = 0.6191 +Query 1/1: Action query time = 0.972 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7263 +t=58: Selected seed 195 with value = 0.7263 +Query 1/1: Action query time = 0.981 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8634 +t=74: Selected seed 195 with value = 0.8634 +Query 1/1: Action query time = 0.989 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=9--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=9--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 9 +# successes: 8 (88.9%) + +Task: put the cream cheese in the bowl +Starting episode 10... +Query 1/1: Action query time = 1.114 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4272 +t=10: Selected seed 195 with value = 0.4272 +Query 1/1: Action query time = 1.141 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5286 +t=26: Selected seed 195 with value = 0.5286 +Query 1/1: Action query time = 1.161 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5922 +t=42: Selected seed 195 with value = 0.5922 +Query 1/1: Action query time = 1.123 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7107 +t=58: Selected seed 195 with value = 0.7107 +Query 1/1: Action query time = 1.089 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8505 +t=74: Selected seed 195 with value = 0.8505 +Query 1/1: Action query time = 1.050 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=90: Selected seed 195 with value = 0.9978 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=10--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=10--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 10 +# successes: 9 (90.0%) + +Task: put the cream cheese in the bowl +Starting episode 11... +Query 1/1: Action query time = 0.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4520 +t=10: Selected seed 195 with value = 0.4520 +Query 1/1: Action query time = 1.172 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5037 +t=26: Selected seed 195 with value = 0.5037 +Query 1/1: Action query time = 1.174 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6019 +t=42: Selected seed 195 with value = 0.6019 +Query 1/1: Action query time = 1.058 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5013 +t=58: Selected seed 195 with value = 0.5013 +Query 1/1: Action query time = 0.961 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5856 +t=74: Selected seed 195 with value = 0.5856 +Query 1/1: Action query time = 0.975 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6856 +t=90: Selected seed 195 with value = 0.6856 +Query 1/1: Action query time = 0.986 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7852 +t=106: Selected seed 195 with value = 0.7852 +Query 1/1: Action query time = 0.978 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9120 +t=122: Selected seed 195 with value = 0.9120 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=11--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=11--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 11 +# successes: 10 (90.9%) + +Task: put the cream cheese in the bowl +Starting episode 12... +Query 1/1: Action query time = 1.040 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4721 +t=10: Selected seed 195 with value = 0.4721 +Query 1/1: Action query time = 1.034 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5036 +t=26: Selected seed 195 with value = 0.5036 +Query 1/1: Action query time = 1.051 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6446 +t=42: Selected seed 195 with value = 0.6446 +Query 1/1: Action query time = 1.120 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7397 +t=58: Selected seed 195 with value = 0.7397 +Query 1/1: Action query time = 1.147 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8776 +t=74: Selected seed 195 with value = 0.8776 +Query 1/1: Action query time = 1.160 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=90: Selected seed 195 with value = 0.9979 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=12--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=12--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 12 +# successes: 11 (91.7%) + +Task: put the cream cheese in the bowl +Starting episode 13... +Query 1/1: Action query time = 0.990 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4564 +t=10: Selected seed 195 with value = 0.4564 +Query 1/1: Action query time = 0.986 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4145 +t=26: Selected seed 195 with value = 0.4145 +Query 1/1: Action query time = 0.984 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6285 +t=42: Selected seed 195 with value = 0.6285 +Query 1/1: Action query time = 1.083 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6642 +t=58: Selected seed 195 with value = 0.6642 +Query 1/1: Action query time = 1.137 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8066 +t=74: Selected seed 195 with value = 0.8066 +Query 1/1: Action query time = 1.159 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9186 +t=90: Selected seed 195 with value = 0.9186 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=13--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=13--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 13 +# successes: 12 (92.3%) + +Task: put the cream cheese in the bowl +Starting episode 14... +Query 1/1: Action query time = 1.151 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4375 +t=10: Selected seed 195 with value = 0.4375 +Query 1/1: Action query time = 1.034 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5557 +t=26: Selected seed 195 with value = 0.5557 +Query 1/1: Action query time = 0.998 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6301 +t=42: Selected seed 195 with value = 0.6301 +Query 1/1: Action query time = 0.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7350 +t=58: Selected seed 195 with value = 0.7350 +Query 1/1: Action query time = 0.983 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9009 +t=74: Selected seed 195 with value = 0.9009 +Query 1/1: Action query time = 0.960 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=14--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=14--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 14 +# successes: 13 (92.9%) + +Task: put the cream cheese in the bowl +Starting episode 15... +Query 1/1: Action query time = 0.980 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4291 +t=10: Selected seed 195 with value = 0.4291 +Query 1/1: Action query time = 1.014 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5397 +t=26: Selected seed 195 with value = 0.5397 +Query 1/1: Action query time = 0.977 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6270 +t=42: Selected seed 195 with value = 0.6270 +Query 1/1: Action query time = 0.974 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6470 +t=58: Selected seed 195 with value = 0.6470 +Query 1/1: Action query time = 0.978 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5872 +t=74: Selected seed 195 with value = 0.5872 +Query 1/1: Action query time = 0.976 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6787 +t=90: Selected seed 195 with value = 0.6787 +Query 1/1: Action query time = 0.974 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7012 +t=106: Selected seed 195 with value = 0.7012 +Query 1/1: Action query time = 0.977 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8611 +t=122: Selected seed 195 with value = 0.8611 +Query 1/1: Action query time = 0.977 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9309 +t=138: Selected seed 195 with value = 0.9309 +Query 1/1: Action query time = 0.995 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.988 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=218: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 0.988 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=234: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 0.967 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=250: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 0.969 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=266: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 0.978 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=282: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 0.968 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=298: Selected seed 195 with value = 0.9961 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=15--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=15--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 15 +# successes: 13 (86.7%) + +Task: put the cream cheese in the bowl +Starting episode 16... +Query 1/1: Action query time = 0.977 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4385 +t=10: Selected seed 195 with value = 0.4385 +Query 1/1: Action query time = 0.970 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5315 +t=26: Selected seed 195 with value = 0.5315 +Query 1/1: Action query time = 0.976 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6286 +t=42: Selected seed 195 with value = 0.6286 +Query 1/1: Action query time = 0.963 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7263 +t=58: Selected seed 195 with value = 0.7263 +Query 1/1: Action query time = 0.987 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8531 +t=74: Selected seed 195 with value = 0.8531 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9975 +t=90: Selected seed 195 with value = 0.9975 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=16--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=16--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 16 +# successes: 14 (87.5%) + +Task: put the cream cheese in the bowl +Starting episode 17... +Query 1/1: Action query time = 1.006 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4431 +t=10: Selected seed 195 with value = 0.4431 +Query 1/1: Action query time = 0.970 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5258 +t=26: Selected seed 195 with value = 0.5258 +Query 1/1: Action query time = 0.996 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6472 +t=42: Selected seed 195 with value = 0.6472 +Query 1/1: Action query time = 0.974 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7314 +t=58: Selected seed 195 with value = 0.7314 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8617 +t=74: Selected seed 195 with value = 0.8617 +Query 1/1: Action query time = 0.967 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9788 +t=90: Selected seed 195 with value = 0.9788 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=17--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=17--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 17 +# successes: 15 (88.2%) + +Task: put the cream cheese in the bowl +Starting episode 18... +Query 1/1: Action query time = 0.992 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4354 +t=10: Selected seed 195 with value = 0.4354 +Query 1/1: Action query time = 0.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5341 +t=26: Selected seed 195 with value = 0.5341 +Query 1/1: Action query time = 0.984 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6280 +t=42: Selected seed 195 with value = 0.6280 +Query 1/1: Action query time = 0.981 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7141 +t=58: Selected seed 195 with value = 0.7141 +Query 1/1: Action query time = 0.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8465 +t=74: Selected seed 195 with value = 0.8465 +Query 1/1: Action query time = 0.987 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9638 +t=90: Selected seed 195 with value = 0.9638 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=18--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=18--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 18 +# successes: 16 (88.9%) + +Task: put the cream cheese in the bowl +Starting episode 19... +Query 1/1: Action query time = 0.999 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4322 +t=10: Selected seed 195 with value = 0.4322 +Query 1/1: Action query time = 0.990 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3639 +t=26: Selected seed 195 with value = 0.3639 +Query 1/1: Action query time = 0.970 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6051 +t=42: Selected seed 195 with value = 0.6051 +Query 1/1: Action query time = 0.979 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5233 +t=58: Selected seed 195 with value = 0.5233 +Query 1/1: Action query time = 0.982 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5968 +t=74: Selected seed 195 with value = 0.5968 +Query 1/1: Action query time = 0.967 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6701 +t=90: Selected seed 195 with value = 0.6701 +Query 1/1: Action query time = 0.988 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7860 +t=106: Selected seed 195 with value = 0.7860 +Query 1/1: Action query time = 0.974 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9280 +t=122: Selected seed 195 with value = 0.9280 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=19--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=19--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 19 +# successes: 17 (89.5%) + +Task: put the cream cheese in the bowl +Starting episode 20... +Query 1/1: Action query time = 0.977 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4443 +t=10: Selected seed 195 with value = 0.4443 +Query 1/1: Action query time = 0.983 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5412 +t=26: Selected seed 195 with value = 0.5412 +Query 1/1: Action query time = 0.988 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6312 +t=42: Selected seed 195 with value = 0.6312 +Query 1/1: Action query time = 0.981 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7189 +t=58: Selected seed 195 with value = 0.7189 +Query 1/1: Action query time = 0.987 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8435 +t=74: Selected seed 195 with value = 0.8435 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9609 +t=90: Selected seed 195 with value = 0.9609 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=20--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=20--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 20 +# successes: 18 (90.0%) + +Task: put the cream cheese in the bowl +Starting episode 21... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4342 +t=10: Selected seed 195 with value = 0.4342 +Query 1/1: Action query time = 0.989 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5414 +t=26: Selected seed 195 with value = 0.5414 +Query 1/1: Action query time = 0.981 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6043 +t=42: Selected seed 195 with value = 0.6043 +Query 1/1: Action query time = 1.074 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5191 +t=58: Selected seed 195 with value = 0.5191 +Query 1/1: Action query time = 1.087 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6100 +t=74: Selected seed 195 with value = 0.6100 +Query 1/1: Action query time = 1.173 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6991 +t=90: Selected seed 195 with value = 0.6991 +Query 1/1: Action query time = 1.189 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8105 +t=106: Selected seed 195 with value = 0.8105 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=21--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=21--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 21 +# successes: 19 (90.5%) + +Task: put the cream cheese in the bowl +Starting episode 22... +Query 1/1: Action query time = 0.992 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4426 +t=10: Selected seed 195 with value = 0.4426 +Query 1/1: Action query time = 0.988 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5197 +t=26: Selected seed 195 with value = 0.5197 +Query 1/1: Action query time = 0.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6282 +t=42: Selected seed 195 with value = 0.6282 +Query 1/1: Action query time = 0.968 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7400 +t=58: Selected seed 195 with value = 0.7400 +Query 1/1: Action query time = 0.990 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8761 +t=74: Selected seed 195 with value = 0.8761 +Query 1/1: Action query time = 0.985 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=22--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=22--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 22 +# successes: 20 (90.9%) + +Task: put the cream cheese in the bowl +Starting episode 23... +Query 1/1: Action query time = 0.990 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4503 +t=10: Selected seed 195 with value = 0.4503 +Query 1/1: Action query time = 0.981 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4677 +t=26: Selected seed 195 with value = 0.4677 +Query 1/1: Action query time = 0.997 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6091 +t=42: Selected seed 195 with value = 0.6091 +Query 1/1: Action query time = 0.968 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6760 +t=58: Selected seed 195 with value = 0.6760 +Query 1/1: Action query time = 0.978 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8162 +t=74: Selected seed 195 with value = 0.8162 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9474 +t=90: Selected seed 195 with value = 0.9474 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=23--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=23--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 23 +# successes: 21 (91.3%) + +Task: put the cream cheese in the bowl +Starting episode 24... +Query 1/1: Action query time = 1.112 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4562 +t=10: Selected seed 195 with value = 0.4562 +Query 1/1: Action query time = 0.982 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5550 +t=26: Selected seed 195 with value = 0.5550 +Query 1/1: Action query time = 0.975 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6292 +t=42: Selected seed 195 with value = 0.6292 +Query 1/1: Action query time = 0.989 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7234 +t=58: Selected seed 195 with value = 0.7234 +Query 1/1: Action query time = 0.977 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8386 +t=74: Selected seed 195 with value = 0.8386 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9830 +t=90: Selected seed 195 with value = 0.9830 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=24--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=24--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 24 +# successes: 22 (91.7%) + +Task: put the cream cheese in the bowl +Starting episode 25... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4215 +t=10: Selected seed 195 with value = 0.4215 +Query 1/1: Action query time = 0.979 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5424 +t=26: Selected seed 195 with value = 0.5424 +Query 1/1: Action query time = 0.969 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6361 +t=42: Selected seed 195 with value = 0.6361 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7631 +t=58: Selected seed 195 with value = 0.7631 +Query 1/1: Action query time = 0.970 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9033 +t=74: Selected seed 195 with value = 0.9033 +Query 1/1: Action query time = 0.976 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=90: Selected seed 195 with value = 0.9980 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=25--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=25--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 25 +# successes: 23 (92.0%) + +Task: put the cream cheese in the bowl +Starting episode 26... +Query 1/1: Action query time = 1.033 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4156 +t=10: Selected seed 195 with value = 0.4156 +Query 1/1: Action query time = 1.098 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3903 +t=26: Selected seed 195 with value = 0.3903 +Query 1/1: Action query time = 1.149 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4289 +t=42: Selected seed 195 with value = 0.4289 +Query 1/1: Action query time = 1.195 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5127 +t=58: Selected seed 195 with value = 0.5127 +Query 1/1: Action query time = 1.139 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5716 +t=74: Selected seed 195 with value = 0.5716 +Query 1/1: Action query time = 0.984 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6553 +t=90: Selected seed 195 with value = 0.6553 +Query 1/1: Action query time = 0.973 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7271 +t=106: Selected seed 195 with value = 0.7271 +Query 1/1: Action query time = 0.980 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7133 +t=122: Selected seed 195 with value = 0.7133 +Query 1/1: Action query time = 0.990 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8665 +t=138: Selected seed 195 with value = 0.8665 +Query 1/1: Action query time = 0.976 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9296 +t=154: Selected seed 195 with value = 0.9296 +Query 1/1: Action query time = 1.180 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9893 +t=170: Selected seed 195 with value = 0.9893 +Query 1/1: Action query time = 1.167 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9888 +t=186: Selected seed 195 with value = 0.9888 +Query 1/1: Action query time = 1.130 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9804 +t=202: Selected seed 195 with value = 0.9804 +Query 1/1: Action query time = 1.067 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9712 +t=218: Selected seed 195 with value = 0.9712 +Query 1/1: Action query time = 1.074 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9972 +t=234: Selected seed 195 with value = 0.9972 +Query 1/1: Action query time = 1.082 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=250: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 1.074 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=266: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 0.980 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9688 +t=282: Selected seed 195 with value = 0.9688 +Query 1/1: Action query time = 0.984 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9828 +t=298: Selected seed 195 with value = 0.9828 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=26--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=26--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 26 +# successes: 23 (88.5%) + +Task: put the cream cheese in the bowl +Starting episode 27... +Query 1/1: Action query time = 0.981 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4568 +t=10: Selected seed 195 with value = 0.4568 +Query 1/1: Action query time = 0.974 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5469 +t=26: Selected seed 195 with value = 0.5469 +Query 1/1: Action query time = 0.984 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6301 +t=42: Selected seed 195 with value = 0.6301 +Query 1/1: Action query time = 0.974 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7117 +t=58: Selected seed 195 with value = 0.7117 +Query 1/1: Action query time = 0.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8261 +t=74: Selected seed 195 with value = 0.8261 +Query 1/1: Action query time = 1.029 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=90: Selected seed 195 with value = 0.9946 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=27--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=27--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 27 +# successes: 24 (88.9%) + +Task: put the cream cheese in the bowl +Starting episode 28... +Query 1/1: Action query time = 1.059 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4253 +t=10: Selected seed 195 with value = 0.4253 +Query 1/1: Action query time = 1.064 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5455 +t=26: Selected seed 195 with value = 0.5455 +Query 1/1: Action query time = 1.046 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6217 +t=42: Selected seed 195 with value = 0.6217 +Query 1/1: Action query time = 0.963 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7346 +t=58: Selected seed 195 with value = 0.7346 +Query 1/1: Action query time = 0.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9059 +t=74: Selected seed 195 with value = 0.9059 +Query 1/1: Action query time = 0.974 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=90: Selected seed 195 with value = 0.9965 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=28--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=28--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 28 +# successes: 25 (89.3%) + +Task: put the cream cheese in the bowl +Starting episode 29... +Query 1/1: Action query time = 0.998 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4318 +t=10: Selected seed 195 with value = 0.4318 +Query 1/1: Action query time = 0.975 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5205 +t=26: Selected seed 195 with value = 0.5205 +Query 1/1: Action query time = 0.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6240 +t=42: Selected seed 195 with value = 0.6240 +Query 1/1: Action query time = 0.973 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4953 +t=58: Selected seed 195 with value = 0.4953 +Query 1/1: Action query time = 0.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5802 +t=74: Selected seed 195 with value = 0.5802 +Query 1/1: Action query time = 0.976 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6590 +t=90: Selected seed 195 with value = 0.6590 +Query 1/1: Action query time = 0.983 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7540 +t=106: Selected seed 195 with value = 0.7540 +Query 1/1: Action query time = 0.973 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9279 +t=122: Selected seed 195 with value = 0.9279 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=29--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=29--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 29 +# successes: 26 (89.7%) + +Task: put the cream cheese in the bowl +Starting episode 30... +Query 1/1: Action query time = 0.994 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4374 +t=10: Selected seed 195 with value = 0.4374 +Query 1/1: Action query time = 0.984 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4633 +t=26: Selected seed 195 with value = 0.4633 +Query 1/1: Action query time = 0.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6261 +t=42: Selected seed 195 with value = 0.6261 +Query 1/1: Action query time = 0.971 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7620 +t=58: Selected seed 195 with value = 0.7620 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9007 +t=74: Selected seed 195 with value = 0.9007 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=30--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=30--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 30 +# successes: 27 (90.0%) + +Task: put the cream cheese in the bowl +Starting episode 31... +Query 1/1: Action query time = 1.197 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4195 +t=10: Selected seed 195 with value = 0.4195 +Query 1/1: Action query time = 1.161 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5385 +t=26: Selected seed 195 with value = 0.5385 +Query 1/1: Action query time = 1.112 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6292 +t=42: Selected seed 195 with value = 0.6292 +Query 1/1: Action query time = 1.062 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7339 +t=58: Selected seed 195 with value = 0.7339 +Query 1/1: Action query time = 1.023 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8599 +t=74: Selected seed 195 with value = 0.8599 +Query 1/1: Action query time = 1.006 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=31--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=31--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 31 +# successes: 28 (90.3%) + +Task: put the cream cheese in the bowl +Starting episode 32... +Query 1/1: Action query time = 0.998 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4463 +t=10: Selected seed 195 with value = 0.4463 +Query 1/1: Action query time = 1.070 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4464 +t=26: Selected seed 195 with value = 0.4464 +Query 1/1: Action query time = 1.029 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5982 +t=42: Selected seed 195 with value = 0.5982 +Query 1/1: Action query time = 0.987 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4770 +t=58: Selected seed 195 with value = 0.4770 +Query 1/1: Action query time = 1.090 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5716 +t=74: Selected seed 195 with value = 0.5716 +Query 1/1: Action query time = 1.049 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6798 +t=90: Selected seed 195 with value = 0.6798 +Query 1/1: Action query time = 1.024 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7783 +t=106: Selected seed 195 with value = 0.7783 +Query 1/1: Action query time = 1.225 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9106 +t=122: Selected seed 195 with value = 0.9106 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=32--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=32--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 32 +# successes: 29 (90.6%) + +Task: put the cream cheese in the bowl +Starting episode 33... +Query 1/1: Action query time = 1.004 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4163 +t=10: Selected seed 195 with value = 0.4163 +Query 1/1: Action query time = 0.982 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4954 +t=26: Selected seed 195 with value = 0.4954 +Query 1/1: Action query time = 0.978 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6198 +t=42: Selected seed 195 with value = 0.6198 +Query 1/1: Action query time = 0.976 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7086 +t=58: Selected seed 195 with value = 0.7086 +Query 1/1: Action query time = 1.011 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8613 +t=74: Selected seed 195 with value = 0.8613 +Query 1/1: Action query time = 1.050 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=90: Selected seed 195 with value = 0.9958 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=33--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=33--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 33 +# successes: 30 (90.9%) + +Task: put the cream cheese in the bowl +Starting episode 34... +Query 1/1: Action query time = 0.998 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4410 +t=10: Selected seed 195 with value = 0.4410 +Query 1/1: Action query time = 1.213 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5275 +t=26: Selected seed 195 with value = 0.5275 +Query 1/1: Action query time = 1.221 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6380 +t=42: Selected seed 195 with value = 0.6380 +Query 1/1: Action query time = 1.160 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7579 +t=58: Selected seed 195 with value = 0.7579 +Query 1/1: Action query time = 1.019 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8975 +t=74: Selected seed 195 with value = 0.8975 +Query 1/1: Action query time = 1.006 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9995 +t=90: Selected seed 195 with value = 0.9995 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=34--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=34--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 34 +# successes: 31 (91.2%) + +Task: put the cream cheese in the bowl +Starting episode 35... +Query 1/1: Action query time = 1.001 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4406 +t=10: Selected seed 195 with value = 0.4406 +Query 1/1: Action query time = 1.050 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4912 +t=26: Selected seed 195 with value = 0.4912 +Query 1/1: Action query time = 0.991 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6358 +t=42: Selected seed 195 with value = 0.6358 +Query 1/1: Action query time = 0.989 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7400 +t=58: Selected seed 195 with value = 0.7400 +Query 1/1: Action query time = 0.985 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8548 +t=74: Selected seed 195 with value = 0.8548 +Query 1/1: Action query time = 0.979 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9706 +t=90: Selected seed 195 with value = 0.9706 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=35--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=35--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 35 +# successes: 32 (91.4%) + +Task: put the cream cheese in the bowl +Starting episode 36... +Query 1/1: Action query time = 0.990 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4235 +t=10: Selected seed 195 with value = 0.4235 +Query 1/1: Action query time = 0.989 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5595 +t=26: Selected seed 195 with value = 0.5595 +Query 1/1: Action query time = 0.976 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6603 +t=42: Selected seed 195 with value = 0.6603 +Query 1/1: Action query time = 0.985 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7554 +t=58: Selected seed 195 with value = 0.7554 +Query 1/1: Action query time = 0.978 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8872 +t=74: Selected seed 195 with value = 0.8872 +Query 1/1: Action query time = 0.979 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9953 +t=90: Selected seed 195 with value = 0.9953 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=36--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=36--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 36 +# successes: 33 (91.7%) + +Task: put the cream cheese in the bowl +Starting episode 37... +Query 1/1: Action query time = 1.025 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4301 +t=10: Selected seed 195 with value = 0.4301 +Query 1/1: Action query time = 1.093 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4724 +t=26: Selected seed 195 with value = 0.4724 +Query 1/1: Action query time = 1.163 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6277 +t=42: Selected seed 195 with value = 0.6277 +Query 1/1: Action query time = 1.197 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7245 +t=58: Selected seed 195 with value = 0.7245 +Query 1/1: Action query time = 0.988 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8712 +t=74: Selected seed 195 with value = 0.8712 +Query 1/1: Action query time = 0.988 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.004 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=106: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 0.974 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=122: Selected seed 195 with value = 0.9976 +Query 1/1: Action query time = 0.988 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=138: Selected seed 195 with value = 0.9980 +Query 1/1: Action query time = 0.971 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9943 +t=154: Selected seed 195 with value = 0.9943 +Query 1/1: Action query time = 0.992 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9952 +t=170: Selected seed 195 with value = 0.9952 +Query 1/1: Action query time = 0.991 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=186: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 0.983 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=202: Selected seed 195 with value = 0.9946 +Query 1/1: Action query time = 0.981 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=234: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 0.981 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9895 +t=250: Selected seed 195 with value = 0.9895 +Query 1/1: Action query time = 1.005 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=282: Selected seed 195 with value = 0.9924 +Query 1/1: Action query time = 1.113 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9896 +t=298: Selected seed 195 with value = 0.9896 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=37--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=37--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 37 +# successes: 33 (89.2%) + +Task: put the cream cheese in the bowl +Starting episode 38... +Query 1/1: Action query time = 0.998 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4545 +t=10: Selected seed 195 with value = 0.4545 +Query 1/1: Action query time = 0.990 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4800 +t=26: Selected seed 195 with value = 0.4800 +Query 1/1: Action query time = 0.991 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5783 +t=42: Selected seed 195 with value = 0.5783 +Query 1/1: Action query time = 0.993 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6861 +t=58: Selected seed 195 with value = 0.6861 +Query 1/1: Action query time = 0.989 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8096 +t=74: Selected seed 195 with value = 0.8096 +Query 1/1: Action query time = 0.996 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9316 +t=90: Selected seed 195 with value = 0.9316 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=38--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=38--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 38 +# successes: 34 (89.5%) + +Task: put the cream cheese in the bowl +Starting episode 39... +Query 1/1: Action query time = 0.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4301 +t=10: Selected seed 195 with value = 0.4301 +Query 1/1: Action query time = 0.980 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5558 +t=26: Selected seed 195 with value = 0.5558 +Query 1/1: Action query time = 0.983 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6489 +t=42: Selected seed 195 with value = 0.6489 +Query 1/1: Action query time = 0.972 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7631 +t=58: Selected seed 195 with value = 0.7631 +Query 1/1: Action query time = 0.982 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9117 +t=74: Selected seed 195 with value = 0.9117 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=39--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=39--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 39 +# successes: 35 (89.7%) + +Task: put the cream cheese in the bowl +Starting episode 40... +Query 1/1: Action query time = 0.997 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4635 +t=10: Selected seed 195 with value = 0.4635 +Query 1/1: Action query time = 0.964 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5480 +t=26: Selected seed 195 with value = 0.5480 +Query 1/1: Action query time = 0.965 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6406 +t=42: Selected seed 195 with value = 0.6406 +Query 1/1: Action query time = 0.978 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6871 +t=58: Selected seed 195 with value = 0.6871 +Query 1/1: Action query time = 0.977 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8252 +t=74: Selected seed 195 with value = 0.8252 +Query 1/1: Action query time = 0.992 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9423 +t=90: Selected seed 195 with value = 0.9423 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=40--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=40--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 40 +# successes: 36 (90.0%) + +Task: put the cream cheese in the bowl +Starting episode 41... +Query 1/1: Action query time = 0.999 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4477 +t=10: Selected seed 195 with value = 0.4477 +Query 1/1: Action query time = 1.004 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4903 +t=26: Selected seed 195 with value = 0.4903 +Query 1/1: Action query time = 0.986 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6182 +t=42: Selected seed 195 with value = 0.6182 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5478 +t=58: Selected seed 195 with value = 0.5478 +Query 1/1: Action query time = 0.999 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6319 +t=74: Selected seed 195 with value = 0.6319 +Query 1/1: Action query time = 0.988 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7027 +t=90: Selected seed 195 with value = 0.7027 +Query 1/1: Action query time = 0.970 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8319 +t=106: Selected seed 195 with value = 0.8319 +Query 1/1: Action query time = 0.976 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8987 +t=122: Selected seed 195 with value = 0.8987 +Query 1/1: Action query time = 0.971 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9186 +t=138: Selected seed 195 with value = 0.9186 +Query 1/1: Action query time = 0.970 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9331 +t=154: Selected seed 195 with value = 0.9331 +Query 1/1: Action query time = 0.973 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9967 +t=202: Selected seed 195 with value = 0.9967 +Query 1/1: Action query time = 0.979 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=218: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 0.975 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=234: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 1.019 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=250: Selected seed 195 with value = 0.9965 +Query 1/1: Action query time = 0.973 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=266: Selected seed 195 with value = 0.9965 +Query 1/1: Action query time = 0.982 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=282: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 0.991 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=298: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=41--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=41--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 41 +# successes: 36 (87.8%) + +Task: put the cream cheese in the bowl +Starting episode 42... +Query 1/1: Action query time = 1.062 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4134 +t=10: Selected seed 195 with value = 0.4134 +Query 1/1: Action query time = 1.002 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5637 +t=26: Selected seed 195 with value = 0.5637 +Query 1/1: Action query time = 1.054 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6265 +t=42: Selected seed 195 with value = 0.6265 +Query 1/1: Action query time = 1.034 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7132 +t=58: Selected seed 195 with value = 0.7132 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8534 +t=74: Selected seed 195 with value = 0.8534 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9413 +t=90: Selected seed 195 with value = 0.9413 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=42--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=42--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 42 +# successes: 37 (88.1%) + +Task: put the cream cheese in the bowl +Starting episode 43... +Query 1/1: Action query time = 0.986 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4395 +t=10: Selected seed 195 with value = 0.4395 +Query 1/1: Action query time = 0.964 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4657 +t=26: Selected seed 195 with value = 0.4657 +Query 1/1: Action query time = 0.978 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5787 +t=42: Selected seed 195 with value = 0.5787 +Query 1/1: Action query time = 0.990 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6823 +t=58: Selected seed 195 with value = 0.6823 +Query 1/1: Action query time = 0.984 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7812 +t=74: Selected seed 195 with value = 0.7812 +Query 1/1: Action query time = 0.990 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8804 +t=90: Selected seed 195 with value = 0.8804 +Query 1/1: Action query time = 0.986 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=106: Selected seed 195 with value = 0.9924 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=43--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=43--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 43 +# successes: 38 (88.4%) + +Task: put the cream cheese in the bowl +Starting episode 44... +Query 1/1: Action query time = 0.996 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4469 +t=10: Selected seed 195 with value = 0.4469 +Query 1/1: Action query time = 0.981 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4901 +t=26: Selected seed 195 with value = 0.4901 +Query 1/1: Action query time = 0.966 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6258 +t=42: Selected seed 195 with value = 0.6258 +Query 1/1: Action query time = 0.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7485 +t=58: Selected seed 195 with value = 0.7485 +Query 1/1: Action query time = 0.988 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8450 +t=74: Selected seed 195 with value = 0.8450 +Query 1/1: Action query time = 0.984 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9738 +t=90: Selected seed 195 with value = 0.9738 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=44--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=44--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 44 +# successes: 39 (88.6%) + +Task: put the cream cheese in the bowl +Starting episode 45... +Query 1/1: Action query time = 0.986 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4676 +t=10: Selected seed 195 with value = 0.4676 +Query 1/1: Action query time = 0.984 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4755 +t=26: Selected seed 195 with value = 0.4755 +Query 1/1: Action query time = 0.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6272 +t=42: Selected seed 195 with value = 0.6272 +Query 1/1: Action query time = 0.991 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7282 +t=58: Selected seed 195 with value = 0.7282 +Query 1/1: Action query time = 0.990 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8776 +t=74: Selected seed 195 with value = 0.8776 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=90: Selected seed 195 with value = 0.9958 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=45--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=45--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 45 +# successes: 40 (88.9%) + +Task: put the cream cheese in the bowl +Starting episode 46... +Query 1/1: Action query time = 0.980 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4891 +t=10: Selected seed 195 with value = 0.4891 +Query 1/1: Action query time = 0.985 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5667 +t=26: Selected seed 195 with value = 0.5667 +Query 1/1: Action query time = 0.967 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6630 +t=42: Selected seed 195 with value = 0.6630 +Query 1/1: Action query time = 0.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7530 +t=58: Selected seed 195 with value = 0.7530 +Query 1/1: Action query time = 0.991 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9212 +t=74: Selected seed 195 with value = 0.9212 +Query 1/1: Action query time = 1.007 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.993 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.997 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.988 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.005 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.995 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=186: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 1.001 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9934 +t=202: Selected seed 195 with value = 0.9934 +Query 1/1: Action query time = 0.997 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9927 +t=218: Selected seed 195 with value = 0.9927 +Query 1/1: Action query time = 0.991 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9923 +t=234: Selected seed 195 with value = 0.9923 +Query 1/1: Action query time = 0.971 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=250: Selected seed 195 with value = 0.9924 +Query 1/1: Action query time = 0.993 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=266: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 0.992 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=282: Selected seed 195 with value = 0.9926 +Query 1/1: Action query time = 0.973 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=298: Selected seed 195 with value = 0.9928 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=46--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=46--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 46 +# successes: 40 (87.0%) + +Task: put the cream cheese in the bowl +Starting episode 47... +Query 1/1: Action query time = 1.056 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4483 +t=10: Selected seed 195 with value = 0.4483 +Query 1/1: Action query time = 1.000 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5151 +t=26: Selected seed 195 with value = 0.5151 +Query 1/1: Action query time = 0.979 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6196 +t=42: Selected seed 195 with value = 0.6196 +Query 1/1: Action query time = 0.988 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4870 +t=58: Selected seed 195 with value = 0.4870 +Query 1/1: Action query time = 0.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8231 +t=74: Selected seed 195 with value = 0.8231 +Query 1/1: Action query time = 0.993 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9440 +t=90: Selected seed 195 with value = 0.9440 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=47--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=47--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 47 +# successes: 41 (87.2%) + +Task: put the cream cheese in the bowl +Starting episode 48... +Query 1/1: Action query time = 0.993 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4213 +t=10: Selected seed 195 with value = 0.4213 +Query 1/1: Action query time = 0.974 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4823 +t=26: Selected seed 195 with value = 0.4823 +Query 1/1: Action query time = 0.985 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5813 +t=42: Selected seed 195 with value = 0.5813 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7244 +t=58: Selected seed 195 with value = 0.7244 +Query 1/1: Action query time = 0.973 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8110 +t=74: Selected seed 195 with value = 0.8110 +Query 1/1: Action query time = 0.966 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9409 +t=90: Selected seed 195 with value = 0.9409 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=48--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=48--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 48 +# successes: 42 (87.5%) + +Task: put the cream cheese in the bowl +Starting episode 49... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4308 +t=10: Selected seed 195 with value = 0.4308 +Query 1/1: Action query time = 0.981 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5380 +t=26: Selected seed 195 with value = 0.5380 +Query 1/1: Action query time = 0.993 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6401 +t=42: Selected seed 195 with value = 0.6401 +Query 1/1: Action query time = 0.978 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7526 +t=58: Selected seed 195 with value = 0.7526 +Query 1/1: Action query time = 0.973 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9043 +t=74: Selected seed 195 with value = 0.9043 +Query 1/1: Action query time = 0.984 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=90: Selected seed 195 with value = 0.9986 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=49--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=49--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 49 +# successes: 43 (87.8%) + +Task: put the cream cheese in the bowl +Starting episode 50... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4235 +t=10: Selected seed 195 with value = 0.4235 +Query 1/1: Action query time = 0.987 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5272 +t=26: Selected seed 195 with value = 0.5272 +Query 1/1: Action query time = 1.043 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6370 +t=42: Selected seed 195 with value = 0.6370 +Query 1/1: Action query time = 0.972 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6883 +t=58: Selected seed 195 with value = 0.6883 +Query 1/1: Action query time = 0.984 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6325 +t=74: Selected seed 195 with value = 0.6325 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6345 +t=90: Selected seed 195 with value = 0.6345 +Query 1/1: Action query time = 0.980 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7197 +t=106: Selected seed 195 with value = 0.7197 +Query 1/1: Action query time = 0.978 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8519 +t=122: Selected seed 195 with value = 0.8519 +Query 1/1: Action query time = 0.985 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7961 +t=138: Selected seed 195 with value = 0.7961 +Query 1/1: Action query time = 0.988 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8584 +t=154: Selected seed 195 with value = 0.8584 +Query 1/1: Action query time = 0.978 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9629 +t=170: Selected seed 195 with value = 0.9629 +Query 1/1: Action query time = 0.983 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.014 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9921 +t=202: Selected seed 195 with value = 0.9921 +Query 1/1: Action query time = 0.980 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.991 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.003 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=250: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 0.975 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9940 +t=266: Selected seed 195 with value = 0.9940 +Query 1/1: Action query time = 1.003 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9805 +t=282: Selected seed 195 with value = 0.9805 +Query 1/1: Action query time = 0.990 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9729 +t=298: Selected seed 195 with value = 0.9729 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--episode=50--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t6/2026_07_31-14_46_45--with_future_img--episode=50--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 50 +# successes: 43 (86.0%) +Current task success rate: 0.86 +Current total success rate: 0.86 +Final results: +Total episodes: 50 +Total successes: 43 +Overall success rate: 0.8600 (86.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_07_31-15_24_36--CLi800_t1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_07_31-15_24_36--CLi800_t1.txt new file mode 100644 index 0000000000000000000000000000000000000000..d7f6a93342198416774fab89ea08a472b4a07f2e --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_07_31-15_24_36--CLi800_t1.txt @@ -0,0 +1,4212 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='CLi800_t1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=True, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 1.507 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4853 +t=10: Selected seed 195 with value = 0.4853 +Query 1/1: Action query time = 0.965 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5754 +t=26: Selected seed 195 with value = 0.5754 +Query 1/1: Action query time = 0.954 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6314 +t=42: Selected seed 195 with value = 0.6314 +Query 1/1: Action query time = 0.963 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6356 +t=58: Selected seed 195 with value = 0.6356 +Query 1/1: Action query time = 0.968 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7971 +t=74: Selected seed 195 with value = 0.7971 +Query 1/1: Action query time = 0.968 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8022 +t=90: Selected seed 195 with value = 0.8022 +Query 1/1: Action query time = 0.960 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9582 +t=106: Selected seed 195 with value = 0.9582 +Query 1/1: Action query time = 0.962 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.957 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=138: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 0.978 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=154: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 0.966 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=170: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 0.981 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9956 +t=186: Selected seed 195 with value = 0.9956 +Query 1/1: Action query time = 0.972 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9884 +t=202: Selected seed 195 with value = 0.9884 +Query 1/1: Action query time = 0.965 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9712 +t=218: Selected seed 195 with value = 0.9712 +Query 1/1: Action query time = 0.971 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9384 +t=234: Selected seed 195 with value = 0.9384 +Query 1/1: Action query time = 0.962 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8770 +t=250: Selected seed 195 with value = 0.8770 +Query 1/1: Action query time = 0.958 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8128 +t=266: Selected seed 195 with value = 0.8128 +Query 1/1: Action query time = 0.947 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7805 +t=282: Selected seed 195 with value = 0.7805 +Query 1/1: Action query time = 0.951 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7848 +t=298: Selected seed 195 with value = 0.7848 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 0.981 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4660 +t=10: Selected seed 195 with value = 0.4660 +Query 1/1: Action query time = 0.963 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5678 +t=26: Selected seed 195 with value = 0.5678 +Query 1/1: Action query time = 0.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5932 +t=42: Selected seed 195 with value = 0.5932 +Query 1/1: Action query time = 0.950 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7389 +t=58: Selected seed 195 with value = 0.7389 +Query 1/1: Action query time = 0.953 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8727 +t=74: Selected seed 195 with value = 0.8727 +Query 1/1: Action query time = 0.952 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=90: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 0.967 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.951 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.961 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.951 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.087 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=202: Selected seed 195 with value = 0.9997 +Query 1/1: Action query time = 1.169 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=218: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 1.139 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.023 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.959 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.961 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 0.972 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4828 +t=10: Selected seed 195 with value = 0.4828 +Query 1/1: Action query time = 0.970 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5619 +t=26: Selected seed 195 with value = 0.5619 +Query 1/1: Action query time = 0.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6140 +t=42: Selected seed 195 with value = 0.6140 +Query 1/1: Action query time = 0.965 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7081 +t=58: Selected seed 195 with value = 0.7081 +Query 1/1: Action query time = 0.988 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8398 +t=74: Selected seed 195 with value = 0.8398 +Query 1/1: Action query time = 0.957 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9427 +t=90: Selected seed 195 with value = 0.9427 +Query 1/1: Action query time = 0.968 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.961 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.166 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.140 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.136 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.105 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.047 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.996 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=234: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 0.967 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=250: Selected seed 195 with value = 0.9971 +Query 1/1: Action query time = 0.973 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=266: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 0.993 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=282: Selected seed 195 with value = 0.9926 +Query 1/1: Action query time = 1.078 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9923 +t=298: Selected seed 195 with value = 0.9923 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 0.972 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4911 +t=10: Selected seed 195 with value = 0.4911 +Query 1/1: Action query time = 0.958 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6030 +t=26: Selected seed 195 with value = 0.6030 +Query 1/1: Action query time = 0.987 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6525 +t=42: Selected seed 195 with value = 0.6525 +Query 1/1: Action query time = 0.971 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7056 +t=58: Selected seed 195 with value = 0.7056 +Query 1/1: Action query time = 0.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8221 +t=74: Selected seed 195 with value = 0.8221 +Query 1/1: Action query time = 0.960 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8662 +t=90: Selected seed 195 with value = 0.8662 +Query 1/1: Action query time = 0.952 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=106: Selected seed 195 with value = 0.9867 +Query 1/1: Action query time = 0.965 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9974 +t=122: Selected seed 195 with value = 0.9974 +Query 1/1: Action query time = 0.953 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.970 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.954 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=218: Selected seed 195 with value = 0.9976 +Query 1/1: Action query time = 0.969 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=234: Selected seed 195 with value = 0.9914 +Query 1/1: Action query time = 0.954 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9822 +t=250: Selected seed 195 with value = 0.9822 +Query 1/1: Action query time = 0.951 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9603 +t=266: Selected seed 195 with value = 0.9603 +Query 1/1: Action query time = 0.943 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9052 +t=282: Selected seed 195 with value = 0.9052 +Query 1/1: Action query time = 0.964 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8265 +t=298: Selected seed 195 with value = 0.8265 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 0.977 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4864 +t=10: Selected seed 195 with value = 0.4864 +Query 1/1: Action query time = 0.962 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5460 +t=26: Selected seed 195 with value = 0.5460 +Query 1/1: Action query time = 0.959 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6419 +t=42: Selected seed 195 with value = 0.6419 +Query 1/1: Action query time = 0.954 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7302 +t=58: Selected seed 195 with value = 0.7302 +Query 1/1: Action query time = 0.964 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8634 +t=74: Selected seed 195 with value = 0.8634 +Query 1/1: Action query time = 0.950 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9574 +t=90: Selected seed 195 with value = 0.9574 +Query 1/1: Action query time = 0.955 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.961 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.087 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.097 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.144 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.198 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.188 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.194 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.168 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=5--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=5--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 5 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 0.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4841 +t=10: Selected seed 195 with value = 0.4841 +Query 1/1: Action query time = 0.972 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5077 +t=26: Selected seed 195 with value = 0.5077 +Query 1/1: Action query time = 0.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6043 +t=42: Selected seed 195 with value = 0.6043 +Query 1/1: Action query time = 0.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7269 +t=58: Selected seed 195 with value = 0.7269 +Query 1/1: Action query time = 0.968 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7726 +t=74: Selected seed 195 with value = 0.7726 +Query 1/1: Action query time = 0.983 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8439 +t=90: Selected seed 195 with value = 0.8439 +Query 1/1: Action query time = 0.998 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9105 +t=106: Selected seed 195 with value = 0.9105 +Query 1/1: Action query time = 0.970 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=122: Selected seed 195 with value = 0.9958 +Query 1/1: Action query time = 0.954 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=138: Selected seed 195 with value = 0.9926 +Query 1/1: Action query time = 0.968 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=154: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 0.959 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=170: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 0.949 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=186: Selected seed 195 with value = 0.9989 +Query 1/1: Action query time = 0.957 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=202: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 0.970 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9918 +t=218: Selected seed 195 with value = 0.9918 +Query 1/1: Action query time = 0.968 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9823 +t=234: Selected seed 195 with value = 0.9823 +Query 1/1: Action query time = 0.957 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9706 +t=250: Selected seed 195 with value = 0.9706 +Query 1/1: Action query time = 0.958 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9346 +t=266: Selected seed 195 with value = 0.9346 +Query 1/1: Action query time = 0.969 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8718 +t=282: Selected seed 195 with value = 0.8718 +Query 1/1: Action query time = 0.977 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8295 +t=298: Selected seed 195 with value = 0.8295 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=6--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=6--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 6 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 1.000 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4762 +t=10: Selected seed 195 with value = 0.4762 +Query 1/1: Action query time = 0.973 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5565 +t=26: Selected seed 195 with value = 0.5565 +Query 1/1: Action query time = 0.949 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6185 +t=42: Selected seed 195 with value = 0.6185 +Query 1/1: Action query time = 0.959 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7353 +t=58: Selected seed 195 with value = 0.7353 +Query 1/1: Action query time = 0.974 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8806 +t=74: Selected seed 195 with value = 0.8806 +Query 1/1: Action query time = 0.964 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.949 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=154: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 0.969 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9778 +t=170: Selected seed 195 with value = 0.9778 +Query 1/1: Action query time = 0.963 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9913 +t=186: Selected seed 195 with value = 0.9913 +Query 1/1: Action query time = 0.975 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9855 +t=202: Selected seed 195 with value = 0.9855 +Query 1/1: Action query time = 0.959 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9893 +t=218: Selected seed 195 with value = 0.9893 +Query 1/1: Action query time = 0.961 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9907 +t=234: Selected seed 195 with value = 0.9907 +Query 1/1: Action query time = 0.966 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9879 +t=250: Selected seed 195 with value = 0.9879 +Query 1/1: Action query time = 0.953 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9849 +t=266: Selected seed 195 with value = 0.9849 +Query 1/1: Action query time = 0.963 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9772 +t=282: Selected seed 195 with value = 0.9772 +Query 1/1: Action query time = 0.956 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9882 +t=298: Selected seed 195 with value = 0.9882 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=7--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=7--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 7 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 0.968 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4755 +t=10: Selected seed 195 with value = 0.4755 +Query 1/1: Action query time = 0.963 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5949 +t=26: Selected seed 195 with value = 0.5949 +Query 1/1: Action query time = 0.951 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6400 +t=42: Selected seed 195 with value = 0.6400 +Query 1/1: Action query time = 0.963 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6585 +t=58: Selected seed 195 with value = 0.6585 +Query 1/1: Action query time = 0.950 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7604 +t=74: Selected seed 195 with value = 0.7604 +Query 1/1: Action query time = 0.970 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7846 +t=90: Selected seed 195 with value = 0.7846 +Query 1/1: Action query time = 0.964 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8268 +t=106: Selected seed 195 with value = 0.8268 +Query 1/1: Action query time = 0.974 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9372 +t=122: Selected seed 195 with value = 0.9372 +Query 1/1: Action query time = 0.954 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=154: Selected seed 195 with value = 0.9971 +Query 1/1: Action query time = 0.959 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=170: Selected seed 195 with value = 0.9980 +Query 1/1: Action query time = 0.960 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=186: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 0.967 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9951 +t=202: Selected seed 195 with value = 0.9951 +Query 1/1: Action query time = 0.960 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=218: Selected seed 195 with value = 0.9864 +Query 1/1: Action query time = 0.972 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9704 +t=234: Selected seed 195 with value = 0.9704 +Query 1/1: Action query time = 0.977 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9408 +t=250: Selected seed 195 with value = 0.9408 +Query 1/1: Action query time = 0.957 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8786 +t=266: Selected seed 195 with value = 0.8786 +Query 1/1: Action query time = 0.965 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8219 +t=282: Selected seed 195 with value = 0.8219 +Query 1/1: Action query time = 0.963 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8040 +t=298: Selected seed 195 with value = 0.8040 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=8--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=8--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 8 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 0.967 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4731 +t=10: Selected seed 195 with value = 0.4731 +Query 1/1: Action query time = 0.962 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5491 +t=26: Selected seed 195 with value = 0.5491 +Query 1/1: Action query time = 0.961 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6220 +t=42: Selected seed 195 with value = 0.6220 +Query 1/1: Action query time = 0.968 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7413 +t=58: Selected seed 195 with value = 0.7413 +Query 1/1: Action query time = 0.962 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8767 +t=74: Selected seed 195 with value = 0.8767 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.957 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9923 +t=250: Selected seed 195 with value = 0.9923 +Query 1/1: Action query time = 0.968 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9780 +t=266: Selected seed 195 with value = 0.9780 +Query 1/1: Action query time = 0.972 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9835 +t=282: Selected seed 195 with value = 0.9835 +Query 1/1: Action query time = 0.967 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9214 +t=298: Selected seed 195 with value = 0.9214 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 9 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 1.196 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5084 +t=10: Selected seed 195 with value = 0.5084 +Query 1/1: Action query time = 0.960 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5476 +t=26: Selected seed 195 with value = 0.5476 +Query 1/1: Action query time = 0.960 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6119 +t=42: Selected seed 195 with value = 0.6119 +Query 1/1: Action query time = 0.977 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7375 +t=58: Selected seed 195 with value = 0.7375 +Query 1/1: Action query time = 0.961 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8321 +t=74: Selected seed 195 with value = 0.8321 +Query 1/1: Action query time = 0.967 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9225 +t=90: Selected seed 195 with value = 0.9225 +Query 1/1: Action query time = 0.961 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9959 +t=106: Selected seed 195 with value = 0.9959 +Query 1/1: Action query time = 0.978 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9936 +t=122: Selected seed 195 with value = 0.9936 +Query 1/1: Action query time = 1.109 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=138: Selected seed 195 with value = 0.9976 +Query 1/1: Action query time = 1.098 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9953 +t=154: Selected seed 195 with value = 0.9953 +Query 1/1: Action query time = 0.949 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9900 +t=170: Selected seed 195 with value = 0.9900 +Query 1/1: Action query time = 0.976 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9808 +t=186: Selected seed 195 with value = 0.9808 +Query 1/1: Action query time = 0.976 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9579 +t=202: Selected seed 195 with value = 0.9579 +Query 1/1: Action query time = 1.113 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9187 +t=218: Selected seed 195 with value = 0.9187 +Query 1/1: Action query time = 1.149 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8483 +t=234: Selected seed 195 with value = 0.8483 +Query 1/1: Action query time = 1.190 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8053 +t=250: Selected seed 195 with value = 0.8053 +Query 1/1: Action query time = 1.187 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7979 +t=266: Selected seed 195 with value = 0.7979 +Query 1/1: Action query time = 1.161 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7968 +t=282: Selected seed 195 with value = 0.7968 +Query 1/1: Action query time = 1.119 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7984 +t=298: Selected seed 195 with value = 0.7984 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=10--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=10--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 10 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 0.969 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4759 +t=10: Selected seed 195 with value = 0.4759 +Query 1/1: Action query time = 0.972 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6026 +t=26: Selected seed 195 with value = 0.6026 +Query 1/1: Action query time = 0.985 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6525 +t=42: Selected seed 195 with value = 0.6525 +Query 1/1: Action query time = 0.956 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7343 +t=58: Selected seed 195 with value = 0.7343 +Query 1/1: Action query time = 0.978 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7895 +t=74: Selected seed 195 with value = 0.7895 +Query 1/1: Action query time = 0.955 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8219 +t=90: Selected seed 195 with value = 0.8219 +Query 1/1: Action query time = 0.972 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9198 +t=106: Selected seed 195 with value = 0.9198 +Query 1/1: Action query time = 0.957 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9194 +t=122: Selected seed 195 with value = 0.9194 +Query 1/1: Action query time = 0.956 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9071 +t=138: Selected seed 195 with value = 0.9071 +Query 1/1: Action query time = 0.966 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9919 +t=154: Selected seed 195 with value = 0.9919 +Query 1/1: Action query time = 0.975 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=170: Selected seed 195 with value = 0.9871 +Query 1/1: Action query time = 0.979 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9863 +t=186: Selected seed 195 with value = 0.9863 +Query 1/1: Action query time = 0.964 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9717 +t=202: Selected seed 195 with value = 0.9717 +Query 1/1: Action query time = 1.121 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9284 +t=218: Selected seed 195 with value = 0.9284 +Query 1/1: Action query time = 1.166 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8588 +t=234: Selected seed 195 with value = 0.8588 +Query 1/1: Action query time = 1.179 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8188 +t=250: Selected seed 195 with value = 0.8188 +Query 1/1: Action query time = 1.108 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8247 +t=266: Selected seed 195 with value = 0.8247 +Query 1/1: Action query time = 1.087 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8360 +t=282: Selected seed 195 with value = 0.8360 +Query 1/1: Action query time = 1.044 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8355 +t=298: Selected seed 195 with value = 0.8355 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=11--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=11--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 11 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 1.192 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4925 +t=10: Selected seed 195 with value = 0.4925 +Query 1/1: Action query time = 1.201 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5207 +t=26: Selected seed 195 with value = 0.5207 +Query 1/1: Action query time = 1.186 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6269 +t=42: Selected seed 195 with value = 0.6269 +Query 1/1: Action query time = 1.159 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6507 +t=58: Selected seed 195 with value = 0.6507 +Query 1/1: Action query time = 1.154 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7532 +t=74: Selected seed 195 with value = 0.7532 +Query 1/1: Action query time = 1.171 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8294 +t=90: Selected seed 195 with value = 0.8294 +Query 1/1: Action query time = 1.196 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8975 +t=106: Selected seed 195 with value = 0.8975 +Query 1/1: Action query time = 1.158 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9878 +t=122: Selected seed 195 with value = 0.9878 +Query 1/1: Action query time = 0.966 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9656 +t=138: Selected seed 195 with value = 0.9656 +Query 1/1: Action query time = 0.980 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.991 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9937 +t=170: Selected seed 195 with value = 0.9937 +Query 1/1: Action query time = 0.963 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9938 +t=186: Selected seed 195 with value = 0.9938 +Query 1/1: Action query time = 0.974 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9884 +t=202: Selected seed 195 with value = 0.9884 +Query 1/1: Action query time = 0.971 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9746 +t=218: Selected seed 195 with value = 0.9746 +Query 1/1: Action query time = 0.984 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9458 +t=234: Selected seed 195 with value = 0.9458 +Query 1/1: Action query time = 0.980 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9056 +t=250: Selected seed 195 with value = 0.9056 +Query 1/1: Action query time = 0.965 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8558 +t=266: Selected seed 195 with value = 0.8558 +Query 1/1: Action query time = 0.980 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8422 +t=282: Selected seed 195 with value = 0.8422 +Query 1/1: Action query time = 0.958 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8333 +t=298: Selected seed 195 with value = 0.8333 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=12--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=12--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 12 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 0.970 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4890 +t=10: Selected seed 195 with value = 0.4890 +Query 1/1: Action query time = 0.958 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5429 +t=26: Selected seed 195 with value = 0.5429 +Query 1/1: Action query time = 0.971 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6131 +t=42: Selected seed 195 with value = 0.6131 +Query 1/1: Action query time = 0.967 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6823 +t=58: Selected seed 195 with value = 0.6823 +Query 1/1: Action query time = 0.981 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7914 +t=74: Selected seed 195 with value = 0.7914 +Query 1/1: Action query time = 0.973 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8393 +t=90: Selected seed 195 with value = 0.8393 +Query 1/1: Action query time = 0.961 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8068 +t=106: Selected seed 195 with value = 0.8068 +Query 1/1: Action query time = 0.965 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9426 +t=122: Selected seed 195 with value = 0.9426 +Query 1/1: Action query time = 0.979 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9819 +t=154: Selected seed 195 with value = 0.9819 +Query 1/1: Action query time = 0.950 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=170: Selected seed 195 with value = 0.9958 +Query 1/1: Action query time = 0.974 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=186: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 0.973 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9756 +t=202: Selected seed 195 with value = 0.9756 +Query 1/1: Action query time = 0.955 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9378 +t=218: Selected seed 195 with value = 0.9378 +Query 1/1: Action query time = 0.970 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8596 +t=234: Selected seed 195 with value = 0.8596 +Query 1/1: Action query time = 0.985 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8041 +t=250: Selected seed 195 with value = 0.8041 +Query 1/1: Action query time = 0.969 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7953 +t=266: Selected seed 195 with value = 0.7953 +Query 1/1: Action query time = 0.961 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7944 +t=282: Selected seed 195 with value = 0.7944 +Query 1/1: Action query time = 0.964 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8084 +t=298: Selected seed 195 with value = 0.8084 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=13--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=13--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 13 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 14... +Query 1/1: Action query time = 0.965 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4791 +t=10: Selected seed 195 with value = 0.4791 +Query 1/1: Action query time = 0.957 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5408 +t=26: Selected seed 195 with value = 0.5408 +Query 1/1: Action query time = 0.958 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6438 +t=42: Selected seed 195 with value = 0.6438 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7358 +t=58: Selected seed 195 with value = 0.7358 +Query 1/1: Action query time = 0.966 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8728 +t=74: Selected seed 195 with value = 0.8728 +Query 1/1: Action query time = 0.975 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9816 +t=90: Selected seed 195 with value = 0.9816 +Query 1/1: Action query time = 0.965 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.989 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.956 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.042 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.030 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=14--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=14--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 14 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 15... +Query 1/1: Action query time = 0.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5004 +t=10: Selected seed 195 with value = 0.5004 +Query 1/1: Action query time = 0.955 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5811 +t=26: Selected seed 195 with value = 0.5811 +Query 1/1: Action query time = 0.979 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6003 +t=42: Selected seed 195 with value = 0.6003 +Query 1/1: Action query time = 0.965 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7438 +t=58: Selected seed 195 with value = 0.7438 +Query 1/1: Action query time = 0.969 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8868 +t=74: Selected seed 195 with value = 0.8868 +Query 1/1: Action query time = 0.954 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.959 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.991 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9950 +t=234: Selected seed 195 with value = 0.9950 +Query 1/1: Action query time = 1.054 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9628 +t=250: Selected seed 195 with value = 0.9628 +Query 1/1: Action query time = 1.209 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9900 +t=266: Selected seed 195 with value = 0.9900 +Query 1/1: Action query time = 1.304 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9474 +t=282: Selected seed 195 with value = 0.9474 +Query 1/1: Action query time = 1.221 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9273 +t=298: Selected seed 195 with value = 0.9273 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=15--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=15--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 15 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 16... +Query 1/1: Action query time = 0.994 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4592 +t=10: Selected seed 195 with value = 0.4592 +Query 1/1: Action query time = 0.984 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5042 +t=26: Selected seed 195 with value = 0.5042 +Query 1/1: Action query time = 0.976 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5984 +t=42: Selected seed 195 with value = 0.5984 +Query 1/1: Action query time = 0.971 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6828 +t=58: Selected seed 195 with value = 0.6828 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8566 +t=74: Selected seed 195 with value = 0.8566 +Query 1/1: Action query time = 0.969 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9653 +t=90: Selected seed 195 with value = 0.9653 +Query 1/1: Action query time = 0.960 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.001 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.035 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.047 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.110 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9654 +t=250: Selected seed 195 with value = 0.9654 +Query 1/1: Action query time = 0.984 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=266: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 0.975 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9692 +t=282: Selected seed 195 with value = 0.9692 +Query 1/1: Action query time = 0.997 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8648 +t=298: Selected seed 195 with value = 0.8648 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=16--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=16--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 16 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 17... +Query 1/1: Action query time = 0.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4783 +t=10: Selected seed 195 with value = 0.4783 +Query 1/1: Action query time = 0.992 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5563 +t=26: Selected seed 195 with value = 0.5563 +Query 1/1: Action query time = 0.964 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6432 +t=42: Selected seed 195 with value = 0.6432 +Query 1/1: Action query time = 0.974 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7759 +t=58: Selected seed 195 with value = 0.7759 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8103 +t=74: Selected seed 195 with value = 0.8103 +Query 1/1: Action query time = 0.988 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9020 +t=90: Selected seed 195 with value = 0.9020 +Query 1/1: Action query time = 0.964 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9941 +t=106: Selected seed 195 with value = 0.9941 +Query 1/1: Action query time = 0.978 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=122: Selected seed 195 with value = 0.9982 +Query 1/1: Action query time = 0.971 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.066 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.111 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.121 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.195 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.148 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=218: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 1.045 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9923 +t=234: Selected seed 195 with value = 0.9923 +Query 1/1: Action query time = 0.968 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9772 +t=250: Selected seed 195 with value = 0.9772 +Query 1/1: Action query time = 0.976 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9362 +t=266: Selected seed 195 with value = 0.9362 +Query 1/1: Action query time = 0.966 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8442 +t=282: Selected seed 195 with value = 0.8442 +Query 1/1: Action query time = 0.974 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7614 +t=298: Selected seed 195 with value = 0.7614 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=17--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=17--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 17 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 18... +Query 1/1: Action query time = 0.999 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4928 +t=10: Selected seed 195 with value = 0.4928 +Query 1/1: Action query time = 0.954 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5934 +t=26: Selected seed 195 with value = 0.5934 +Query 1/1: Action query time = 0.977 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6160 +t=42: Selected seed 195 with value = 0.6160 +Query 1/1: Action query time = 0.952 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7493 +t=58: Selected seed 195 with value = 0.7493 +Query 1/1: Action query time = 0.970 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8882 +t=74: Selected seed 195 with value = 0.8882 +Query 1/1: Action query time = 0.963 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9904 +t=90: Selected seed 195 with value = 0.9904 +Query 1/1: Action query time = 0.986 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.954 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9853 +t=218: Selected seed 195 with value = 0.9853 +Query 1/1: Action query time = 0.967 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9290 +t=234: Selected seed 195 with value = 0.9290 +Query 1/1: Action query time = 0.975 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9165 +t=250: Selected seed 195 with value = 0.9165 +Query 1/1: Action query time = 0.991 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.959 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=18--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=18--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 18 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 19... +Query 1/1: Action query time = 1.000 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4884 +t=10: Selected seed 195 with value = 0.4884 +Query 1/1: Action query time = 0.972 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5843 +t=26: Selected seed 195 with value = 0.5843 +Query 1/1: Action query time = 0.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6113 +t=42: Selected seed 195 with value = 0.6113 +Query 1/1: Action query time = 0.974 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7538 +t=58: Selected seed 195 with value = 0.7538 +Query 1/1: Action query time = 0.982 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8943 +t=74: Selected seed 195 with value = 0.8943 +Query 1/1: Action query time = 0.992 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=90: Selected seed 195 with value = 0.9870 +Query 1/1: Action query time = 0.977 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.955 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.961 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.959 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9970 +t=186: Selected seed 195 with value = 0.9970 +Query 1/1: Action query time = 0.993 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9632 +t=202: Selected seed 195 with value = 0.9632 +Query 1/1: Action query time = 0.967 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9768 +t=218: Selected seed 195 with value = 0.9768 +Query 1/1: Action query time = 0.969 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8565 +t=234: Selected seed 195 with value = 0.8565 +Query 1/1: Action query time = 0.984 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8377 +t=250: Selected seed 195 with value = 0.8377 +Query 1/1: Action query time = 0.969 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8779 +t=266: Selected seed 195 with value = 0.8779 +Query 1/1: Action query time = 0.975 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9661 +t=282: Selected seed 195 with value = 0.9661 +Query 1/1: Action query time = 0.963 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=19--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=19--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 19 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 20... +Query 1/1: Action query time = 0.977 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4669 +t=10: Selected seed 195 with value = 0.4669 +Query 1/1: Action query time = 0.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5683 +t=26: Selected seed 195 with value = 0.5683 +Query 1/1: Action query time = 0.969 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6080 +t=42: Selected seed 195 with value = 0.6080 +Query 1/1: Action query time = 0.982 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7421 +t=58: Selected seed 195 with value = 0.7421 +Query 1/1: Action query time = 0.990 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8763 +t=74: Selected seed 195 with value = 0.8763 +Query 1/1: Action query time = 0.959 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=90: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 0.970 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.970 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.960 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=266: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 0.953 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=298: Selected seed 195 with value = 0.9993 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=20--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=20--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 20 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 21... +Query 1/1: Action query time = 0.982 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4613 +t=10: Selected seed 195 with value = 0.4613 +Query 1/1: Action query time = 0.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5518 +t=26: Selected seed 195 with value = 0.5518 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6220 +t=42: Selected seed 195 with value = 0.6220 +Query 1/1: Action query time = 0.986 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7776 +t=58: Selected seed 195 with value = 0.7776 +Query 1/1: Action query time = 0.961 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8879 +t=74: Selected seed 195 with value = 0.8879 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9167 +t=90: Selected seed 195 with value = 0.9167 +Query 1/1: Action query time = 0.987 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8940 +t=106: Selected seed 195 with value = 0.8940 +Query 1/1: Action query time = 0.971 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8923 +t=122: Selected seed 195 with value = 0.8923 +Query 1/1: Action query time = 0.967 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9258 +t=138: Selected seed 195 with value = 0.9258 +Query 1/1: Action query time = 0.975 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=170: Selected seed 195 with value = 0.9867 +Query 1/1: Action query time = 0.968 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.001 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=218: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 0.972 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9902 +t=234: Selected seed 195 with value = 0.9902 +Query 1/1: Action query time = 0.976 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9745 +t=250: Selected seed 195 with value = 0.9745 +Query 1/1: Action query time = 0.975 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9464 +t=266: Selected seed 195 with value = 0.9464 +Query 1/1: Action query time = 0.965 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9018 +t=282: Selected seed 195 with value = 0.9018 +Query 1/1: Action query time = 0.963 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8386 +t=298: Selected seed 195 with value = 0.8386 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=21--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=21--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 21 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 22... +Query 1/1: Action query time = 0.980 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4894 +t=10: Selected seed 195 with value = 0.4894 +Query 1/1: Action query time = 0.972 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4807 +t=26: Selected seed 195 with value = 0.4807 +Query 1/1: Action query time = 0.983 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6004 +t=42: Selected seed 195 with value = 0.6004 +Query 1/1: Action query time = 0.968 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7408 +t=58: Selected seed 195 with value = 0.7408 +Query 1/1: Action query time = 0.970 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8671 +t=74: Selected seed 195 with value = 0.8671 +Query 1/1: Action query time = 0.979 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=90: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 0.965 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.984 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.961 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=22--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=22--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 22 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 23... +Query 1/1: Action query time = 0.991 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4965 +t=10: Selected seed 195 with value = 0.4965 +Query 1/1: Action query time = 0.969 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5847 +t=26: Selected seed 195 with value = 0.5847 +Query 1/1: Action query time = 0.979 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6117 +t=42: Selected seed 195 with value = 0.6117 +Query 1/1: Action query time = 0.977 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7506 +t=58: Selected seed 195 with value = 0.7506 +Query 1/1: Action query time = 0.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8885 +t=74: Selected seed 195 with value = 0.8885 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9880 +t=90: Selected seed 195 with value = 0.9880 +Query 1/1: Action query time = 0.965 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.970 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.989 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=298: Selected seed 195 with value = 0.9985 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=23--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=23--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 23 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 24... +Query 1/1: Action query time = 0.974 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4796 +t=10: Selected seed 195 with value = 0.4796 +Query 1/1: Action query time = 0.987 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5543 +t=26: Selected seed 195 with value = 0.5543 +Query 1/1: Action query time = 0.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6433 +t=42: Selected seed 195 with value = 0.6433 +Query 1/1: Action query time = 1.036 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7387 +t=58: Selected seed 195 with value = 0.7387 +Query 1/1: Action query time = 1.006 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8001 +t=74: Selected seed 195 with value = 0.8001 +Query 1/1: Action query time = 1.004 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8992 +t=90: Selected seed 195 with value = 0.8992 +Query 1/1: Action query time = 1.001 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=106: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 0.996 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=122: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 0.973 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.010 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=234: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 0.950 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9911 +t=250: Selected seed 195 with value = 0.9911 +Query 1/1: Action query time = 0.959 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9746 +t=266: Selected seed 195 with value = 0.9746 +Query 1/1: Action query time = 0.975 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9124 +t=282: Selected seed 195 with value = 0.9124 +Query 1/1: Action query time = 0.967 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8211 +t=298: Selected seed 195 with value = 0.8211 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=24--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=24--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 24 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 25... +Query 1/1: Action query time = 0.996 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4597 +t=10: Selected seed 195 with value = 0.4597 +Query 1/1: Action query time = 0.964 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5499 +t=26: Selected seed 195 with value = 0.5499 +Query 1/1: Action query time = 0.969 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6062 +t=42: Selected seed 195 with value = 0.6062 +Query 1/1: Action query time = 0.976 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7211 +t=58: Selected seed 195 with value = 0.7211 +Query 1/1: Action query time = 0.964 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8321 +t=74: Selected seed 195 with value = 0.8321 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9295 +t=90: Selected seed 195 with value = 0.9295 +Query 1/1: Action query time = 0.967 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9941 +t=106: Selected seed 195 with value = 0.9941 +Query 1/1: Action query time = 0.981 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9952 +t=122: Selected seed 195 with value = 0.9952 +Query 1/1: Action query time = 0.969 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=138: Selected seed 195 with value = 0.9955 +Query 1/1: Action query time = 0.971 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=154: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 1.036 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=170: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 0.964 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9890 +t=186: Selected seed 195 with value = 0.9890 +Query 1/1: Action query time = 0.981 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9771 +t=202: Selected seed 195 with value = 0.9771 +Query 1/1: Action query time = 0.980 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9554 +t=218: Selected seed 195 with value = 0.9554 +Query 1/1: Action query time = 0.969 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9159 +t=234: Selected seed 195 with value = 0.9159 +Query 1/1: Action query time = 0.979 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8593 +t=250: Selected seed 195 with value = 0.8593 +Query 1/1: Action query time = 0.979 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8330 +t=266: Selected seed 195 with value = 0.8330 +Query 1/1: Action query time = 0.960 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8197 +t=282: Selected seed 195 with value = 0.8197 +Query 1/1: Action query time = 0.972 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8127 +t=298: Selected seed 195 with value = 0.8127 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=25--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=25--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 25 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 26... +Query 1/1: Action query time = 0.983 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4894 +t=10: Selected seed 195 with value = 0.4894 +Query 1/1: Action query time = 0.967 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5502 +t=26: Selected seed 195 with value = 0.5502 +Query 1/1: Action query time = 0.971 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5960 +t=42: Selected seed 195 with value = 0.5960 +Query 1/1: Action query time = 0.964 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7513 +t=58: Selected seed 195 with value = 0.7513 +Query 1/1: Action query time = 0.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8903 +t=74: Selected seed 195 with value = 0.8903 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.957 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=250: Selected seed 195 with value = 0.9886 +Query 1/1: Action query time = 0.969 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9358 +t=266: Selected seed 195 with value = 0.9358 +Query 1/1: Action query time = 0.960 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8655 +t=282: Selected seed 195 with value = 0.8655 +Query 1/1: Action query time = 0.970 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9430 +t=298: Selected seed 195 with value = 0.9430 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=26--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=26--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 26 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 27... +Query 1/1: Action query time = 0.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5008 +t=10: Selected seed 195 with value = 0.5008 +Query 1/1: Action query time = 0.969 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5993 +t=26: Selected seed 195 with value = 0.5993 +Query 1/1: Action query time = 0.966 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6381 +t=42: Selected seed 195 with value = 0.6381 +Query 1/1: Action query time = 0.977 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6394 +t=58: Selected seed 195 with value = 0.6394 +Query 1/1: Action query time = 0.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7496 +t=74: Selected seed 195 with value = 0.7496 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7798 +t=90: Selected seed 195 with value = 0.7798 +Query 1/1: Action query time = 0.984 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8375 +t=106: Selected seed 195 with value = 0.8375 +Query 1/1: Action query time = 1.037 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9251 +t=122: Selected seed 195 with value = 0.9251 +Query 1/1: Action query time = 1.040 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9995 +t=138: Selected seed 195 with value = 0.9995 +Query 1/1: Action query time = 1.028 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9893 +t=154: Selected seed 195 with value = 0.9893 +Query 1/1: Action query time = 1.029 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9904 +t=170: Selected seed 195 with value = 0.9904 +Query 1/1: Action query time = 1.021 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9801 +t=186: Selected seed 195 with value = 0.9801 +Query 1/1: Action query time = 1.031 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9388 +t=202: Selected seed 195 with value = 0.9388 +Query 1/1: Action query time = 1.040 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8838 +t=218: Selected seed 195 with value = 0.8838 +Query 1/1: Action query time = 1.047 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8269 +t=234: Selected seed 195 with value = 0.8269 +Query 1/1: Action query time = 1.076 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8001 +t=250: Selected seed 195 with value = 0.8001 +Query 1/1: Action query time = 1.098 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8011 +t=266: Selected seed 195 with value = 0.8011 +Query 1/1: Action query time = 1.114 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8237 +t=282: Selected seed 195 with value = 0.8237 +Query 1/1: Action query time = 0.969 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8388 +t=298: Selected seed 195 with value = 0.8388 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=27--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=27--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 27 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 28... +Query 1/1: Action query time = 0.972 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4673 +t=10: Selected seed 195 with value = 0.4673 +Query 1/1: Action query time = 0.983 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5836 +t=26: Selected seed 195 with value = 0.5836 +Query 1/1: Action query time = 0.986 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6419 +t=42: Selected seed 195 with value = 0.6419 +Query 1/1: Action query time = 1.060 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6856 +t=58: Selected seed 195 with value = 0.6856 +Query 1/1: Action query time = 1.049 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8226 +t=74: Selected seed 195 with value = 0.8226 +Query 1/1: Action query time = 1.052 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8966 +t=90: Selected seed 195 with value = 0.8966 +Query 1/1: Action query time = 1.062 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9904 +t=106: Selected seed 195 with value = 0.9904 +Query 1/1: Action query time = 1.105 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9494 +t=122: Selected seed 195 with value = 0.9494 +Query 1/1: Action query time = 1.115 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9817 +t=138: Selected seed 195 with value = 0.9817 +Query 1/1: Action query time = 1.132 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9833 +t=154: Selected seed 195 with value = 0.9833 +Query 1/1: Action query time = 1.100 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=170: Selected seed 195 with value = 0.9864 +Query 1/1: Action query time = 1.100 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9835 +t=186: Selected seed 195 with value = 0.9835 +Query 1/1: Action query time = 1.114 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9752 +t=202: Selected seed 195 with value = 0.9752 +Query 1/1: Action query time = 1.123 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9611 +t=218: Selected seed 195 with value = 0.9611 +Query 1/1: Action query time = 1.107 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9386 +t=234: Selected seed 195 with value = 0.9386 +Query 1/1: Action query time = 0.978 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9161 +t=250: Selected seed 195 with value = 0.9161 +Query 1/1: Action query time = 0.975 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8999 +t=266: Selected seed 195 with value = 0.8999 +Query 1/1: Action query time = 0.965 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8848 +t=282: Selected seed 195 with value = 0.8848 +Query 1/1: Action query time = 0.971 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8682 +t=298: Selected seed 195 with value = 0.8682 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=28--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=28--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 28 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 29... +Query 1/1: Action query time = 1.145 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4953 +t=10: Selected seed 195 with value = 0.4953 +Query 1/1: Action query time = 1.130 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5619 +t=26: Selected seed 195 with value = 0.5619 +Query 1/1: Action query time = 1.155 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6279 +t=42: Selected seed 195 with value = 0.6279 +Query 1/1: Action query time = 1.101 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6942 +t=58: Selected seed 195 with value = 0.6942 +Query 1/1: Action query time = 1.077 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8212 +t=74: Selected seed 195 with value = 0.8212 +Query 1/1: Action query time = 1.099 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9070 +t=90: Selected seed 195 with value = 0.9070 +Query 1/1: Action query time = 1.074 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=106: Selected seed 195 with value = 0.9914 +Query 1/1: Action query time = 1.100 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=122: Selected seed 195 with value = 0.9864 +Query 1/1: Action query time = 1.146 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9906 +t=138: Selected seed 195 with value = 0.9906 +Query 1/1: Action query time = 1.182 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=154: Selected seed 195 with value = 0.9926 +Query 1/1: Action query time = 1.168 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9892 +t=170: Selected seed 195 with value = 0.9892 +Query 1/1: Action query time = 1.151 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9816 +t=186: Selected seed 195 with value = 0.9816 +Query 1/1: Action query time = 1.154 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9655 +t=202: Selected seed 195 with value = 0.9655 +Query 1/1: Action query time = 1.143 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9152 +t=218: Selected seed 195 with value = 0.9152 +Query 1/1: Action query time = 1.126 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8795 +t=234: Selected seed 195 with value = 0.8795 +Query 1/1: Action query time = 0.978 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8309 +t=250: Selected seed 195 with value = 0.8309 +Query 1/1: Action query time = 0.978 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8220 +t=266: Selected seed 195 with value = 0.8220 +Query 1/1: Action query time = 0.965 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8313 +t=282: Selected seed 195 with value = 0.8313 +Query 1/1: Action query time = 1.128 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8430 +t=298: Selected seed 195 with value = 0.8430 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=29--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=29--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 29 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 30... +Query 1/1: Action query time = 1.108 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4927 +t=10: Selected seed 195 with value = 0.4927 +Query 1/1: Action query time = 1.084 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5746 +t=26: Selected seed 195 with value = 0.5746 +Query 1/1: Action query time = 1.038 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6514 +t=42: Selected seed 195 with value = 0.6514 +Query 1/1: Action query time = 1.108 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6191 +t=58: Selected seed 195 with value = 0.6191 +Query 1/1: Action query time = 1.102 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8035 +t=74: Selected seed 195 with value = 0.8035 +Query 1/1: Action query time = 1.123 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7987 +t=90: Selected seed 195 with value = 0.7987 +Query 1/1: Action query time = 1.111 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9226 +t=106: Selected seed 195 with value = 0.9226 +Query 1/1: Action query time = 1.118 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.987 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.984 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.195 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9996 +t=202: Selected seed 195 with value = 0.9996 +Query 1/1: Action query time = 1.208 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9951 +t=218: Selected seed 195 with value = 0.9951 +Query 1/1: Action query time = 1.187 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9876 +t=234: Selected seed 195 with value = 0.9876 +Query 1/1: Action query time = 1.212 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9695 +t=250: Selected seed 195 with value = 0.9695 +Query 1/1: Action query time = 1.165 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9268 +t=266: Selected seed 195 with value = 0.9268 +Query 1/1: Action query time = 1.203 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8176 +t=282: Selected seed 195 with value = 0.8176 +Query 1/1: Action query time = 1.222 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7749 +t=298: Selected seed 195 with value = 0.7749 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=30--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=30--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 30 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 31... +Query 1/1: Action query time = 0.991 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4939 +t=10: Selected seed 195 with value = 0.4939 +Query 1/1: Action query time = 0.967 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5554 +t=26: Selected seed 195 with value = 0.5554 +Query 1/1: Action query time = 0.998 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6234 +t=42: Selected seed 195 with value = 0.6234 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7325 +t=58: Selected seed 195 with value = 0.7325 +Query 1/1: Action query time = 0.973 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8358 +t=74: Selected seed 195 with value = 0.8358 +Query 1/1: Action query time = 0.973 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9566 +t=90: Selected seed 195 with value = 0.9566 +Query 1/1: Action query time = 0.978 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.013 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.035 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.026 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.064 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.095 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.181 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.011 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.981 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=31--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=31--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 31 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 32... +Query 1/1: Action query time = 1.136 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4811 +t=10: Selected seed 195 with value = 0.4811 +Query 1/1: Action query time = 1.085 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5813 +t=26: Selected seed 195 with value = 0.5813 +Query 1/1: Action query time = 1.047 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6210 +t=42: Selected seed 195 with value = 0.6210 +Query 1/1: Action query time = 1.049 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7644 +t=58: Selected seed 195 with value = 0.7644 +Query 1/1: Action query time = 0.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9049 +t=74: Selected seed 195 with value = 0.9049 +Query 1/1: Action query time = 0.963 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=90: Selected seed 195 with value = 0.9999 +Query 1/1: Action query time = 0.974 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.189 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.210 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.181 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.185 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.156 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=202: Selected seed 195 with value = 0.9924 +Query 1/1: Action query time = 1.159 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9792 +t=218: Selected seed 195 with value = 0.9792 +Query 1/1: Action query time = 1.133 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=234: Selected seed 195 with value = 0.9871 +Query 1/1: Action query time = 1.155 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9602 +t=250: Selected seed 195 with value = 0.9602 +Query 1/1: Action query time = 1.160 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9674 +t=266: Selected seed 195 with value = 0.9674 +Query 1/1: Action query time = 1.201 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9714 +t=282: Selected seed 195 with value = 0.9714 +Query 1/1: Action query time = 1.186 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9364 +t=298: Selected seed 195 with value = 0.9364 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=32--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=32--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 32 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 33... +Query 1/1: Action query time = 0.989 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5087 +t=10: Selected seed 195 with value = 0.5087 +Query 1/1: Action query time = 0.974 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5451 +t=26: Selected seed 195 with value = 0.5451 +Query 1/1: Action query time = 0.994 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6222 +t=42: Selected seed 195 with value = 0.6222 +Query 1/1: Action query time = 0.962 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7199 +t=58: Selected seed 195 with value = 0.7199 +Query 1/1: Action query time = 0.970 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8104 +t=74: Selected seed 195 with value = 0.8104 +Query 1/1: Action query time = 0.966 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8796 +t=90: Selected seed 195 with value = 0.8796 +Query 1/1: Action query time = 0.979 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9733 +t=106: Selected seed 195 with value = 0.9733 +Query 1/1: Action query time = 0.970 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9573 +t=122: Selected seed 195 with value = 0.9573 +Query 1/1: Action query time = 0.981 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9932 +t=138: Selected seed 195 with value = 0.9932 +Query 1/1: Action query time = 0.982 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=154: Selected seed 195 with value = 0.9914 +Query 1/1: Action query time = 0.973 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9856 +t=170: Selected seed 195 with value = 0.9856 +Query 1/1: Action query time = 0.966 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9768 +t=186: Selected seed 195 with value = 0.9768 +Query 1/1: Action query time = 0.969 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9600 +t=202: Selected seed 195 with value = 0.9600 +Query 1/1: Action query time = 0.971 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9260 +t=218: Selected seed 195 with value = 0.9260 +Query 1/1: Action query time = 0.987 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8944 +t=234: Selected seed 195 with value = 0.8944 +Query 1/1: Action query time = 0.982 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8650 +t=250: Selected seed 195 with value = 0.8650 +Query 1/1: Action query time = 0.969 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8507 +t=266: Selected seed 195 with value = 0.8507 +Query 1/1: Action query time = 0.983 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8461 +t=282: Selected seed 195 with value = 0.8461 +Query 1/1: Action query time = 0.974 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8289 +t=298: Selected seed 195 with value = 0.8289 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=33--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=33--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 33 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 34... +Query 1/1: Action query time = 0.983 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4681 +t=10: Selected seed 195 with value = 0.4681 +Query 1/1: Action query time = 1.002 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5511 +t=26: Selected seed 195 with value = 0.5511 +Query 1/1: Action query time = 0.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6088 +t=42: Selected seed 195 with value = 0.6088 +Query 1/1: Action query time = 0.986 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7243 +t=58: Selected seed 195 with value = 0.7243 +Query 1/1: Action query time = 0.969 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8648 +t=74: Selected seed 195 with value = 0.8648 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9921 +t=90: Selected seed 195 with value = 0.9921 +Query 1/1: Action query time = 0.980 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.960 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=138: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 0.983 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=154: Selected seed 195 with value = 0.9983 +Query 1/1: Action query time = 0.969 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=170: Selected seed 195 with value = 0.9976 +Query 1/1: Action query time = 0.968 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=186: Selected seed 195 with value = 0.9963 +Query 1/1: Action query time = 0.969 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=202: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 0.968 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9929 +t=218: Selected seed 195 with value = 0.9929 +Query 1/1: Action query time = 0.957 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9927 +t=234: Selected seed 195 with value = 0.9927 +Query 1/1: Action query time = 0.961 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9941 +t=250: Selected seed 195 with value = 0.9941 +Query 1/1: Action query time = 0.960 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9951 +t=266: Selected seed 195 with value = 0.9951 +Query 1/1: Action query time = 0.975 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9841 +t=282: Selected seed 195 with value = 0.9841 +Query 1/1: Action query time = 0.980 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8979 +t=298: Selected seed 195 with value = 0.8979 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=34--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=34--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 34 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 35... +Query 1/1: Action query time = 0.980 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4901 +t=10: Selected seed 195 with value = 0.4901 +Query 1/1: Action query time = 0.966 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5310 +t=26: Selected seed 195 with value = 0.5310 +Query 1/1: Action query time = 0.970 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6070 +t=42: Selected seed 195 with value = 0.6070 +Query 1/1: Action query time = 0.971 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6847 +t=58: Selected seed 195 with value = 0.6847 +Query 1/1: Action query time = 0.988 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7516 +t=74: Selected seed 195 with value = 0.7516 +Query 1/1: Action query time = 1.019 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8223 +t=90: Selected seed 195 with value = 0.8223 +Query 1/1: Action query time = 1.029 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8666 +t=106: Selected seed 195 with value = 0.8666 +Query 1/1: Action query time = 1.081 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9145 +t=122: Selected seed 195 with value = 0.9145 +Query 1/1: Action query time = 1.104 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9844 +t=138: Selected seed 195 with value = 0.9844 +Query 1/1: Action query time = 1.107 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=154: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 1.124 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=170: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 1.098 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.054 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=202: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 1.083 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9882 +t=218: Selected seed 195 with value = 0.9882 +Query 1/1: Action query time = 1.103 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9786 +t=234: Selected seed 195 with value = 0.9786 +Query 1/1: Action query time = 1.135 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9474 +t=250: Selected seed 195 with value = 0.9474 +Query 1/1: Action query time = 1.154 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8772 +t=266: Selected seed 195 with value = 0.8772 +Query 1/1: Action query time = 0.976 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8107 +t=282: Selected seed 195 with value = 0.8107 +Query 1/1: Action query time = 0.968 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7759 +t=298: Selected seed 195 with value = 0.7759 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=35--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=35--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 35 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 36... +Query 1/1: Action query time = 0.992 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4801 +t=10: Selected seed 195 with value = 0.4801 +Query 1/1: Action query time = 0.980 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5703 +t=26: Selected seed 195 with value = 0.5703 +Query 1/1: Action query time = 0.985 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5928 +t=42: Selected seed 195 with value = 0.5928 +Query 1/1: Action query time = 0.973 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7157 +t=58: Selected seed 195 with value = 0.7157 +Query 1/1: Action query time = 0.974 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8452 +t=74: Selected seed 195 with value = 0.8452 +Query 1/1: Action query time = 0.983 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9657 +t=90: Selected seed 195 with value = 0.9657 +Query 1/1: Action query time = 0.981 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=154: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 0.980 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9837 +t=170: Selected seed 195 with value = 0.9837 +Query 1/1: Action query time = 0.979 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9851 +t=186: Selected seed 195 with value = 0.9851 +Query 1/1: Action query time = 0.971 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9929 +t=202: Selected seed 195 with value = 0.9929 +Query 1/1: Action query time = 0.992 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9841 +t=218: Selected seed 195 with value = 0.9841 +Query 1/1: Action query time = 0.971 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9190 +t=234: Selected seed 195 with value = 0.9190 +Query 1/1: Action query time = 0.963 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8712 +t=250: Selected seed 195 with value = 0.8712 +Query 1/1: Action query time = 0.968 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8935 +t=266: Selected seed 195 with value = 0.8935 +Query 1/1: Action query time = 0.995 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9800 +t=282: Selected seed 195 with value = 0.9800 +Query 1/1: Action query time = 0.966 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=298: Selected seed 195 with value = 0.9870 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=36--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=36--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 36 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 37... +Query 1/1: Action query time = 1.036 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4648 +t=10: Selected seed 195 with value = 0.4648 +Query 1/1: Action query time = 1.013 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5836 +t=26: Selected seed 195 with value = 0.5836 +Query 1/1: Action query time = 1.048 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6238 +t=42: Selected seed 195 with value = 0.6238 +Query 1/1: Action query time = 1.075 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6307 +t=58: Selected seed 195 with value = 0.6307 +Query 1/1: Action query time = 1.118 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7937 +t=74: Selected seed 195 with value = 0.7937 +Query 1/1: Action query time = 1.115 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8085 +t=90: Selected seed 195 with value = 0.8085 +Query 1/1: Action query time = 1.085 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9084 +t=106: Selected seed 195 with value = 0.9084 +Query 1/1: Action query time = 0.983 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9941 +t=122: Selected seed 195 with value = 0.9941 +Query 1/1: Action query time = 0.974 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9900 +t=138: Selected seed 195 with value = 0.9900 +Query 1/1: Action query time = 0.973 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9915 +t=154: Selected seed 195 with value = 0.9915 +Query 1/1: Action query time = 0.960 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9811 +t=170: Selected seed 195 with value = 0.9811 +Query 1/1: Action query time = 0.961 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9692 +t=186: Selected seed 195 with value = 0.9692 +Query 1/1: Action query time = 0.973 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9485 +t=202: Selected seed 195 with value = 0.9485 +Query 1/1: Action query time = 0.960 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9203 +t=218: Selected seed 195 with value = 0.9203 +Query 1/1: Action query time = 0.968 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8945 +t=234: Selected seed 195 with value = 0.8945 +Query 1/1: Action query time = 0.981 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8685 +t=250: Selected seed 195 with value = 0.8685 +Query 1/1: Action query time = 0.982 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8483 +t=266: Selected seed 195 with value = 0.8483 +Query 1/1: Action query time = 0.959 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8179 +t=282: Selected seed 195 with value = 0.8179 +Query 1/1: Action query time = 0.964 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9022 +t=298: Selected seed 195 with value = 0.9022 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=37--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=37--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 37 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 38... +Query 1/1: Action query time = 0.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4872 +t=10: Selected seed 195 with value = 0.4872 +Query 1/1: Action query time = 0.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5868 +t=26: Selected seed 195 with value = 0.5868 +Query 1/1: Action query time = 0.961 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6435 +t=42: Selected seed 195 with value = 0.6435 +Query 1/1: Action query time = 0.977 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6507 +t=58: Selected seed 195 with value = 0.6507 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7907 +t=74: Selected seed 195 with value = 0.7907 +Query 1/1: Action query time = 0.982 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8349 +t=90: Selected seed 195 with value = 0.8349 +Query 1/1: Action query time = 0.989 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9526 +t=106: Selected seed 195 with value = 0.9526 +Query 1/1: Action query time = 0.984 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=122: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 0.984 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.994 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.992 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9975 +t=170: Selected seed 195 with value = 0.9975 +Query 1/1: Action query time = 0.988 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9894 +t=186: Selected seed 195 with value = 0.9894 +Query 1/1: Action query time = 0.991 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9789 +t=202: Selected seed 195 with value = 0.9789 +Query 1/1: Action query time = 0.977 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9564 +t=218: Selected seed 195 with value = 0.9564 +Query 1/1: Action query time = 1.011 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8986 +t=234: Selected seed 195 with value = 0.8986 +Query 1/1: Action query time = 0.979 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8211 +t=250: Selected seed 195 with value = 0.8211 +Query 1/1: Action query time = 0.990 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7874 +t=266: Selected seed 195 with value = 0.7874 +Query 1/1: Action query time = 0.969 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7842 +t=282: Selected seed 195 with value = 0.7842 +Query 1/1: Action query time = 0.984 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7869 +t=298: Selected seed 195 with value = 0.7869 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=38--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=38--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 38 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 39... +Query 1/1: Action query time = 1.168 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4806 +t=10: Selected seed 195 with value = 0.4806 +Query 1/1: Action query time = 1.148 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4847 +t=26: Selected seed 195 with value = 0.4847 +Query 1/1: Action query time = 1.128 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6016 +t=42: Selected seed 195 with value = 0.6016 +Query 1/1: Action query time = 1.137 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7153 +t=58: Selected seed 195 with value = 0.7153 +Query 1/1: Action query time = 1.107 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8552 +t=74: Selected seed 195 with value = 0.8552 +Query 1/1: Action query time = 1.126 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9758 +t=90: Selected seed 195 with value = 0.9758 +Query 1/1: Action query time = 1.108 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.146 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.118 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.178 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.151 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.142 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.223 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.202 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.201 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=266: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 0.977 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9534 +t=282: Selected seed 195 with value = 0.9534 +Query 1/1: Action query time = 0.975 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9213 +t=298: Selected seed 195 with value = 0.9213 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=39--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=39--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 39 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 40... +Query 1/1: Action query time = 1.152 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4965 +t=10: Selected seed 195 with value = 0.4965 +Query 1/1: Action query time = 1.215 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5436 +t=26: Selected seed 195 with value = 0.5436 +Query 1/1: Action query time = 1.196 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6198 +t=42: Selected seed 195 with value = 0.6198 +Query 1/1: Action query time = 1.208 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7870 +t=58: Selected seed 195 with value = 0.7870 +Query 1/1: Action query time = 1.212 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8373 +t=74: Selected seed 195 with value = 0.8373 +Query 1/1: Action query time = 1.171 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8908 +t=90: Selected seed 195 with value = 0.8908 +Query 1/1: Action query time = 1.174 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9768 +t=106: Selected seed 195 with value = 0.9768 +Query 1/1: Action query time = 1.099 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.051 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=138: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 0.987 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.987 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=170: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 0.972 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=186: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 0.979 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=202: Selected seed 195 with value = 0.9963 +Query 1/1: Action query time = 0.974 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9896 +t=218: Selected seed 195 with value = 0.9896 +Query 1/1: Action query time = 0.967 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9801 +t=234: Selected seed 195 with value = 0.9801 +Query 1/1: Action query time = 0.963 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9633 +t=250: Selected seed 195 with value = 0.9633 +Query 1/1: Action query time = 0.973 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9153 +t=266: Selected seed 195 with value = 0.9153 +Query 1/1: Action query time = 0.972 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8523 +t=282: Selected seed 195 with value = 0.8523 +Query 1/1: Action query time = 0.959 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8040 +t=298: Selected seed 195 with value = 0.8040 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=40--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=40--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 40 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 41... +Query 1/1: Action query time = 1.007 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4846 +t=10: Selected seed 195 with value = 0.4846 +Query 1/1: Action query time = 0.993 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5304 +t=26: Selected seed 195 with value = 0.5304 +Query 1/1: Action query time = 0.976 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5963 +t=42: Selected seed 195 with value = 0.5963 +Query 1/1: Action query time = 0.984 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7232 +t=58: Selected seed 195 with value = 0.7232 +Query 1/1: Action query time = 0.967 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8448 +t=74: Selected seed 195 with value = 0.8448 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9730 +t=90: Selected seed 195 with value = 0.9730 +Query 1/1: Action query time = 0.974 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.989 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.070 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.053 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.051 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.086 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.101 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.100 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.113 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=41--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=41--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 41 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 42... +Query 1/1: Action query time = 0.993 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4939 +t=10: Selected seed 195 with value = 0.4939 +Query 1/1: Action query time = 0.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5815 +t=26: Selected seed 195 with value = 0.5815 +Query 1/1: Action query time = 0.950 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6162 +t=42: Selected seed 195 with value = 0.6162 +Query 1/1: Action query time = 1.012 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7209 +t=58: Selected seed 195 with value = 0.7209 +Query 1/1: Action query time = 1.076 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8544 +t=74: Selected seed 195 with value = 0.8544 +Query 1/1: Action query time = 1.049 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9634 +t=90: Selected seed 195 with value = 0.9634 +Query 1/1: Action query time = 1.058 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9974 +t=154: Selected seed 195 with value = 0.9974 +Query 1/1: Action query time = 0.983 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9941 +t=170: Selected seed 195 with value = 0.9941 +Query 1/1: Action query time = 0.979 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9919 +t=186: Selected seed 195 with value = 0.9919 +Query 1/1: Action query time = 1.029 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9920 +t=202: Selected seed 195 with value = 0.9920 +Query 1/1: Action query time = 1.060 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9942 +t=218: Selected seed 195 with value = 0.9942 +Query 1/1: Action query time = 1.084 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=234: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 1.105 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=250: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 1.108 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.138 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=282: Selected seed 195 with value = 0.9962 +Query 1/1: Action query time = 1.188 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9392 +t=298: Selected seed 195 with value = 0.9392 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=42--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=42--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 42 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 43... +Query 1/1: Action query time = 0.988 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4658 +t=10: Selected seed 195 with value = 0.4658 +Query 1/1: Action query time = 0.981 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5939 +t=26: Selected seed 195 with value = 0.5939 +Query 1/1: Action query time = 0.996 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6330 +t=42: Selected seed 195 with value = 0.6330 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7379 +t=58: Selected seed 195 with value = 0.7379 +Query 1/1: Action query time = 0.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8377 +t=74: Selected seed 195 with value = 0.8377 +Query 1/1: Action query time = 0.964 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9402 +t=90: Selected seed 195 with value = 0.9402 +Query 1/1: Action query time = 0.959 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9967 +t=106: Selected seed 195 with value = 0.9967 +Query 1/1: Action query time = 0.965 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9970 +t=170: Selected seed 195 with value = 0.9970 +Query 1/1: Action query time = 0.974 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9879 +t=186: Selected seed 195 with value = 0.9879 +Query 1/1: Action query time = 1.112 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9783 +t=202: Selected seed 195 with value = 0.9783 +Query 1/1: Action query time = 1.115 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9594 +t=218: Selected seed 195 with value = 0.9594 +Query 1/1: Action query time = 1.132 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9035 +t=234: Selected seed 195 with value = 0.9035 +Query 1/1: Action query time = 1.283 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8293 +t=250: Selected seed 195 with value = 0.8293 +Query 1/1: Action query time = 1.341 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7914 +t=266: Selected seed 195 with value = 0.7914 +Query 1/1: Action query time = 1.369 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7808 +t=282: Selected seed 195 with value = 0.7808 +Query 1/1: Action query time = 1.351 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7916 +t=298: Selected seed 195 with value = 0.7916 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=43--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=43--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 43 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 44... +Query 1/1: Action query time = 0.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4854 +t=10: Selected seed 195 with value = 0.4854 +Query 1/1: Action query time = 0.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5745 +t=26: Selected seed 195 with value = 0.5745 +Query 1/1: Action query time = 0.984 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6072 +t=42: Selected seed 195 with value = 0.6072 +Query 1/1: Action query time = 0.981 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7459 +t=58: Selected seed 195 with value = 0.7459 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8956 +t=74: Selected seed 195 with value = 0.8956 +Query 1/1: Action query time = 0.969 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.962 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.001 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.042 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.050 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.038 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.108 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=250: Selected seed 195 with value = 0.9958 +Query 1/1: Action query time = 0.990 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9580 +t=266: Selected seed 195 with value = 0.9580 +Query 1/1: Action query time = 0.987 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9813 +t=282: Selected seed 195 with value = 0.9813 +Query 1/1: Action query time = 0.985 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9263 +t=298: Selected seed 195 with value = 0.9263 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=44--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=44--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 44 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 45... +Query 1/1: Action query time = 0.993 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5004 +t=10: Selected seed 195 with value = 0.5004 +Query 1/1: Action query time = 0.966 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5512 +t=26: Selected seed 195 with value = 0.5512 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6284 +t=42: Selected seed 195 with value = 0.6284 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7726 +t=58: Selected seed 195 with value = 0.7726 +Query 1/1: Action query time = 0.977 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8140 +t=74: Selected seed 195 with value = 0.8140 +Query 1/1: Action query time = 0.976 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8831 +t=90: Selected seed 195 with value = 0.8831 +Query 1/1: Action query time = 0.957 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9741 +t=106: Selected seed 195 with value = 0.9741 +Query 1/1: Action query time = 0.971 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.990 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.020 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.097 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=186: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 1.154 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9921 +t=202: Selected seed 195 with value = 0.9921 +Query 1/1: Action query time = 1.209 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9837 +t=218: Selected seed 195 with value = 0.9837 +Query 1/1: Action query time = 1.154 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9561 +t=234: Selected seed 195 with value = 0.9561 +Query 1/1: Action query time = 0.991 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9188 +t=250: Selected seed 195 with value = 0.9188 +Query 1/1: Action query time = 0.973 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8513 +t=266: Selected seed 195 with value = 0.8513 +Query 1/1: Action query time = 0.962 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8017 +t=282: Selected seed 195 with value = 0.8017 +Query 1/1: Action query time = 0.962 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7831 +t=298: Selected seed 195 with value = 0.7831 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=45--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=45--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 45 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 46... +Query 1/1: Action query time = 0.983 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4803 +t=10: Selected seed 195 with value = 0.4803 +Query 1/1: Action query time = 1.004 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5546 +t=26: Selected seed 195 with value = 0.5546 +Query 1/1: Action query time = 0.983 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5998 +t=42: Selected seed 195 with value = 0.5998 +Query 1/1: Action query time = 0.988 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7042 +t=58: Selected seed 195 with value = 0.7042 +Query 1/1: Action query time = 1.021 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8704 +t=74: Selected seed 195 with value = 0.8704 +Query 1/1: Action query time = 1.029 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9688 +t=90: Selected seed 195 with value = 0.9688 +Query 1/1: Action query time = 1.032 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.080 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.104 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.097 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.097 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.067 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=186: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 1.044 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=202: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 1.018 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9628 +t=218: Selected seed 195 with value = 0.9628 +Query 1/1: Action query time = 0.966 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8544 +t=234: Selected seed 195 with value = 0.8544 +Query 1/1: Action query time = 0.976 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8815 +t=250: Selected seed 195 with value = 0.8815 +Query 1/1: Action query time = 0.978 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9507 +t=266: Selected seed 195 with value = 0.9507 +Query 1/1: Action query time = 0.968 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=282: Selected seed 195 with value = 0.9982 +Query 1/1: Action query time = 0.965 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=298: Selected seed 195 with value = 0.9997 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=46--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=46--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 46 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 47... +Query 1/1: Action query time = 1.072 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4987 +t=10: Selected seed 195 with value = 0.4987 +Query 1/1: Action query time = 1.073 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5419 +t=26: Selected seed 195 with value = 0.5419 +Query 1/1: Action query time = 1.084 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6321 +t=42: Selected seed 195 with value = 0.6321 +Query 1/1: Action query time = 1.190 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7018 +t=58: Selected seed 195 with value = 0.7018 +Query 1/1: Action query time = 1.227 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7306 +t=74: Selected seed 195 with value = 0.7306 +Query 1/1: Action query time = 1.199 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8027 +t=90: Selected seed 195 with value = 0.8027 +Query 1/1: Action query time = 1.207 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8902 +t=106: Selected seed 195 with value = 0.8902 +Query 1/1: Action query time = 1.227 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9464 +t=122: Selected seed 195 with value = 0.9464 +Query 1/1: Action query time = 1.155 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9923 +t=138: Selected seed 195 with value = 0.9923 +Query 1/1: Action query time = 1.180 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=154: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 1.193 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=170: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 1.203 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9975 +t=186: Selected seed 195 with value = 0.9975 +Query 1/1: Action query time = 1.171 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9936 +t=202: Selected seed 195 with value = 0.9936 +Query 1/1: Action query time = 1.173 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9888 +t=218: Selected seed 195 with value = 0.9888 +Query 1/1: Action query time = 1.163 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9725 +t=234: Selected seed 195 with value = 0.9725 +Query 1/1: Action query time = 0.980 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9498 +t=250: Selected seed 195 with value = 0.9498 +Query 1/1: Action query time = 0.980 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8826 +t=266: Selected seed 195 with value = 0.8826 +Query 1/1: Action query time = 0.982 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8188 +t=282: Selected seed 195 with value = 0.8188 +Query 1/1: Action query time = 0.982 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7879 +t=298: Selected seed 195 with value = 0.7879 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=47--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=47--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 47 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 48... +Query 1/1: Action query time = 1.230 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4798 +t=10: Selected seed 195 with value = 0.4798 +Query 1/1: Action query time = 1.230 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5713 +t=26: Selected seed 195 with value = 0.5713 +Query 1/1: Action query time = 1.177 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6338 +t=42: Selected seed 195 with value = 0.6338 +Query 1/1: Action query time = 1.224 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5604 +t=58: Selected seed 195 with value = 0.5604 +Query 1/1: Action query time = 1.186 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7816 +t=74: Selected seed 195 with value = 0.7816 +Query 1/1: Action query time = 1.160 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8051 +t=90: Selected seed 195 with value = 0.8051 +Query 1/1: Action query time = 1.170 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8746 +t=106: Selected seed 195 with value = 0.8746 +Query 1/1: Action query time = 1.169 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9604 +t=122: Selected seed 195 with value = 0.9604 +Query 1/1: Action query time = 1.155 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9938 +t=138: Selected seed 195 with value = 0.9938 +Query 1/1: Action query time = 1.103 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=154: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 1.063 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9947 +t=170: Selected seed 195 with value = 0.9947 +Query 1/1: Action query time = 0.978 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9834 +t=186: Selected seed 195 with value = 0.9834 +Query 1/1: Action query time = 0.979 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9549 +t=202: Selected seed 195 with value = 0.9549 +Query 1/1: Action query time = 0.971 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9204 +t=218: Selected seed 195 with value = 0.9204 +Query 1/1: Action query time = 0.978 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8444 +t=234: Selected seed 195 with value = 0.8444 +Query 1/1: Action query time = 0.978 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7858 +t=250: Selected seed 195 with value = 0.7858 +Query 1/1: Action query time = 0.968 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7909 +t=266: Selected seed 195 with value = 0.7909 +Query 1/1: Action query time = 0.971 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7774 +t=282: Selected seed 195 with value = 0.7774 +Query 1/1: Action query time = 0.974 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7803 +t=298: Selected seed 195 with value = 0.7803 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=48--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=48--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 48 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 49... +Query 1/1: Action query time = 0.986 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4697 +t=10: Selected seed 195 with value = 0.4697 +Query 1/1: Action query time = 0.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5704 +t=26: Selected seed 195 with value = 0.5704 +Query 1/1: Action query time = 0.987 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6424 +t=42: Selected seed 195 with value = 0.6424 +Query 1/1: Action query time = 0.968 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7729 +t=58: Selected seed 195 with value = 0.7729 +Query 1/1: Action query time = 0.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8203 +t=74: Selected seed 195 with value = 0.8203 +Query 1/1: Action query time = 0.968 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9030 +t=90: Selected seed 195 with value = 0.9030 +Query 1/1: Action query time = 1.026 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=106: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 1.038 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.109 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.087 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=154: Selected seed 195 with value = 0.9983 +Query 1/1: Action query time = 1.163 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9931 +t=170: Selected seed 195 with value = 0.9931 +Query 1/1: Action query time = 1.187 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9847 +t=186: Selected seed 195 with value = 0.9847 +Query 1/1: Action query time = 1.190 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9662 +t=202: Selected seed 195 with value = 0.9662 +Query 1/1: Action query time = 1.147 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9197 +t=218: Selected seed 195 with value = 0.9197 +Query 1/1: Action query time = 1.130 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8672 +t=234: Selected seed 195 with value = 0.8672 +Query 1/1: Action query time = 1.143 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7998 +t=250: Selected seed 195 with value = 0.7998 +Query 1/1: Action query time = 1.115 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7761 +t=266: Selected seed 195 with value = 0.7761 +Query 1/1: Action query time = 0.965 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9026 +t=282: Selected seed 195 with value = 0.9026 +Query 1/1: Action query time = 0.995 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9465 +t=298: Selected seed 195 with value = 0.9465 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=49--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=49--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 49 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 50... +Query 1/1: Action query time = 1.008 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5139 +t=10: Selected seed 195 with value = 0.5139 +Query 1/1: Action query time = 0.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5663 +t=26: Selected seed 195 with value = 0.5663 +Query 1/1: Action query time = 0.981 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5773 +t=42: Selected seed 195 with value = 0.5773 +Query 1/1: Action query time = 0.961 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6429 +t=58: Selected seed 195 with value = 0.6429 +Query 1/1: Action query time = 0.985 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8173 +t=74: Selected seed 195 with value = 0.8173 +Query 1/1: Action query time = 0.974 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8664 +t=90: Selected seed 195 with value = 0.8664 +Query 1/1: Action query time = 0.980 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9834 +t=106: Selected seed 195 with value = 0.9834 +Query 1/1: Action query time = 0.971 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9539 +t=122: Selected seed 195 with value = 0.9539 +Query 1/1: Action query time = 0.977 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9532 +t=138: Selected seed 195 with value = 0.9532 +Query 1/1: Action query time = 0.976 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9778 +t=154: Selected seed 195 with value = 0.9778 +Query 1/1: Action query time = 0.985 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9750 +t=170: Selected seed 195 with value = 0.9750 +Query 1/1: Action query time = 0.977 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9213 +t=186: Selected seed 195 with value = 0.9213 +Query 1/1: Action query time = 0.977 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8771 +t=202: Selected seed 195 with value = 0.8771 +Query 1/1: Action query time = 0.983 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8969 +t=218: Selected seed 195 with value = 0.8969 +Query 1/1: Action query time = 0.979 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9562 +t=234: Selected seed 195 with value = 0.9562 +Query 1/1: Action query time = 0.974 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9591 +t=250: Selected seed 195 with value = 0.9591 +Query 1/1: Action query time = 0.967 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9630 +t=266: Selected seed 195 with value = 0.9630 +Query 1/1: Action query time = 0.987 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9617 +t=282: Selected seed 195 with value = 0.9617 +Query 1/1: Action query time = 0.966 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9611 +t=298: Selected seed 195 with value = 0.9611 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--episode=50--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLi800_t1/2026_07_31-15_24_36--with_future_img--episode=50--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 50 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 50 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-00_10_26--realcl_i800_t2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-00_10_26--realcl_i800_t2.txt new file mode 100644 index 0000000000000000000000000000000000000000..041d19a1a43f342fb9cf711dcc59539e1438d5cd --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-00_10_26--realcl_i800_t2.txt @@ -0,0 +1,1732 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl_i800_t2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=True, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 1.780 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4880 +t=10: Selected seed 195 with value = 0.4880 +Query 1/1: Action query time = 1.291 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5786 +t=26: Selected seed 195 with value = 0.5786 +Query 1/1: Action query time = 1.400 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6747 +t=42: Selected seed 195 with value = 0.6747 +Query 1/1: Action query time = 1.190 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7538 +t=58: Selected seed 195 with value = 0.7538 +Query 1/1: Action query time = 1.195 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9114 +t=74: Selected seed 195 with value = 0.9114 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 1.502 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4304 +t=10: Selected seed 195 with value = 0.4304 +Query 1/1: Action query time = 1.399 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4195 +t=26: Selected seed 195 with value = 0.4195 +Query 1/1: Action query time = 1.407 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5733 +t=42: Selected seed 195 with value = 0.5733 +Query 1/1: Action query time = 1.284 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6794 +t=58: Selected seed 195 with value = 0.6794 +Query 1/1: Action query time = 1.097 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8041 +t=74: Selected seed 195 with value = 0.8041 +Query 1/1: Action query time = 1.099 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9492 +t=90: Selected seed 195 with value = 0.9492 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 0.997 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4213 +t=10: Selected seed 195 with value = 0.4213 +Query 1/1: Action query time = 0.990 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5009 +t=26: Selected seed 195 with value = 0.5009 +Query 1/1: Action query time = 0.971 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5548 +t=42: Selected seed 195 with value = 0.5548 +Query 1/1: Action query time = 0.969 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7041 +t=58: Selected seed 195 with value = 0.7041 +Query 1/1: Action query time = 1.006 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8823 +t=74: Selected seed 195 with value = 0.8823 +Query 1/1: Action query time = 0.993 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=90: Selected seed 195 with value = 0.9979 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 1.145 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3698 +t=10: Selected seed 195 with value = 0.3698 +Query 1/1: Action query time = 0.997 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4848 +t=26: Selected seed 195 with value = 0.4848 +Query 1/1: Action query time = 1.006 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5991 +t=42: Selected seed 195 with value = 0.5991 +Query 1/1: Action query time = 0.941 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6869 +t=58: Selected seed 195 with value = 0.6869 +Query 1/1: Action query time = 1.143 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8036 +t=74: Selected seed 195 with value = 0.8036 +Query 1/1: Action query time = 1.176 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9422 +t=90: Selected seed 195 with value = 0.9422 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=4--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=4--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 5... +Query 1/1: Action query time = 1.197 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4672 +t=10: Selected seed 195 with value = 0.4672 +Query 1/1: Action query time = 1.131 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5499 +t=26: Selected seed 195 with value = 0.5499 +Query 1/1: Action query time = 1.139 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6241 +t=42: Selected seed 195 with value = 0.6241 +Query 1/1: Action query time = 0.998 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7216 +t=58: Selected seed 195 with value = 0.7216 +Query 1/1: Action query time = 1.038 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8311 +t=74: Selected seed 195 with value = 0.8311 +Query 1/1: Action query time = 0.992 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9716 +t=90: Selected seed 195 with value = 0.9716 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=5--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=5--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 6... +Query 1/1: Action query time = 1.191 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4249 +t=10: Selected seed 195 with value = 0.4249 +Query 1/1: Action query time = 1.020 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5186 +t=26: Selected seed 195 with value = 0.5186 +Query 1/1: Action query time = 1.090 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5795 +t=42: Selected seed 195 with value = 0.5795 +Query 1/1: Action query time = 1.154 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7212 +t=58: Selected seed 195 with value = 0.7212 +Query 1/1: Action query time = 1.197 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8463 +t=74: Selected seed 195 with value = 0.8463 +Query 1/1: Action query time = 0.959 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=90: Selected seed 195 with value = 0.9955 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=6--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=6--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 7... +Query 1/1: Action query time = 1.034 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3152 +t=10: Selected seed 195 with value = 0.3152 +Query 1/1: Action query time = 0.996 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4132 +t=26: Selected seed 195 with value = 0.4132 +Query 1/1: Action query time = 1.142 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5545 +t=42: Selected seed 195 with value = 0.5545 +Query 1/1: Action query time = 1.141 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6523 +t=58: Selected seed 195 with value = 0.6523 +Query 1/1: Action query time = 1.063 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7817 +t=74: Selected seed 195 with value = 0.7817 +Query 1/1: Action query time = 1.175 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9178 +t=90: Selected seed 195 with value = 0.9178 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=7--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=7--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 8... +Query 1/1: Action query time = 0.959 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3861 +t=10: Selected seed 195 with value = 0.3861 +Query 1/1: Action query time = 0.953 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4718 +t=26: Selected seed 195 with value = 0.4718 +Query 1/1: Action query time = 1.150 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5820 +t=42: Selected seed 195 with value = 0.5820 +Query 1/1: Action query time = 1.142 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4908 +t=58: Selected seed 195 with value = 0.4908 +Query 1/1: Action query time = 1.007 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5336 +t=74: Selected seed 195 with value = 0.5336 +Query 1/1: Action query time = 1.019 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6318 +t=90: Selected seed 195 with value = 0.6318 +Query 1/1: Action query time = 1.006 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5442 +t=106: Selected seed 195 with value = 0.5442 +Query 1/1: Action query time = 0.955 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5391 +t=122: Selected seed 195 with value = 0.5391 +Query 1/1: Action query time = 0.952 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9166 +t=138: Selected seed 195 with value = 0.9166 +Query 1/1: Action query time = 0.962 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9525 +t=154: Selected seed 195 with value = 0.9525 +Query 1/1: Action query time = 0.996 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9580 +t=170: Selected seed 195 with value = 0.9580 +Query 1/1: Action query time = 0.957 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9562 +t=186: Selected seed 195 with value = 0.9562 +Query 1/1: Action query time = 0.970 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9464 +t=202: Selected seed 195 with value = 0.9464 +Query 1/1: Action query time = 0.959 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9532 +t=218: Selected seed 195 with value = 0.9532 +Query 1/1: Action query time = 0.956 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9502 +t=234: Selected seed 195 with value = 0.9502 +Query 1/1: Action query time = 1.217 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9226 +t=250: Selected seed 195 with value = 0.9226 +Query 1/1: Action query time = 1.175 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4506 +t=266: Selected seed 195 with value = 0.4506 +Query 1/1: Action query time = 1.179 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8112 +t=282: Selected seed 195 with value = 0.8112 +Query 1/1: Action query time = 1.192 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8014 +t=298: Selected seed 195 with value = 0.8014 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=8--success=False--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=8--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 8 +# successes: 7 (87.5%) + +Task: put the wine bottle on top of the cabinet +Starting episode 9... +Query 1/1: Action query time = 1.052 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3686 +t=10: Selected seed 195 with value = 0.3686 +Query 1/1: Action query time = 1.036 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4550 +t=26: Selected seed 195 with value = 0.4550 +Query 1/1: Action query time = 1.062 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5591 +t=42: Selected seed 195 with value = 0.5591 +Query 1/1: Action query time = 1.120 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6952 +t=58: Selected seed 195 with value = 0.6952 +Query 1/1: Action query time = 1.200 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8312 +t=74: Selected seed 195 with value = 0.8312 +Query 1/1: Action query time = 1.290 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9817 +t=90: Selected seed 195 with value = 0.9817 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=9--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=9--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 9 +# successes: 8 (88.9%) + +Task: put the wine bottle on top of the cabinet +Starting episode 10... +Query 1/1: Action query time = 1.008 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4276 +t=10: Selected seed 195 with value = 0.4276 +Query 1/1: Action query time = 0.955 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5102 +t=26: Selected seed 195 with value = 0.5102 +Query 1/1: Action query time = 1.262 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6156 +t=42: Selected seed 195 with value = 0.6156 +Query 1/1: Action query time = 1.323 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7244 +t=58: Selected seed 195 with value = 0.7244 +Query 1/1: Action query time = 1.149 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8691 +t=74: Selected seed 195 with value = 0.8691 +Query 1/1: Action query time = 1.052 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=10--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=10--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 10 +# successes: 9 (90.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 11... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3969 +t=10: Selected seed 195 with value = 0.3969 +Query 1/1: Action query time = 0.954 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4856 +t=26: Selected seed 195 with value = 0.4856 +Query 1/1: Action query time = 0.970 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5836 +t=42: Selected seed 195 with value = 0.5836 +Query 1/1: Action query time = 0.996 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6719 +t=58: Selected seed 195 with value = 0.6719 +Query 1/1: Action query time = 1.016 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7841 +t=74: Selected seed 195 with value = 0.7841 +Query 1/1: Action query time = 1.164 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9180 +t=90: Selected seed 195 with value = 0.9180 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=11--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=11--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 11 +# successes: 10 (90.9%) + +Task: put the wine bottle on top of the cabinet +Starting episode 12... +Query 1/1: Action query time = 1.081 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4430 +t=10: Selected seed 195 with value = 0.4430 +Query 1/1: Action query time = 1.136 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5502 +t=26: Selected seed 195 with value = 0.5502 +Query 1/1: Action query time = 1.162 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6344 +t=42: Selected seed 195 with value = 0.6344 +Query 1/1: Action query time = 1.202 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7407 +t=58: Selected seed 195 with value = 0.7407 +Query 1/1: Action query time = 1.147 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8831 +t=74: Selected seed 195 with value = 0.8831 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=12--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=12--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 12 +# successes: 11 (91.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 13... +Query 1/1: Action query time = 0.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4677 +t=10: Selected seed 195 with value = 0.4677 +Query 1/1: Action query time = 0.978 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5386 +t=26: Selected seed 195 with value = 0.5386 +Query 1/1: Action query time = 0.992 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6124 +t=42: Selected seed 195 with value = 0.6124 +Query 1/1: Action query time = 0.977 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7155 +t=58: Selected seed 195 with value = 0.7155 +Query 1/1: Action query time = 0.957 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8257 +t=74: Selected seed 195 with value = 0.8257 +Query 1/1: Action query time = 1.131 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9820 +t=90: Selected seed 195 with value = 0.9820 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=13--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=13--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 13 +# successes: 12 (92.3%) + +Task: put the wine bottle on top of the cabinet +Starting episode 14... +Query 1/1: Action query time = 1.009 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4228 +t=10: Selected seed 195 with value = 0.4228 +Query 1/1: Action query time = 0.944 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4921 +t=26: Selected seed 195 with value = 0.4921 +Query 1/1: Action query time = 0.946 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5572 +t=42: Selected seed 195 with value = 0.5572 +Query 1/1: Action query time = 0.971 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6484 +t=58: Selected seed 195 with value = 0.6484 +Query 1/1: Action query time = 1.208 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7321 +t=74: Selected seed 195 with value = 0.7321 +Query 1/1: Action query time = 1.269 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9303 +t=90: Selected seed 195 with value = 0.9303 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=14--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=14--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 14 +# successes: 13 (92.9%) + +Task: put the wine bottle on top of the cabinet +Starting episode 15... +Query 1/1: Action query time = 1.012 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4221 +t=10: Selected seed 195 with value = 0.4221 +Query 1/1: Action query time = 0.967 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5148 +t=26: Selected seed 195 with value = 0.5148 +Query 1/1: Action query time = 0.968 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5940 +t=42: Selected seed 195 with value = 0.5940 +Query 1/1: Action query time = 0.960 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7373 +t=58: Selected seed 195 with value = 0.7373 +Query 1/1: Action query time = 1.073 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8500 +t=74: Selected seed 195 with value = 0.8500 +Query 1/1: Action query time = 1.027 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=15--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=15--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 15 +# successes: 14 (93.3%) + +Task: put the wine bottle on top of the cabinet +Starting episode 16... +Query 1/1: Action query time = 0.957 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4069 +t=10: Selected seed 195 with value = 0.4069 +Query 1/1: Action query time = 0.959 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4456 +t=26: Selected seed 195 with value = 0.4456 +Query 1/1: Action query time = 1.024 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5473 +t=42: Selected seed 195 with value = 0.5473 +Query 1/1: Action query time = 1.032 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6512 +t=58: Selected seed 195 with value = 0.6512 +Query 1/1: Action query time = 1.093 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7311 +t=74: Selected seed 195 with value = 0.7311 +Query 1/1: Action query time = 0.978 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8635 +t=90: Selected seed 195 with value = 0.8635 +Query 1/1: Action query time = 1.062 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=16--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=16--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 16 +# successes: 15 (93.8%) + +Task: put the wine bottle on top of the cabinet +Starting episode 17... +Query 1/1: Action query time = 1.271 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4016 +t=10: Selected seed 195 with value = 0.4016 +Query 1/1: Action query time = 1.215 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4593 +t=26: Selected seed 195 with value = 0.4593 +Query 1/1: Action query time = 1.163 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5621 +t=42: Selected seed 195 with value = 0.5621 +Query 1/1: Action query time = 1.177 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6519 +t=58: Selected seed 195 with value = 0.6519 +Query 1/1: Action query time = 1.129 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8173 +t=74: Selected seed 195 with value = 0.8173 +Query 1/1: Action query time = 1.365 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9569 +t=90: Selected seed 195 with value = 0.9569 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=17--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=17--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 17 +# successes: 16 (94.1%) + +Task: put the wine bottle on top of the cabinet +Starting episode 18... +Query 1/1: Action query time = 1.003 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3606 +t=10: Selected seed 195 with value = 0.3606 +Query 1/1: Action query time = 1.018 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4745 +t=26: Selected seed 195 with value = 0.4745 +Query 1/1: Action query time = 1.036 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5987 +t=42: Selected seed 195 with value = 0.5987 +Query 1/1: Action query time = 0.993 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6797 +t=58: Selected seed 195 with value = 0.6797 +Query 1/1: Action query time = 0.962 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7890 +t=74: Selected seed 195 with value = 0.7890 +Query 1/1: Action query time = 0.962 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9409 +t=90: Selected seed 195 with value = 0.9409 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=18--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=18--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 18 +# successes: 17 (94.4%) + +Task: put the wine bottle on top of the cabinet +Starting episode 19... +Query 1/1: Action query time = 0.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4461 +t=10: Selected seed 195 with value = 0.4461 +Query 1/1: Action query time = 0.962 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5239 +t=26: Selected seed 195 with value = 0.5239 +Query 1/1: Action query time = 1.003 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6178 +t=42: Selected seed 195 with value = 0.6178 +Query 1/1: Action query time = 1.020 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7117 +t=58: Selected seed 195 with value = 0.7117 +Query 1/1: Action query time = 1.234 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8794 +t=74: Selected seed 195 with value = 0.8794 +Query 1/1: Action query time = 1.062 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=19--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=19--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 19 +# successes: 18 (94.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 20... +Query 1/1: Action query time = 1.049 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4257 +t=10: Selected seed 195 with value = 0.4257 +Query 1/1: Action query time = 0.948 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4968 +t=26: Selected seed 195 with value = 0.4968 +Query 1/1: Action query time = 0.957 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5706 +t=42: Selected seed 195 with value = 0.5706 +Query 1/1: Action query time = 0.956 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6971 +t=58: Selected seed 195 with value = 0.6971 +Query 1/1: Action query time = 0.993 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8167 +t=74: Selected seed 195 with value = 0.8167 +Query 1/1: Action query time = 1.008 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9808 +t=90: Selected seed 195 with value = 0.9808 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=20--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=20--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 20 +# successes: 19 (95.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 21... +Query 1/1: Action query time = 1.102 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4697 +t=10: Selected seed 195 with value = 0.4697 +Query 1/1: Action query time = 1.216 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5442 +t=26: Selected seed 195 with value = 0.5442 +Query 1/1: Action query time = 1.211 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6532 +t=42: Selected seed 195 with value = 0.6532 +Query 1/1: Action query time = 1.094 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7680 +t=58: Selected seed 195 with value = 0.7680 +Query 1/1: Action query time = 1.100 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9062 +t=74: Selected seed 195 with value = 0.9062 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=21--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=21--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 21 +# successes: 20 (95.2%) + +Task: put the wine bottle on top of the cabinet +Starting episode 22... +Query 1/1: Action query time = 0.968 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4555 +t=10: Selected seed 195 with value = 0.4555 +Query 1/1: Action query time = 1.188 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5511 +t=26: Selected seed 195 with value = 0.5511 +Query 1/1: Action query time = 1.172 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6405 +t=42: Selected seed 195 with value = 0.6405 +Query 1/1: Action query time = 1.161 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7401 +t=58: Selected seed 195 with value = 0.7401 +Query 1/1: Action query time = 1.303 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8540 +t=74: Selected seed 195 with value = 0.8540 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=22--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=22--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 22 +# successes: 21 (95.5%) + +Task: put the wine bottle on top of the cabinet +Starting episode 23... +Query 1/1: Action query time = 0.976 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4201 +t=10: Selected seed 195 with value = 0.4201 +Query 1/1: Action query time = 1.227 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4563 +t=26: Selected seed 195 with value = 0.4563 +Query 1/1: Action query time = 1.202 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5390 +t=42: Selected seed 195 with value = 0.5390 +Query 1/1: Action query time = 1.206 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6669 +t=58: Selected seed 195 with value = 0.6669 +Query 1/1: Action query time = 0.977 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7621 +t=74: Selected seed 195 with value = 0.7621 +Query 1/1: Action query time = 1.130 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9218 +t=90: Selected seed 195 with value = 0.9218 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=23--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=23--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 23 +# successes: 22 (95.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 24... +Query 1/1: Action query time = 1.010 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4026 +t=10: Selected seed 195 with value = 0.4026 +Query 1/1: Action query time = 1.123 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5220 +t=26: Selected seed 195 with value = 0.5220 +Query 1/1: Action query time = 0.993 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6123 +t=42: Selected seed 195 with value = 0.6123 +Query 1/1: Action query time = 1.163 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7237 +t=58: Selected seed 195 with value = 0.7237 +Query 1/1: Action query time = 1.241 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8496 +t=74: Selected seed 195 with value = 0.8496 +Query 1/1: Action query time = 1.010 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=90: Selected seed 195 with value = 0.9977 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=24--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=24--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 24 +# successes: 23 (95.8%) + +Task: put the wine bottle on top of the cabinet +Starting episode 25... +Query 1/1: Action query time = 1.005 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3219 +t=10: Selected seed 195 with value = 0.3219 +Query 1/1: Action query time = 0.980 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4858 +t=26: Selected seed 195 with value = 0.4858 +Query 1/1: Action query time = 0.945 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5189 +t=42: Selected seed 195 with value = 0.5189 +Query 1/1: Action query time = 0.954 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5981 +t=58: Selected seed 195 with value = 0.5981 +Query 1/1: Action query time = 1.128 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7050 +t=74: Selected seed 195 with value = 0.7050 +Query 1/1: Action query time = 1.207 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8752 +t=90: Selected seed 195 with value = 0.8752 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=25--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=25--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 25 +# successes: 24 (96.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 26... +Query 1/1: Action query time = 1.114 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4521 +t=10: Selected seed 195 with value = 0.4521 +Query 1/1: Action query time = 1.142 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5117 +t=26: Selected seed 195 with value = 0.5117 +Query 1/1: Action query time = 1.147 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6431 +t=42: Selected seed 195 with value = 0.6431 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7223 +t=58: Selected seed 195 with value = 0.7223 +Query 1/1: Action query time = 0.962 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8680 +t=74: Selected seed 195 with value = 0.8680 +Query 1/1: Action query time = 0.966 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=26--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=26--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 26 +# successes: 25 (96.2%) + +Task: put the wine bottle on top of the cabinet +Starting episode 27... +Query 1/1: Action query time = 0.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3187 +t=10: Selected seed 195 with value = 0.3187 +Query 1/1: Action query time = 1.208 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3717 +t=26: Selected seed 195 with value = 0.3717 +Query 1/1: Action query time = 1.364 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5365 +t=42: Selected seed 195 with value = 0.5365 +Query 1/1: Action query time = 1.284 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6416 +t=58: Selected seed 195 with value = 0.6416 +Query 1/1: Action query time = 1.429 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7871 +t=74: Selected seed 195 with value = 0.7871 +Query 1/1: Action query time = 1.453 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9348 +t=90: Selected seed 195 with value = 0.9348 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=27--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=27--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 27 +# successes: 26 (96.3%) + +Task: put the wine bottle on top of the cabinet +Starting episode 28... +Query 1/1: Action query time = 1.196 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4192 +t=10: Selected seed 195 with value = 0.4192 +Query 1/1: Action query time = 1.173 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5091 +t=26: Selected seed 195 with value = 0.5091 +Query 1/1: Action query time = 1.170 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5995 +t=42: Selected seed 195 with value = 0.5995 +Query 1/1: Action query time = 1.071 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7089 +t=58: Selected seed 195 with value = 0.7089 +Query 1/1: Action query time = 1.017 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8323 +t=74: Selected seed 195 with value = 0.8323 +Query 1/1: Action query time = 0.979 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9750 +t=90: Selected seed 195 with value = 0.9750 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=28--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=28--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 28 +# successes: 27 (96.4%) + +Task: put the wine bottle on top of the cabinet +Starting episode 29... +Query 1/1: Action query time = 1.002 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3356 +t=10: Selected seed 195 with value = 0.3356 +Query 1/1: Action query time = 1.081 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3427 +t=26: Selected seed 195 with value = 0.3427 +Query 1/1: Action query time = 1.283 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4084 +t=42: Selected seed 195 with value = 0.4084 +Query 1/1: Action query time = 1.113 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4946 +t=58: Selected seed 195 with value = 0.4946 +Query 1/1: Action query time = 1.034 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5253 +t=74: Selected seed 195 with value = 0.5253 +Query 1/1: Action query time = 0.967 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5876 +t=90: Selected seed 195 with value = 0.5876 +Query 1/1: Action query time = 1.009 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6860 +t=106: Selected seed 195 with value = 0.6860 +Query 1/1: Action query time = 0.985 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8156 +t=122: Selected seed 195 with value = 0.8156 +Query 1/1: Action query time = 0.963 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9566 +t=138: Selected seed 195 with value = 0.9566 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=29--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=29--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 29 +# successes: 28 (96.6%) + +Task: put the wine bottle on top of the cabinet +Starting episode 30... +Query 1/1: Action query time = 1.321 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4528 +t=10: Selected seed 195 with value = 0.4528 +Query 1/1: Action query time = 1.393 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5367 +t=26: Selected seed 195 with value = 0.5367 +Query 1/1: Action query time = 1.419 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6272 +t=42: Selected seed 195 with value = 0.6272 +Query 1/1: Action query time = 1.099 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7765 +t=58: Selected seed 195 with value = 0.7765 +Query 1/1: Action query time = 1.188 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8912 +t=74: Selected seed 195 with value = 0.8912 +Query 1/1: Action query time = 1.101 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=30--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=30--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 30 +# successes: 29 (96.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 31... +Query 1/1: Action query time = 1.205 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3561 +t=10: Selected seed 195 with value = 0.3561 +Query 1/1: Action query time = 0.980 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3946 +t=26: Selected seed 195 with value = 0.3946 +Query 1/1: Action query time = 0.999 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4772 +t=42: Selected seed 195 with value = 0.4772 +Query 1/1: Action query time = 1.022 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6644 +t=58: Selected seed 195 with value = 0.6644 +Query 1/1: Action query time = 0.999 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8005 +t=74: Selected seed 195 with value = 0.8005 +Query 1/1: Action query time = 1.113 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9280 +t=90: Selected seed 195 with value = 0.9280 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=31--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=31--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 31 +# successes: 30 (96.8%) + +Task: put the wine bottle on top of the cabinet +Starting episode 32... +Query 1/1: Action query time = 1.419 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4123 +t=10: Selected seed 195 with value = 0.4123 +Query 1/1: Action query time = 1.153 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4924 +t=26: Selected seed 195 with value = 0.4924 +Query 1/1: Action query time = 0.979 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5597 +t=42: Selected seed 195 with value = 0.5597 +Query 1/1: Action query time = 0.977 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6982 +t=58: Selected seed 195 with value = 0.6982 +Query 1/1: Action query time = 0.995 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8423 +t=74: Selected seed 195 with value = 0.8423 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=32--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=32--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 32 +# successes: 31 (96.9%) + +Task: put the wine bottle on top of the cabinet +Starting episode 33... +Query 1/1: Action query time = 1.200 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4365 +t=10: Selected seed 195 with value = 0.4365 +Query 1/1: Action query time = 1.133 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4994 +t=26: Selected seed 195 with value = 0.4994 +Query 1/1: Action query time = 1.115 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5458 +t=42: Selected seed 195 with value = 0.5458 +Query 1/1: Action query time = 1.056 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6210 +t=58: Selected seed 195 with value = 0.6210 +Query 1/1: Action query time = 0.984 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7779 +t=74: Selected seed 195 with value = 0.7779 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8990 +t=90: Selected seed 195 with value = 0.8990 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=33--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=33--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 33 +# successes: 32 (97.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 34... +Query 1/1: Action query time = 1.128 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3834 +t=10: Selected seed 195 with value = 0.3834 +Query 1/1: Action query time = 1.000 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4348 +t=26: Selected seed 195 with value = 0.4348 +Query 1/1: Action query time = 1.000 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5089 +t=42: Selected seed 195 with value = 0.5089 +Query 1/1: Action query time = 0.973 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7104 +t=58: Selected seed 195 with value = 0.7104 +Query 1/1: Action query time = 0.952 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8606 +t=74: Selected seed 195 with value = 0.8606 +Query 1/1: Action query time = 1.059 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=90: Selected seed 195 with value = 0.9969 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=34--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=34--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 34 +# successes: 33 (97.1%) + +Task: put the wine bottle on top of the cabinet +Starting episode 35... +Query 1/1: Action query time = 0.988 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3330 +t=10: Selected seed 195 with value = 0.3330 +Query 1/1: Action query time = 0.984 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4218 +t=26: Selected seed 195 with value = 0.4218 +Query 1/1: Action query time = 1.007 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5233 +t=42: Selected seed 195 with value = 0.5233 +Query 1/1: Action query time = 1.034 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6689 +t=58: Selected seed 195 with value = 0.6689 +Query 1/1: Action query time = 0.995 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8460 +t=74: Selected seed 195 with value = 0.8460 +Query 1/1: Action query time = 1.001 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9797 +t=90: Selected seed 195 with value = 0.9797 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=35--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=35--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 35 +# successes: 34 (97.1%) + +Task: put the wine bottle on top of the cabinet +Starting episode 36... +Query 1/1: Action query time = 0.999 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4067 +t=10: Selected seed 195 with value = 0.4067 +Query 1/1: Action query time = 0.976 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4654 +t=26: Selected seed 195 with value = 0.4654 +Query 1/1: Action query time = 0.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5774 +t=42: Selected seed 195 with value = 0.5774 +Query 1/1: Action query time = 0.971 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6537 +t=58: Selected seed 195 with value = 0.6537 +Query 1/1: Action query time = 0.963 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8447 +t=74: Selected seed 195 with value = 0.8447 +Query 1/1: Action query time = 1.258 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9334 +t=90: Selected seed 195 with value = 0.9334 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=36--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=36--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 36 +# successes: 35 (97.2%) + +Task: put the wine bottle on top of the cabinet +Starting episode 37... +Query 1/1: Action query time = 0.989 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3713 +t=10: Selected seed 195 with value = 0.3713 +Query 1/1: Action query time = 1.123 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4741 +t=26: Selected seed 195 with value = 0.4741 +Query 1/1: Action query time = 0.993 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5310 +t=42: Selected seed 195 with value = 0.5310 +Query 1/1: Action query time = 0.996 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5950 +t=58: Selected seed 195 with value = 0.5950 +Query 1/1: Action query time = 0.988 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7272 +t=74: Selected seed 195 with value = 0.7272 +Query 1/1: Action query time = 1.003 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8755 +t=90: Selected seed 195 with value = 0.8755 +Query 1/1: Action query time = 0.970 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=37--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=37--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 37 +# successes: 36 (97.3%) + +Task: put the wine bottle on top of the cabinet +Starting episode 38... +Query 1/1: Action query time = 1.029 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3209 +t=10: Selected seed 195 with value = 0.3209 +Query 1/1: Action query time = 1.120 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3396 +t=26: Selected seed 195 with value = 0.3396 +Query 1/1: Action query time = 1.109 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4014 +t=42: Selected seed 195 with value = 0.4014 +Query 1/1: Action query time = 1.190 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4712 +t=58: Selected seed 195 with value = 0.4712 +Query 1/1: Action query time = 1.197 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5446 +t=74: Selected seed 195 with value = 0.5446 +Query 1/1: Action query time = 1.187 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6156 +t=90: Selected seed 195 with value = 0.6156 +Query 1/1: Action query time = 1.141 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6938 +t=106: Selected seed 195 with value = 0.6938 +Query 1/1: Action query time = 1.190 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8174 +t=122: Selected seed 195 with value = 0.8174 +Query 1/1: Action query time = 1.387 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9956 +t=138: Selected seed 195 with value = 0.9956 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=38--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=38--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 38 +# successes: 37 (97.4%) + +Task: put the wine bottle on top of the cabinet +Starting episode 39... +Query 1/1: Action query time = 1.368 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4317 +t=10: Selected seed 195 with value = 0.4317 +Query 1/1: Action query time = 1.419 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5042 +t=26: Selected seed 195 with value = 0.5042 +Query 1/1: Action query time = 1.405 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6084 +t=42: Selected seed 195 with value = 0.6084 +Query 1/1: Action query time = 1.306 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7337 +t=58: Selected seed 195 with value = 0.7337 +Query 1/1: Action query time = 1.131 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8489 +t=74: Selected seed 195 with value = 0.8489 +Query 1/1: Action query time = 1.078 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9938 +t=90: Selected seed 195 with value = 0.9938 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=39--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=39--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 39 +# successes: 38 (97.4%) + +Task: put the wine bottle on top of the cabinet +Starting episode 40... +Query 1/1: Action query time = 1.275 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3163 +t=10: Selected seed 195 with value = 0.3163 +Query 1/1: Action query time = 1.142 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4500 +t=26: Selected seed 195 with value = 0.4500 +Query 1/1: Action query time = 1.136 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5173 +t=42: Selected seed 195 with value = 0.5173 +Query 1/1: Action query time = 1.078 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6437 +t=58: Selected seed 195 with value = 0.6437 +Query 1/1: Action query time = 0.994 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7277 +t=74: Selected seed 195 with value = 0.7277 +Query 1/1: Action query time = 1.000 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8486 +t=90: Selected seed 195 with value = 0.8486 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=40--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=40--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 40 +# successes: 39 (97.5%) + +Task: put the wine bottle on top of the cabinet +Starting episode 41... +Query 1/1: Action query time = 0.983 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3690 +t=10: Selected seed 195 with value = 0.3690 +Query 1/1: Action query time = 0.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4092 +t=26: Selected seed 195 with value = 0.4092 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4984 +t=42: Selected seed 195 with value = 0.4984 +Query 1/1: Action query time = 0.976 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6180 +t=58: Selected seed 195 with value = 0.6180 +Query 1/1: Action query time = 0.989 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7364 +t=74: Selected seed 195 with value = 0.7364 +Query 1/1: Action query time = 0.970 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8684 +t=90: Selected seed 195 with value = 0.8684 +Query 1/1: Action query time = 1.133 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=41--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=41--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 41 +# successes: 40 (97.6%) + +Task: put the wine bottle on top of the cabinet +Starting episode 42... +Query 1/1: Action query time = 1.226 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4280 +t=10: Selected seed 195 with value = 0.4280 +Query 1/1: Action query time = 1.189 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4808 +t=26: Selected seed 195 with value = 0.4808 +Query 1/1: Action query time = 1.138 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5034 +t=42: Selected seed 195 with value = 0.5034 +Query 1/1: Action query time = 1.191 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5884 +t=58: Selected seed 195 with value = 0.5884 +Query 1/1: Action query time = 1.209 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6907 +t=74: Selected seed 195 with value = 0.6907 +Query 1/1: Action query time = 1.058 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8126 +t=90: Selected seed 195 with value = 0.8126 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=42--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=42--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 42 +# successes: 41 (97.6%) + +Task: put the wine bottle on top of the cabinet +Starting episode 43... +Query 1/1: Action query time = 1.016 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4057 +t=10: Selected seed 195 with value = 0.4057 +Query 1/1: Action query time = 0.984 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4900 +t=26: Selected seed 195 with value = 0.4900 +Query 1/1: Action query time = 0.963 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5604 +t=42: Selected seed 195 with value = 0.5604 +Query 1/1: Action query time = 0.984 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6082 +t=58: Selected seed 195 with value = 0.6082 +Query 1/1: Action query time = 0.980 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7219 +t=74: Selected seed 195 with value = 0.7219 +Query 1/1: Action query time = 0.990 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8095 +t=90: Selected seed 195 with value = 0.8095 +Query 1/1: Action query time = 0.995 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9437 +t=106: Selected seed 195 with value = 0.9437 +Query 1/1: Action query time = 1.056 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=122: Selected seed 195 with value = 0.9983 +Query 1/1: Action query time = 0.977 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=154: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 1.131 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=170: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 1.268 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=186: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 1.227 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9798 +t=202: Selected seed 195 with value = 0.9798 +Query 1/1: Action query time = 1.247 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9410 +t=218: Selected seed 195 with value = 0.9410 +Query 1/1: Action query time = 1.318 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9106 +t=234: Selected seed 195 with value = 0.9106 +Query 1/1: Action query time = 1.267 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8815 +t=250: Selected seed 195 with value = 0.8815 +Query 1/1: Action query time = 0.981 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8837 +t=266: Selected seed 195 with value = 0.8837 +Query 1/1: Action query time = 1.026 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8719 +t=282: Selected seed 195 with value = 0.8719 +Query 1/1: Action query time = 1.036 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8512 +t=298: Selected seed 195 with value = 0.8512 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=43--success=False--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=43--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 43 +# successes: 41 (95.3%) + +Task: put the wine bottle on top of the cabinet +Starting episode 44... +Query 1/1: Action query time = 0.986 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4086 +t=10: Selected seed 195 with value = 0.4086 +Query 1/1: Action query time = 0.952 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4057 +t=26: Selected seed 195 with value = 0.4057 +Query 1/1: Action query time = 0.968 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5618 +t=42: Selected seed 195 with value = 0.5618 +Query 1/1: Action query time = 0.992 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6672 +t=58: Selected seed 195 with value = 0.6672 +Query 1/1: Action query time = 0.970 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8293 +t=74: Selected seed 195 with value = 0.8293 +Query 1/1: Action query time = 0.961 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9434 +t=90: Selected seed 195 with value = 0.9434 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=44--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=44--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 44 +# successes: 42 (95.5%) + +Task: put the wine bottle on top of the cabinet +Starting episode 45... +Query 1/1: Action query time = 0.975 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3191 +t=10: Selected seed 195 with value = 0.3191 +Query 1/1: Action query time = 0.957 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4790 +t=26: Selected seed 195 with value = 0.4790 +Query 1/1: Action query time = 0.961 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6347 +t=42: Selected seed 195 with value = 0.6347 +Query 1/1: Action query time = 1.096 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7471 +t=58: Selected seed 195 with value = 0.7471 +Query 1/1: Action query time = 1.017 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8802 +t=74: Selected seed 195 with value = 0.8802 +Query 1/1: Action query time = 0.960 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=45--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=45--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 45 +# successes: 43 (95.6%) + +Task: put the wine bottle on top of the cabinet +Starting episode 46... +Query 1/1: Action query time = 1.037 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4765 +t=10: Selected seed 195 with value = 0.4765 +Query 1/1: Action query time = 0.973 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5389 +t=26: Selected seed 195 with value = 0.5389 +Query 1/1: Action query time = 1.033 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6390 +t=42: Selected seed 195 with value = 0.6390 +Query 1/1: Action query time = 1.072 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7327 +t=58: Selected seed 195 with value = 0.7327 +Query 1/1: Action query time = 1.072 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8864 +t=74: Selected seed 195 with value = 0.8864 +Query 1/1: Action query time = 1.058 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=46--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=46--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 46 +# successes: 44 (95.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 47... +Query 1/1: Action query time = 1.298 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4497 +t=10: Selected seed 195 with value = 0.4497 +Query 1/1: Action query time = 1.172 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5361 +t=26: Selected seed 195 with value = 0.5361 +Query 1/1: Action query time = 1.096 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6078 +t=42: Selected seed 195 with value = 0.6078 +Query 1/1: Action query time = 0.967 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7198 +t=58: Selected seed 195 with value = 0.7198 +Query 1/1: Action query time = 0.967 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9095 +t=74: Selected seed 195 with value = 0.9095 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=47--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=47--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 47 +# successes: 45 (95.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 48... +Query 1/1: Action query time = 1.018 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4544 +t=10: Selected seed 195 with value = 0.4544 +Query 1/1: Action query time = 0.980 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5382 +t=26: Selected seed 195 with value = 0.5382 +Query 1/1: Action query time = 0.994 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5953 +t=42: Selected seed 195 with value = 0.5953 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7270 +t=58: Selected seed 195 with value = 0.7270 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8547 +t=74: Selected seed 195 with value = 0.8547 +Query 1/1: Action query time = 0.957 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=90: Selected seed 195 with value = 0.9968 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=48--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=48--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 48 +# successes: 46 (95.8%) + +Task: put the wine bottle on top of the cabinet +Starting episode 49... +Query 1/1: Action query time = 1.168 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3474 +t=10: Selected seed 195 with value = 0.3474 +Query 1/1: Action query time = 0.963 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4500 +t=26: Selected seed 195 with value = 0.4500 +Query 1/1: Action query time = 0.955 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=42: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 1.222 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7147 +t=58: Selected seed 195 with value = 0.7147 +Query 1/1: Action query time = 1.203 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8687 +t=74: Selected seed 195 with value = 0.8687 +Query 1/1: Action query time = 1.089 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=49--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=49--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 49 +# successes: 47 (95.9%) + +Task: put the wine bottle on top of the cabinet +Starting episode 50... +Query 1/1: Action query time = 0.990 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4454 +t=10: Selected seed 195 with value = 0.4454 +Query 1/1: Action query time = 0.959 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5062 +t=26: Selected seed 195 with value = 0.5062 +Query 1/1: Action query time = 0.954 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6264 +t=42: Selected seed 195 with value = 0.6264 +Query 1/1: Action query time = 0.977 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6965 +t=58: Selected seed 195 with value = 0.6965 +Query 1/1: Action query time = 0.985 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8344 +t=74: Selected seed 195 with value = 0.8344 +Query 1/1: Action query time = 0.967 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=90: Selected seed 195 with value = 0.9909 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--episode=50--success=True--task=put_the_wine_bottle_on_top_of_the_cabine.mp4 +Saved rollout MP4 at path ./rollouts/realcl_i800_t2/2026_08_01-00_10_26--with_future_img--episode=50--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 50 +# successes: 48 (96.0%) +Current task success rate: 0.96 +Current total success rate: 0.96 +Final results: +Total episodes: 50 +Total successes: 48 +Overall success rate: 0.9600 (96.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-04_17_57--task1_new_ft_iter100.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-04_17_57--task1_new_ft_iter100.txt new file mode 100644 index 0000000000000000000000000000000000000000..b55709fe4b14d5acffe52fce48f782161f3e6ca2 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-04_17_57--task1_new_ft_iter100.txt @@ -0,0 +1,1768 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task1_new_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='task1_new_ft_iter100', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=True, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 1.575 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4891 +t=10: Selected seed 195 with value = 0.4891 +Query 1/1: Action query time = 0.967 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5980 +t=26: Selected seed 195 with value = 0.5980 +Query 1/1: Action query time = 0.964 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7262 +t=42: Selected seed 195 with value = 0.7262 +Query 1/1: Action query time = 0.981 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8231 +t=58: Selected seed 195 with value = 0.8231 +Query 1/1: Action query time = 0.968 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9786 +t=74: Selected seed 195 with value = 0.9786 +Query 1/1: Action query time = 0.952 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9905 +t=90: Selected seed 195 with value = 0.9905 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 0.993 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5225 +t=10: Selected seed 195 with value = 0.5225 +Query 1/1: Action query time = 0.957 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5726 +t=26: Selected seed 195 with value = 0.5726 +Query 1/1: Action query time = 0.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6829 +t=42: Selected seed 195 with value = 0.6829 +Query 1/1: Action query time = 0.973 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7987 +t=58: Selected seed 195 with value = 0.7987 +Query 1/1: Action query time = 0.984 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9590 +t=74: Selected seed 195 with value = 0.9590 +Query 1/1: Action query time = 0.967 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 0.963 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5295 +t=10: Selected seed 195 with value = 0.5295 +Query 1/1: Action query time = 0.954 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5836 +t=26: Selected seed 195 with value = 0.5836 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6660 +t=42: Selected seed 195 with value = 0.6660 +Query 1/1: Action query time = 0.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8304 +t=58: Selected seed 195 with value = 0.8304 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=74: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 0.958 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 0.982 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5033 +t=10: Selected seed 195 with value = 0.5033 +Query 1/1: Action query time = 0.967 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6072 +t=26: Selected seed 195 with value = 0.6072 +Query 1/1: Action query time = 0.969 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6792 +t=42: Selected seed 195 with value = 0.6792 +Query 1/1: Action query time = 0.964 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8131 +t=58: Selected seed 195 with value = 0.8131 +Query 1/1: Action query time = 0.983 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9442 +t=74: Selected seed 195 with value = 0.9442 +Query 1/1: Action query time = 0.956 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9934 +t=90: Selected seed 195 with value = 0.9934 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=4--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=4--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 0.975 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5010 +t=10: Selected seed 195 with value = 0.5010 +Query 1/1: Action query time = 0.979 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6067 +t=26: Selected seed 195 with value = 0.6067 +Query 1/1: Action query time = 0.960 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7079 +t=42: Selected seed 195 with value = 0.7079 +Query 1/1: Action query time = 0.960 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8347 +t=58: Selected seed 195 with value = 0.8347 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9687 +t=74: Selected seed 195 with value = 0.9687 +Query 1/1: Action query time = 0.970 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=90: Selected seed 195 with value = 0.9980 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=5--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=5--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 0.968 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5033 +t=10: Selected seed 195 with value = 0.5033 +Query 1/1: Action query time = 0.974 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5846 +t=26: Selected seed 195 with value = 0.5846 +Query 1/1: Action query time = 0.966 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6941 +t=42: Selected seed 195 with value = 0.6941 +Query 1/1: Action query time = 0.957 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8000 +t=58: Selected seed 195 with value = 0.8000 +Query 1/1: Action query time = 0.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9627 +t=74: Selected seed 195 with value = 0.9627 +Query 1/1: Action query time = 0.970 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=90: Selected seed 195 with value = 0.9989 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=6--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=6--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 0.991 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4032 +t=10: Selected seed 195 with value = 0.4032 +Query 1/1: Action query time = 0.986 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4998 +t=26: Selected seed 195 with value = 0.4998 +Query 1/1: Action query time = 0.977 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6238 +t=42: Selected seed 195 with value = 0.6238 +Query 1/1: Action query time = 0.976 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7565 +t=58: Selected seed 195 with value = 0.7565 +Query 1/1: Action query time = 0.956 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9066 +t=74: Selected seed 195 with value = 0.9066 +Query 1/1: Action query time = 0.968 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=7--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=7--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 0.969 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4929 +t=10: Selected seed 195 with value = 0.4929 +Query 1/1: Action query time = 0.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5895 +t=26: Selected seed 195 with value = 0.5895 +Query 1/1: Action query time = 0.981 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7309 +t=42: Selected seed 195 with value = 0.7309 +Query 1/1: Action query time = 0.961 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8582 +t=58: Selected seed 195 with value = 0.8582 +Query 1/1: Action query time = 0.977 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9571 +t=74: Selected seed 195 with value = 0.9571 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=8--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=8--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 0.980 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5046 +t=10: Selected seed 195 with value = 0.5046 +Query 1/1: Action query time = 0.960 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5867 +t=26: Selected seed 195 with value = 0.5867 +Query 1/1: Action query time = 0.976 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6501 +t=42: Selected seed 195 with value = 0.6501 +Query 1/1: Action query time = 0.976 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7823 +t=58: Selected seed 195 with value = 0.7823 +Query 1/1: Action query time = 0.967 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9258 +t=74: Selected seed 195 with value = 0.9258 +Query 1/1: Action query time = 0.988 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9996 +t=90: Selected seed 195 with value = 0.9996 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=9--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=9--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 0.972 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5124 +t=10: Selected seed 195 with value = 0.5124 +Query 1/1: Action query time = 0.966 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6159 +t=26: Selected seed 195 with value = 0.6159 +Query 1/1: Action query time = 0.950 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7180 +t=42: Selected seed 195 with value = 0.7180 +Query 1/1: Action query time = 0.974 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8167 +t=58: Selected seed 195 with value = 0.8167 +Query 1/1: Action query time = 0.963 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9684 +t=74: Selected seed 195 with value = 0.9684 +Query 1/1: Action query time = 0.978 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=90: Selected seed 195 with value = 0.9966 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=10--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=10--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 0.999 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4730 +t=10: Selected seed 195 with value = 0.4730 +Query 1/1: Action query time = 0.975 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5871 +t=26: Selected seed 195 with value = 0.5871 +Query 1/1: Action query time = 0.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6741 +t=42: Selected seed 195 with value = 0.6741 +Query 1/1: Action query time = 0.958 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7851 +t=58: Selected seed 195 with value = 0.7851 +Query 1/1: Action query time = 0.980 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9081 +t=74: Selected seed 195 with value = 0.9081 +Query 1/1: Action query time = 0.968 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9714 +t=90: Selected seed 195 with value = 0.9714 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=11--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=11--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 0.978 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5227 +t=10: Selected seed 195 with value = 0.5227 +Query 1/1: Action query time = 0.962 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=26: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 0.958 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6928 +t=42: Selected seed 195 with value = 0.6928 +Query 1/1: Action query time = 0.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8083 +t=58: Selected seed 195 with value = 0.8083 +Query 1/1: Action query time = 0.983 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9405 +t=74: Selected seed 195 with value = 0.9405 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9967 +t=90: Selected seed 195 with value = 0.9967 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=12--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=12--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 0.982 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5220 +t=10: Selected seed 195 with value = 0.5220 +Query 1/1: Action query time = 0.971 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5893 +t=26: Selected seed 195 with value = 0.5893 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7049 +t=42: Selected seed 195 with value = 0.7049 +Query 1/1: Action query time = 0.967 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8245 +t=58: Selected seed 195 with value = 0.8245 +Query 1/1: Action query time = 0.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9579 +t=74: Selected seed 195 with value = 0.9579 +Query 1/1: Action query time = 0.960 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=13--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=13--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) + +Task: put the bowl on the stove +Starting episode 14... +Query 1/1: Action query time = 0.964 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5305 +t=10: Selected seed 195 with value = 0.5305 +Query 1/1: Action query time = 0.967 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6106 +t=26: Selected seed 195 with value = 0.6106 +Query 1/1: Action query time = 0.957 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7106 +t=42: Selected seed 195 with value = 0.7106 +Query 1/1: Action query time = 0.959 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7744 +t=58: Selected seed 195 with value = 0.7744 +Query 1/1: Action query time = 0.977 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9630 +t=74: Selected seed 195 with value = 0.9630 +Query 1/1: Action query time = 0.955 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=14--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=14--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 14 +# successes: 14 (100.0%) + +Task: put the bowl on the stove +Starting episode 15... +Query 1/1: Action query time = 0.981 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5058 +t=10: Selected seed 195 with value = 0.5058 +Query 1/1: Action query time = 0.963 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6173 +t=26: Selected seed 195 with value = 0.6173 +Query 1/1: Action query time = 0.958 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7484 +t=42: Selected seed 195 with value = 0.7484 +Query 1/1: Action query time = 0.970 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8847 +t=58: Selected seed 195 with value = 0.8847 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9927 +t=74: Selected seed 195 with value = 0.9927 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=15--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=15--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 15 +# successes: 15 (100.0%) + +Task: put the bowl on the stove +Starting episode 16... +Query 1/1: Action query time = 0.992 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4014 +t=10: Selected seed 195 with value = 0.4014 +Query 1/1: Action query time = 0.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4876 +t=26: Selected seed 195 with value = 0.4876 +Query 1/1: Action query time = 0.971 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6206 +t=42: Selected seed 195 with value = 0.6206 +Query 1/1: Action query time = 0.973 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7564 +t=58: Selected seed 195 with value = 0.7564 +Query 1/1: Action query time = 0.974 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8840 +t=74: Selected seed 195 with value = 0.8840 +Query 1/1: Action query time = 0.978 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=16--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=16--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 16 +# successes: 16 (100.0%) + +Task: put the bowl on the stove +Starting episode 17... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4758 +t=10: Selected seed 195 with value = 0.4758 +Query 1/1: Action query time = 0.976 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6138 +t=26: Selected seed 195 with value = 0.6138 +Query 1/1: Action query time = 0.979 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6995 +t=42: Selected seed 195 with value = 0.6995 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8124 +t=58: Selected seed 195 with value = 0.8124 +Query 1/1: Action query time = 0.988 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9498 +t=74: Selected seed 195 with value = 0.9498 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=90: Selected seed 195 with value = 0.9957 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=17--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=17--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 17 +# successes: 17 (100.0%) + +Task: put the bowl on the stove +Starting episode 18... +Query 1/1: Action query time = 0.992 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4968 +t=10: Selected seed 195 with value = 0.4968 +Query 1/1: Action query time = 0.979 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5873 +t=26: Selected seed 195 with value = 0.5873 +Query 1/1: Action query time = 0.959 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6768 +t=42: Selected seed 195 with value = 0.6768 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8140 +t=58: Selected seed 195 with value = 0.8140 +Query 1/1: Action query time = 0.964 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9434 +t=74: Selected seed 195 with value = 0.9434 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=90: Selected seed 195 with value = 0.9994 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=18--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=18--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 18 +# successes: 18 (100.0%) + +Task: put the bowl on the stove +Starting episode 19... +Query 1/1: Action query time = 0.993 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4990 +t=10: Selected seed 195 with value = 0.4990 +Query 1/1: Action query time = 0.985 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5955 +t=26: Selected seed 195 with value = 0.5955 +Query 1/1: Action query time = 0.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6871 +t=42: Selected seed 195 with value = 0.6871 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7705 +t=58: Selected seed 195 with value = 0.7705 +Query 1/1: Action query time = 0.987 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9382 +t=74: Selected seed 195 with value = 0.9382 +Query 1/1: Action query time = 0.979 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=19--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=19--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 19 +# successes: 19 (100.0%) + +Task: put the bowl on the stove +Starting episode 20... +Query 1/1: Action query time = 0.978 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5312 +t=10: Selected seed 195 with value = 0.5312 +Query 1/1: Action query time = 0.963 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6207 +t=26: Selected seed 195 with value = 0.6207 +Query 1/1: Action query time = 0.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7381 +t=42: Selected seed 195 with value = 0.7381 +Query 1/1: Action query time = 0.976 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8558 +t=58: Selected seed 195 with value = 0.8558 +Query 1/1: Action query time = 0.966 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9853 +t=74: Selected seed 195 with value = 0.9853 +Query 1/1: Action query time = 0.960 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=90: Selected seed 195 with value = 0.9991 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=20--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=20--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 20 +# successes: 20 (100.0%) + +Task: put the bowl on the stove +Starting episode 21... +Query 1/1: Action query time = 0.976 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4894 +t=10: Selected seed 195 with value = 0.4894 +Query 1/1: Action query time = 0.961 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5878 +t=26: Selected seed 195 with value = 0.5878 +Query 1/1: Action query time = 0.962 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6857 +t=42: Selected seed 195 with value = 0.6857 +Query 1/1: Action query time = 0.981 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7770 +t=58: Selected seed 195 with value = 0.7770 +Query 1/1: Action query time = 1.004 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9394 +t=74: Selected seed 195 with value = 0.9394 +Query 1/1: Action query time = 0.976 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9907 +t=90: Selected seed 195 with value = 0.9907 +Query 1/1: Action query time = 0.964 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.978 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.965 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.970 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.970 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9775 +t=298: Selected seed 195 with value = 0.9775 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=21--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=21--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 21 +# successes: 20 (95.2%) + +Task: put the bowl on the stove +Starting episode 22... +Query 1/1: Action query time = 0.991 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4122 +t=10: Selected seed 195 with value = 0.4122 +Query 1/1: Action query time = 0.979 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4992 +t=26: Selected seed 195 with value = 0.4992 +Query 1/1: Action query time = 0.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6026 +t=42: Selected seed 195 with value = 0.6026 +Query 1/1: Action query time = 0.972 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6949 +t=58: Selected seed 195 with value = 0.6949 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8081 +t=74: Selected seed 195 with value = 0.8081 +Query 1/1: Action query time = 0.961 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9598 +t=90: Selected seed 195 with value = 0.9598 +Query 1/1: Action query time = 0.962 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=106: Selected seed 195 with value = 0.9963 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=22--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=22--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 22 +# successes: 21 (95.5%) + +Task: put the bowl on the stove +Starting episode 23... +Query 1/1: Action query time = 0.973 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4739 +t=10: Selected seed 195 with value = 0.4739 +Query 1/1: Action query time = 0.984 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5801 +t=26: Selected seed 195 with value = 0.5801 +Query 1/1: Action query time = 0.963 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6896 +t=42: Selected seed 195 with value = 0.6896 +Query 1/1: Action query time = 0.978 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7922 +t=58: Selected seed 195 with value = 0.7922 +Query 1/1: Action query time = 0.973 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9592 +t=74: Selected seed 195 with value = 0.9592 +Query 1/1: Action query time = 0.984 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=23--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=23--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 23 +# successes: 22 (95.7%) + +Task: put the bowl on the stove +Starting episode 24... +Query 1/1: Action query time = 0.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5207 +t=10: Selected seed 195 with value = 0.5207 +Query 1/1: Action query time = 0.978 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6094 +t=26: Selected seed 195 with value = 0.6094 +Query 1/1: Action query time = 0.984 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7059 +t=42: Selected seed 195 with value = 0.7059 +Query 1/1: Action query time = 0.962 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8338 +t=58: Selected seed 195 with value = 0.8338 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9535 +t=74: Selected seed 195 with value = 0.9535 +Query 1/1: Action query time = 0.965 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=90: Selected seed 195 with value = 0.9965 +Query 1/1: Action query time = 0.968 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.971 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.966 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.977 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.982 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.988 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=24--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=24--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 24 +# successes: 22 (91.7%) + +Task: put the bowl on the stove +Starting episode 25... +Query 1/1: Action query time = 1.008 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4987 +t=10: Selected seed 195 with value = 0.4987 +Query 1/1: Action query time = 0.990 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5668 +t=26: Selected seed 195 with value = 0.5668 +Query 1/1: Action query time = 0.983 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6789 +t=42: Selected seed 195 with value = 0.6789 +Query 1/1: Action query time = 0.984 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7840 +t=58: Selected seed 195 with value = 0.7840 +Query 1/1: Action query time = 0.990 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9510 +t=74: Selected seed 195 with value = 0.9510 +Query 1/1: Action query time = 0.961 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9996 +t=90: Selected seed 195 with value = 0.9996 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=25--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=25--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 25 +# successes: 23 (92.0%) + +Task: put the bowl on the stove +Starting episode 26... +Query 1/1: Action query time = 0.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5307 +t=10: Selected seed 195 with value = 0.5307 +Query 1/1: Action query time = 0.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6253 +t=26: Selected seed 195 with value = 0.6253 +Query 1/1: Action query time = 0.983 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7270 +t=42: Selected seed 195 with value = 0.7270 +Query 1/1: Action query time = 0.972 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8431 +t=58: Selected seed 195 with value = 0.8431 +Query 1/1: Action query time = 0.978 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9754 +t=74: Selected seed 195 with value = 0.9754 +Query 1/1: Action query time = 0.986 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=26--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=26--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 26 +# successes: 24 (92.3%) + +Task: put the bowl on the stove +Starting episode 27... +Query 1/1: Action query time = 0.963 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5186 +t=10: Selected seed 195 with value = 0.5186 +Query 1/1: Action query time = 0.976 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6232 +t=26: Selected seed 195 with value = 0.6232 +Query 1/1: Action query time = 0.997 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7284 +t=42: Selected seed 195 with value = 0.7284 +Query 1/1: Action query time = 0.970 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8803 +t=58: Selected seed 195 with value = 0.8803 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=27--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=27--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 27 +# successes: 25 (92.6%) + +Task: put the bowl on the stove +Starting episode 28... +Query 1/1: Action query time = 0.975 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4719 +t=10: Selected seed 195 with value = 0.4719 +Query 1/1: Action query time = 0.990 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6035 +t=26: Selected seed 195 with value = 0.6035 +Query 1/1: Action query time = 0.970 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7129 +t=42: Selected seed 195 with value = 0.7129 +Query 1/1: Action query time = 0.960 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8215 +t=58: Selected seed 195 with value = 0.8215 +Query 1/1: Action query time = 0.984 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9675 +t=74: Selected seed 195 with value = 0.9675 +Query 1/1: Action query time = 0.988 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=90: Selected seed 195 with value = 0.9962 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=28--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=28--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 28 +# successes: 26 (92.9%) + +Task: put the bowl on the stove +Starting episode 29... +Query 1/1: Action query time = 0.992 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5012 +t=10: Selected seed 195 with value = 0.5012 +Query 1/1: Action query time = 0.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6137 +t=26: Selected seed 195 with value = 0.6137 +Query 1/1: Action query time = 0.993 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7198 +t=42: Selected seed 195 with value = 0.7198 +Query 1/1: Action query time = 0.995 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8086 +t=58: Selected seed 195 with value = 0.8086 +Query 1/1: Action query time = 0.968 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9413 +t=74: Selected seed 195 with value = 0.9413 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=90: Selected seed 195 with value = 0.9998 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=29--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=29--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 29 +# successes: 27 (93.1%) + +Task: put the bowl on the stove +Starting episode 30... +Query 1/1: Action query time = 0.996 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5527 +t=10: Selected seed 195 with value = 0.5527 +Query 1/1: Action query time = 0.986 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6207 +t=26: Selected seed 195 with value = 0.6207 +Query 1/1: Action query time = 0.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7223 +t=42: Selected seed 195 with value = 0.7223 +Query 1/1: Action query time = 0.971 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8221 +t=58: Selected seed 195 with value = 0.8221 +Query 1/1: Action query time = 0.968 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9575 +t=74: Selected seed 195 with value = 0.9575 +Query 1/1: Action query time = 0.982 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=90: Selected seed 195 with value = 0.9945 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=30--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=30--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 30 +# successes: 28 (93.3%) + +Task: put the bowl on the stove +Starting episode 31... +Query 1/1: Action query time = 0.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5135 +t=10: Selected seed 195 with value = 0.5135 +Query 1/1: Action query time = 0.974 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6016 +t=26: Selected seed 195 with value = 0.6016 +Query 1/1: Action query time = 0.976 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7029 +t=42: Selected seed 195 with value = 0.7029 +Query 1/1: Action query time = 0.986 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8246 +t=58: Selected seed 195 with value = 0.8246 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9537 +t=74: Selected seed 195 with value = 0.9537 +Query 1/1: Action query time = 0.979 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=31--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=31--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 31 +# successes: 29 (93.5%) + +Task: put the bowl on the stove +Starting episode 32... +Query 1/1: Action query time = 0.992 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4996 +t=10: Selected seed 195 with value = 0.4996 +Query 1/1: Action query time = 0.971 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5838 +t=26: Selected seed 195 with value = 0.5838 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6681 +t=42: Selected seed 195 with value = 0.6681 +Query 1/1: Action query time = 0.978 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7807 +t=58: Selected seed 195 with value = 0.7807 +Query 1/1: Action query time = 0.982 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9080 +t=74: Selected seed 195 with value = 0.9080 +Query 1/1: Action query time = 0.984 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9730 +t=90: Selected seed 195 with value = 0.9730 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=32--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=32--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 32 +# successes: 30 (93.8%) + +Task: put the bowl on the stove +Starting episode 33... +Query 1/1: Action query time = 0.983 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5189 +t=10: Selected seed 195 with value = 0.5189 +Query 1/1: Action query time = 0.982 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6037 +t=26: Selected seed 195 with value = 0.6037 +Query 1/1: Action query time = 0.981 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7356 +t=42: Selected seed 195 with value = 0.7356 +Query 1/1: Action query time = 0.981 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8346 +t=58: Selected seed 195 with value = 0.8346 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9636 +t=74: Selected seed 195 with value = 0.9636 +Query 1/1: Action query time = 0.985 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=33--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=33--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 33 +# successes: 31 (93.9%) + +Task: put the bowl on the stove +Starting episode 34... +Query 1/1: Action query time = 0.988 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4797 +t=10: Selected seed 195 with value = 0.4797 +Query 1/1: Action query time = 0.971 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5896 +t=26: Selected seed 195 with value = 0.5896 +Query 1/1: Action query time = 0.985 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7104 +t=42: Selected seed 195 with value = 0.7104 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8086 +t=58: Selected seed 195 with value = 0.8086 +Query 1/1: Action query time = 0.994 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9481 +t=74: Selected seed 195 with value = 0.9481 +Query 1/1: Action query time = 0.974 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=34--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=34--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 34 +# successes: 32 (94.1%) + +Task: put the bowl on the stove +Starting episode 35... +Query 1/1: Action query time = 1.001 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5240 +t=10: Selected seed 195 with value = 0.5240 +Query 1/1: Action query time = 0.982 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6135 +t=26: Selected seed 195 with value = 0.6135 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7250 +t=42: Selected seed 195 with value = 0.7250 +Query 1/1: Action query time = 0.968 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8386 +t=58: Selected seed 195 with value = 0.8386 +Query 1/1: Action query time = 0.986 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9704 +t=74: Selected seed 195 with value = 0.9704 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9913 +t=90: Selected seed 195 with value = 0.9913 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=35--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=35--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 35 +# successes: 33 (94.3%) + +Task: put the bowl on the stove +Starting episode 36... +Query 1/1: Action query time = 0.980 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5100 +t=10: Selected seed 195 with value = 0.5100 +Query 1/1: Action query time = 0.992 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5963 +t=26: Selected seed 195 with value = 0.5963 +Query 1/1: Action query time = 0.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7061 +t=42: Selected seed 195 with value = 0.7061 +Query 1/1: Action query time = 0.993 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8550 +t=58: Selected seed 195 with value = 0.8550 +Query 1/1: Action query time = 0.992 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9819 +t=74: Selected seed 195 with value = 0.9819 +Query 1/1: Action query time = 0.994 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=36--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=36--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 36 +# successes: 34 (94.4%) + +Task: put the bowl on the stove +Starting episode 37... +Query 1/1: Action query time = 0.994 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5160 +t=10: Selected seed 195 with value = 0.5160 +Query 1/1: Action query time = 0.986 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6261 +t=26: Selected seed 195 with value = 0.6261 +Query 1/1: Action query time = 0.983 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7480 +t=42: Selected seed 195 with value = 0.7480 +Query 1/1: Action query time = 0.973 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8630 +t=58: Selected seed 195 with value = 0.8630 +Query 1/1: Action query time = 0.984 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9887 +t=74: Selected seed 195 with value = 0.9887 +Query 1/1: Action query time = 0.974 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=37--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=37--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 37 +# successes: 35 (94.6%) + +Task: put the bowl on the stove +Starting episode 38... +Query 1/1: Action query time = 0.993 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5221 +t=10: Selected seed 195 with value = 0.5221 +Query 1/1: Action query time = 0.990 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6027 +t=26: Selected seed 195 with value = 0.6027 +Query 1/1: Action query time = 0.982 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7010 +t=42: Selected seed 195 with value = 0.7010 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8204 +t=58: Selected seed 195 with value = 0.8204 +Query 1/1: Action query time = 0.986 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9664 +t=74: Selected seed 195 with value = 0.9664 +Query 1/1: Action query time = 0.973 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=38--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=38--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 38 +# successes: 36 (94.7%) + +Task: put the bowl on the stove +Starting episode 39... +Query 1/1: Action query time = 0.989 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4845 +t=10: Selected seed 195 with value = 0.4845 +Query 1/1: Action query time = 0.981 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5579 +t=26: Selected seed 195 with value = 0.5579 +Query 1/1: Action query time = 0.959 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6894 +t=42: Selected seed 195 with value = 0.6894 +Query 1/1: Action query time = 0.971 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7925 +t=58: Selected seed 195 with value = 0.7925 +Query 1/1: Action query time = 1.001 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9615 +t=74: Selected seed 195 with value = 0.9615 +Query 1/1: Action query time = 0.962 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=90: Selected seed 195 with value = 0.9968 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=39--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=39--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 39 +# successes: 37 (94.9%) + +Task: put the bowl on the stove +Starting episode 40... +Query 1/1: Action query time = 0.993 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5306 +t=10: Selected seed 195 with value = 0.5306 +Query 1/1: Action query time = 0.964 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5984 +t=26: Selected seed 195 with value = 0.5984 +Query 1/1: Action query time = 0.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6724 +t=42: Selected seed 195 with value = 0.6724 +Query 1/1: Action query time = 0.972 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8200 +t=58: Selected seed 195 with value = 0.8200 +Query 1/1: Action query time = 0.958 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9271 +t=74: Selected seed 195 with value = 0.9271 +Query 1/1: Action query time = 1.000 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9894 +t=90: Selected seed 195 with value = 0.9894 +Query 1/1: Action query time = 0.976 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.974 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.995 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.983 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.972 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.976 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.973 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.989 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.985 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.968 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.967 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.990 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=40--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=40--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 40 +# successes: 37 (92.5%) + +Task: put the bowl on the stove +Starting episode 41... +Query 1/1: Action query time = 0.992 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4288 +t=10: Selected seed 195 with value = 0.4288 +Query 1/1: Action query time = 0.974 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5251 +t=26: Selected seed 195 with value = 0.5251 +Query 1/1: Action query time = 0.981 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6548 +t=42: Selected seed 195 with value = 0.6548 +Query 1/1: Action query time = 0.987 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7504 +t=58: Selected seed 195 with value = 0.7504 +Query 1/1: Action query time = 0.977 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8980 +t=74: Selected seed 195 with value = 0.8980 +Query 1/1: Action query time = 0.975 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=41--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=41--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 41 +# successes: 38 (92.7%) + +Task: put the bowl on the stove +Starting episode 42... +Query 1/1: Action query time = 0.983 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4490 +t=10: Selected seed 195 with value = 0.4490 +Query 1/1: Action query time = 0.973 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5100 +t=26: Selected seed 195 with value = 0.5100 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7055 +t=42: Selected seed 195 with value = 0.7055 +Query 1/1: Action query time = 0.982 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8101 +t=58: Selected seed 195 with value = 0.8101 +Query 1/1: Action query time = 0.990 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9815 +t=74: Selected seed 195 with value = 0.9815 +Query 1/1: Action query time = 0.988 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9932 +t=90: Selected seed 195 with value = 0.9932 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=42--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=42--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 42 +# successes: 39 (92.9%) + +Task: put the bowl on the stove +Starting episode 43... +Query 1/1: Action query time = 0.976 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5010 +t=10: Selected seed 195 with value = 0.5010 +Query 1/1: Action query time = 0.991 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5931 +t=26: Selected seed 195 with value = 0.5931 +Query 1/1: Action query time = 0.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7018 +t=42: Selected seed 195 with value = 0.7018 +Query 1/1: Action query time = 0.993 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8211 +t=58: Selected seed 195 with value = 0.8211 +Query 1/1: Action query time = 0.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9569 +t=74: Selected seed 195 with value = 0.9569 +Query 1/1: Action query time = 0.971 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=90: Selected seed 195 with value = 0.9994 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=43--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=43--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 43 +# successes: 40 (93.0%) + +Task: put the bowl on the stove +Starting episode 44... +Query 1/1: Action query time = 0.986 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5352 +t=10: Selected seed 195 with value = 0.5352 +Query 1/1: Action query time = 0.970 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6135 +t=26: Selected seed 195 with value = 0.6135 +Query 1/1: Action query time = 0.964 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7133 +t=42: Selected seed 195 with value = 0.7133 +Query 1/1: Action query time = 0.979 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8241 +t=58: Selected seed 195 with value = 0.8241 +Query 1/1: Action query time = 0.983 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9508 +t=74: Selected seed 195 with value = 0.9508 +Query 1/1: Action query time = 0.971 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=90: Selected seed 195 with value = 0.9990 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=44--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=44--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 44 +# successes: 41 (93.2%) + +Task: put the bowl on the stove +Starting episode 45... +Query 1/1: Action query time = 0.980 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4963 +t=10: Selected seed 195 with value = 0.4963 +Query 1/1: Action query time = 0.976 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6022 +t=26: Selected seed 195 with value = 0.6022 +Query 1/1: Action query time = 0.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7014 +t=42: Selected seed 195 with value = 0.7014 +Query 1/1: Action query time = 0.986 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7687 +t=58: Selected seed 195 with value = 0.7687 +Query 1/1: Action query time = 0.981 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9437 +t=74: Selected seed 195 with value = 0.9437 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=45--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=45--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 45 +# successes: 42 (93.3%) + +Task: put the bowl on the stove +Starting episode 46... +Query 1/1: Action query time = 0.995 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4269 +t=10: Selected seed 195 with value = 0.4269 +Query 1/1: Action query time = 0.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5495 +t=26: Selected seed 195 with value = 0.5495 +Query 1/1: Action query time = 0.963 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6269 +t=42: Selected seed 195 with value = 0.6269 +Query 1/1: Action query time = 0.982 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7218 +t=58: Selected seed 195 with value = 0.7218 +Query 1/1: Action query time = 0.984 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8515 +t=74: Selected seed 195 with value = 0.8515 +Query 1/1: Action query time = 0.981 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=46--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=46--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 46 +# successes: 43 (93.5%) + +Task: put the bowl on the stove +Starting episode 47... +Query 1/1: Action query time = 0.997 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4550 +t=10: Selected seed 195 with value = 0.4550 +Query 1/1: Action query time = 0.980 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5913 +t=26: Selected seed 195 with value = 0.5913 +Query 1/1: Action query time = 0.966 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6703 +t=42: Selected seed 195 with value = 0.6703 +Query 1/1: Action query time = 0.992 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7987 +t=58: Selected seed 195 with value = 0.7987 +Query 1/1: Action query time = 0.973 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9162 +t=74: Selected seed 195 with value = 0.9162 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=90: Selected seed 195 with value = 0.9989 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=47--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=47--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 47 +# successes: 44 (93.6%) + +Task: put the bowl on the stove +Starting episode 48... +Query 1/1: Action query time = 0.971 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5193 +t=10: Selected seed 195 with value = 0.5193 +Query 1/1: Action query time = 0.970 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6293 +t=26: Selected seed 195 with value = 0.6293 +Query 1/1: Action query time = 0.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7224 +t=42: Selected seed 195 with value = 0.7224 +Query 1/1: Action query time = 0.980 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8666 +t=58: Selected seed 195 with value = 0.8666 +Query 1/1: Action query time = 0.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=74: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 0.991 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=48--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=48--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 48 +# successes: 45 (93.8%) + +Task: put the bowl on the stove +Starting episode 49... +Query 1/1: Action query time = 0.974 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4977 +t=10: Selected seed 195 with value = 0.4977 +Query 1/1: Action query time = 0.963 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5878 +t=26: Selected seed 195 with value = 0.5878 +Query 1/1: Action query time = 0.970 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6902 +t=42: Selected seed 195 with value = 0.6902 +Query 1/1: Action query time = 0.971 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8164 +t=58: Selected seed 195 with value = 0.8164 +Query 1/1: Action query time = 0.980 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9618 +t=74: Selected seed 195 with value = 0.9618 +Query 1/1: Action query time = 0.987 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=49--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=49--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 49 +# successes: 46 (93.9%) + +Task: put the bowl on the stove +Starting episode 50... +Query 1/1: Action query time = 0.973 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5526 +t=10: Selected seed 195 with value = 0.5526 +Query 1/1: Action query time = 0.989 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6488 +t=26: Selected seed 195 with value = 0.6488 +Query 1/1: Action query time = 0.977 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7109 +t=42: Selected seed 195 with value = 0.7109 +Query 1/1: Action query time = 0.965 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8476 +t=58: Selected seed 195 with value = 0.8476 +Query 1/1: Action query time = 0.966 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=74: Selected seed 195 with value = 0.9966 +Query 1/1: Action query time = 0.965 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--episode=50--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/task1_new_ft_iter100/2026_08_01-04_17_57--with_future_img--episode=50--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 50 +# successes: 47 (94.0%) +Current task success rate: 0.94 +Current total success rate: 0.94 +Final results: +Total episodes: 50 +Total successes: 47 +Overall success rate: 0.9400 (94.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-11_38_09--forget_cl_iter350_par_t0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-11_38_09--forget_cl_iter350_par_t0.txt new file mode 100644 index 0000000000000000000000000000000000000000..294a45de215d93c1e5b2663ae083ab8927bc167b --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-11_38_09--forget_cl_iter350_par_t0.txt @@ -0,0 +1,2160 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_from100_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='forget_cl_iter350_par_t0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=True, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 1.557 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3719 +t=10: Selected seed 195 with value = 0.3719 +Query 1/1: Action query time = 1.095 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4272 +t=26: Selected seed 195 with value = 0.4272 +Query 1/1: Action query time = 2.215 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5093 +t=42: Selected seed 195 with value = 0.5093 +Query 1/1: Action query time = 2.708 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5926 +t=58: Selected seed 195 with value = 0.5926 +Query 1/1: Action query time = 2.230 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6435 +t=74: Selected seed 195 with value = 0.6435 +Query 1/1: Action query time = 2.359 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7045 +t=90: Selected seed 195 with value = 0.7045 +Query 1/1: Action query time = 2.663 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8866 +t=106: Selected seed 195 with value = 0.8866 +Query 1/1: Action query time = 1.996 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=1--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.393 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3629 +t=10: Selected seed 195 with value = 0.3629 +Query 1/1: Action query time = 2.743 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3996 +t=26: Selected seed 195 with value = 0.3996 +Query 1/1: Action query time = 3.410 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4696 +t=42: Selected seed 195 with value = 0.4696 +Query 1/1: Action query time = 2.978 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5490 +t=58: Selected seed 195 with value = 0.5490 +Query 1/1: Action query time = 1.985 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6615 +t=74: Selected seed 195 with value = 0.6615 +Query 1/1: Action query time = 1.242 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7649 +t=90: Selected seed 195 with value = 0.7649 +Query 1/1: Action query time = 1.930 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9342 +t=106: Selected seed 195 with value = 0.9342 +Query 1/1: Action query time = 2.875 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=2--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 1.887 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4020 +t=10: Selected seed 195 with value = 0.4020 +Query 1/1: Action query time = 1.697 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4376 +t=26: Selected seed 195 with value = 0.4376 +Query 1/1: Action query time = 1.460 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4940 +t=42: Selected seed 195 with value = 0.4940 +Query 1/1: Action query time = 2.039 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6058 +t=58: Selected seed 195 with value = 0.6058 +Query 1/1: Action query time = 2.609 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6639 +t=74: Selected seed 195 with value = 0.6639 +Query 1/1: Action query time = 2.811 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7698 +t=90: Selected seed 195 with value = 0.7698 +Query 1/1: Action query time = 2.210 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9210 +t=106: Selected seed 195 with value = 0.9210 +Query 1/1: Action query time = 2.380 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=3--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 4... +Query 1/1: Action query time = 1.884 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3458 +t=10: Selected seed 195 with value = 0.3458 +Query 1/1: Action query time = 2.641 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3916 +t=26: Selected seed 195 with value = 0.3916 +Query 1/1: Action query time = 2.654 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5129 +t=42: Selected seed 195 with value = 0.5129 +Query 1/1: Action query time = 2.773 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5419 +t=58: Selected seed 195 with value = 0.5419 +Query 1/1: Action query time = 2.288 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6503 +t=74: Selected seed 195 with value = 0.6503 +Query 1/1: Action query time = 1.394 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7950 +t=90: Selected seed 195 with value = 0.7950 +Query 1/1: Action query time = 1.705 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9105 +t=106: Selected seed 195 with value = 0.9105 +Query 1/1: Action query time = 0.994 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=4--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=4--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 5... +Query 1/1: Action query time = 2.790 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3867 +t=10: Selected seed 195 with value = 0.3867 +Query 1/1: Action query time = 1.925 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4449 +t=26: Selected seed 195 with value = 0.4449 +Query 1/1: Action query time = 2.724 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5169 +t=42: Selected seed 195 with value = 0.5169 +Query 1/1: Action query time = 2.455 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5869 +t=58: Selected seed 195 with value = 0.5869 +Query 1/1: Action query time = 2.113 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6938 +t=74: Selected seed 195 with value = 0.6938 +Query 1/1: Action query time = 1.689 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8185 +t=90: Selected seed 195 with value = 0.8185 +Query 1/1: Action query time = 2.048 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9482 +t=106: Selected seed 195 with value = 0.9482 +Query 1/1: Action query time = 2.789 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=5--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=5--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 6... +Query 1/1: Action query time = 2.427 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3585 +t=10: Selected seed 195 with value = 0.3585 +Query 1/1: Action query time = 2.441 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4107 +t=26: Selected seed 195 with value = 0.4107 +Query 1/1: Action query time = 2.828 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4913 +t=42: Selected seed 195 with value = 0.4913 +Query 1/1: Action query time = 1.294 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5585 +t=58: Selected seed 195 with value = 0.5585 +Query 1/1: Action query time = 1.585 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6897 +t=74: Selected seed 195 with value = 0.6897 +Query 1/1: Action query time = 2.067 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7388 +t=90: Selected seed 195 with value = 0.7388 +Query 1/1: Action query time = 2.305 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7359 +t=106: Selected seed 195 with value = 0.7359 +Query 1/1: Action query time = 2.437 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9464 +t=122: Selected seed 195 with value = 0.9464 +Query 1/1: Action query time = 2.386 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=6--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=6--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 7... +Query 1/1: Action query time = 1.813 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3376 +t=10: Selected seed 195 with value = 0.3376 +Query 1/1: Action query time = 2.177 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3174 +t=26: Selected seed 195 with value = 0.3174 +Query 1/1: Action query time = 2.556 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4249 +t=42: Selected seed 195 with value = 0.4249 +Query 1/1: Action query time = 1.682 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5628 +t=58: Selected seed 195 with value = 0.5628 +Query 1/1: Action query time = 2.263 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7025 +t=74: Selected seed 195 with value = 0.7025 +Query 1/1: Action query time = 2.324 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8310 +t=90: Selected seed 195 with value = 0.8310 +Query 1/1: Action query time = 2.549 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9393 +t=106: Selected seed 195 with value = 0.9393 +Query 1/1: Action query time = 2.452 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=122: Selected seed 195 with value = 0.9973 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=7--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=7--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 8... +Query 1/1: Action query time = 2.411 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3504 +t=10: Selected seed 195 with value = 0.3504 +Query 1/1: Action query time = 2.481 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4016 +t=26: Selected seed 195 with value = 0.4016 +Query 1/1: Action query time = 2.548 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4699 +t=42: Selected seed 195 with value = 0.4699 +Query 1/1: Action query time = 1.957 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5449 +t=58: Selected seed 195 with value = 0.5449 +Query 1/1: Action query time = 2.134 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6514 +t=74: Selected seed 195 with value = 0.6514 +Query 1/1: Action query time = 2.026 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7551 +t=90: Selected seed 195 with value = 0.7551 +Query 1/1: Action query time = 1.664 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9256 +t=106: Selected seed 195 with value = 0.9256 +Query 1/1: Action query time = 1.454 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=8--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=8--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 9... +Query 1/1: Action query time = 2.575 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3552 +t=10: Selected seed 195 with value = 0.3552 +Query 1/1: Action query time = 2.507 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4468 +t=26: Selected seed 195 with value = 0.4468 +Query 1/1: Action query time = 2.034 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4943 +t=42: Selected seed 195 with value = 0.4943 +Query 1/1: Action query time = 1.717 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6132 +t=58: Selected seed 195 with value = 0.6132 +Query 1/1: Action query time = 1.667 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7214 +t=74: Selected seed 195 with value = 0.7214 +Query 1/1: Action query time = 2.816 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8178 +t=90: Selected seed 195 with value = 0.8178 +Query 1/1: Action query time = 2.554 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9415 +t=106: Selected seed 195 with value = 0.9415 +Query 1/1: Action query time = 2.274 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=9--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=9--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 10... +Query 1/1: Action query time = 2.163 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4201 +t=10: Selected seed 195 with value = 0.4201 +Query 1/1: Action query time = 2.337 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4795 +t=26: Selected seed 195 with value = 0.4795 +Query 1/1: Action query time = 2.354 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5458 +t=42: Selected seed 195 with value = 0.5458 +Query 1/1: Action query time = 2.243 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6308 +t=58: Selected seed 195 with value = 0.6308 +Query 1/1: Action query time = 1.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7316 +t=74: Selected seed 195 with value = 0.7316 +Query 1/1: Action query time = 2.312 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8119 +t=90: Selected seed 195 with value = 0.8119 +Query 1/1: Action query time = 2.295 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9799 +t=106: Selected seed 195 with value = 0.9799 +Query 1/1: Action query time = 1.740 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=10--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=10--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 11... +Query 1/1: Action query time = 3.045 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3635 +t=10: Selected seed 195 with value = 0.3635 +Query 1/1: Action query time = 2.612 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3755 +t=26: Selected seed 195 with value = 0.3755 +Query 1/1: Action query time = 2.499 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4532 +t=42: Selected seed 195 with value = 0.4532 +Query 1/1: Action query time = 2.372 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5197 +t=58: Selected seed 195 with value = 0.5197 +Query 1/1: Action query time = 1.772 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5845 +t=74: Selected seed 195 with value = 0.5845 +Query 1/1: Action query time = 1.960 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6143 +t=90: Selected seed 195 with value = 0.6143 +Query 1/1: Action query time = 2.518 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8661 +t=106: Selected seed 195 with value = 0.8661 +Query 1/1: Action query time = 1.947 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=122: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 1.768 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9900 +t=138: Selected seed 195 with value = 0.9900 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=11--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=11--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 12... +Query 1/1: Action query time = 1.803 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3657 +t=10: Selected seed 195 with value = 0.3657 +Query 1/1: Action query time = 1.627 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4239 +t=26: Selected seed 195 with value = 0.4239 +Query 1/1: Action query time = 2.330 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5047 +t=42: Selected seed 195 with value = 0.5047 +Query 1/1: Action query time = 2.638 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5802 +t=58: Selected seed 195 with value = 0.5802 +Query 1/1: Action query time = 1.631 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7140 +t=74: Selected seed 195 with value = 0.7140 +Query 1/1: Action query time = 2.288 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8219 +t=90: Selected seed 195 with value = 0.8219 +Query 1/1: Action query time = 2.474 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8911 +t=106: Selected seed 195 with value = 0.8911 +Query 1/1: Action query time = 2.606 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9802 +t=122: Selected seed 195 with value = 0.9802 +Query 1/1: Action query time = 2.521 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=12--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=12--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 13... +Query 1/1: Action query time = 1.629 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3697 +t=10: Selected seed 195 with value = 0.3697 +Query 1/1: Action query time = 1.915 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4019 +t=26: Selected seed 195 with value = 0.4019 +Query 1/1: Action query time = 2.233 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4512 +t=42: Selected seed 195 with value = 0.4512 +Query 1/1: Action query time = 2.583 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5327 +t=58: Selected seed 195 with value = 0.5327 +Query 1/1: Action query time = 2.562 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6250 +t=74: Selected seed 195 with value = 0.6250 +Query 1/1: Action query time = 2.168 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7489 +t=90: Selected seed 195 with value = 0.7489 +Query 1/1: Action query time = 2.191 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9099 +t=106: Selected seed 195 with value = 0.9099 +Query 1/1: Action query time = 2.288 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=13--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=13--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 14... +Query 1/1: Action query time = 1.829 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3744 +t=10: Selected seed 195 with value = 0.3744 +Query 1/1: Action query time = 1.779 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4088 +t=26: Selected seed 195 with value = 0.4088 +Query 1/1: Action query time = 1.427 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4757 +t=42: Selected seed 195 with value = 0.4757 +Query 1/1: Action query time = 2.173 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5519 +t=58: Selected seed 195 with value = 0.5519 +Query 1/1: Action query time = 2.683 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6535 +t=74: Selected seed 195 with value = 0.6535 +Query 1/1: Action query time = 2.488 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7484 +t=90: Selected seed 195 with value = 0.7484 +Query 1/1: Action query time = 2.856 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9231 +t=106: Selected seed 195 with value = 0.9231 +Query 1/1: Action query time = 2.599 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=14--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=14--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 14 +# successes: 14 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 15... +Query 1/1: Action query time = 2.199 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3213 +t=10: Selected seed 195 with value = 0.3213 +Query 1/1: Action query time = 2.038 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4194 +t=26: Selected seed 195 with value = 0.4194 +Query 1/1: Action query time = 2.114 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4787 +t=42: Selected seed 195 with value = 0.4787 +Query 1/1: Action query time = 1.889 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5602 +t=58: Selected seed 195 with value = 0.5602 +Query 1/1: Action query time = 2.066 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6223 +t=74: Selected seed 195 with value = 0.6223 +Query 1/1: Action query time = 2.538 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7235 +t=90: Selected seed 195 with value = 0.7235 +Query 1/1: Action query time = 2.730 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8502 +t=106: Selected seed 195 with value = 0.8502 +Query 1/1: Action query time = 2.785 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9625 +t=122: Selected seed 195 with value = 0.9625 +Query 1/1: Action query time = 2.235 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=138: Selected seed 195 with value = 0.9981 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=15--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=15--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 15 +# successes: 15 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 16... +Query 1/1: Action query time = 2.460 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3506 +t=10: Selected seed 195 with value = 0.3506 +Query 1/1: Action query time = 1.672 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4170 +t=26: Selected seed 195 with value = 0.4170 +Query 1/1: Action query time = 1.781 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4764 +t=42: Selected seed 195 with value = 0.4764 +Query 1/1: Action query time = 1.856 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5455 +t=58: Selected seed 195 with value = 0.5455 +Query 1/1: Action query time = 2.254 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6445 +t=74: Selected seed 195 with value = 0.6445 +Query 1/1: Action query time = 2.124 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7469 +t=90: Selected seed 195 with value = 0.7469 +Query 1/1: Action query time = 2.232 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9077 +t=106: Selected seed 195 with value = 0.9077 +Query 1/1: Action query time = 2.632 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=16--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=16--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 16 +# successes: 16 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 17... +Query 1/1: Action query time = 2.401 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3794 +t=10: Selected seed 195 with value = 0.3794 +Query 1/1: Action query time = 2.343 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4321 +t=26: Selected seed 195 with value = 0.4321 +Query 1/1: Action query time = 2.049 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5226 +t=42: Selected seed 195 with value = 0.5226 +Query 1/1: Action query time = 2.287 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6079 +t=58: Selected seed 195 with value = 0.6079 +Query 1/1: Action query time = 2.002 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7096 +t=74: Selected seed 195 with value = 0.7096 +Query 1/1: Action query time = 2.114 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8452 +t=90: Selected seed 195 with value = 0.8452 +Query 1/1: Action query time = 2.558 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9841 +t=106: Selected seed 195 with value = 0.9841 +Query 1/1: Action query time = 2.450 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=17--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=17--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 17 +# successes: 17 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 18... +Query 1/1: Action query time = 1.677 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3588 +t=10: Selected seed 195 with value = 0.3588 +Query 1/1: Action query time = 1.523 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4258 +t=26: Selected seed 195 with value = 0.4258 +Query 1/1: Action query time = 2.358 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5182 +t=42: Selected seed 195 with value = 0.5182 +Query 1/1: Action query time = 3.034 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5305 +t=58: Selected seed 195 with value = 0.5305 +Query 1/1: Action query time = 2.833 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5429 +t=74: Selected seed 195 with value = 0.5429 +Query 1/1: Action query time = 2.518 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7564 +t=90: Selected seed 195 with value = 0.7564 +Query 1/1: Action query time = 1.936 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8996 +t=106: Selected seed 195 with value = 0.8996 +Query 1/1: Action query time = 1.856 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9944 +t=122: Selected seed 195 with value = 0.9944 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=18--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=18--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 18 +# successes: 18 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 19... +Query 1/1: Action query time = 2.850 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3502 +t=10: Selected seed 195 with value = 0.3502 +Query 1/1: Action query time = 2.622 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4676 +t=26: Selected seed 195 with value = 0.4676 +Query 1/1: Action query time = 2.588 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5262 +t=42: Selected seed 195 with value = 0.5262 +Query 1/1: Action query time = 1.939 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6084 +t=58: Selected seed 195 with value = 0.6084 +Query 1/1: Action query time = 1.658 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7247 +t=74: Selected seed 195 with value = 0.7247 +Query 1/1: Action query time = 2.682 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7998 +t=90: Selected seed 195 with value = 0.7998 +Query 1/1: Action query time = 2.476 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6258 +t=106: Selected seed 195 with value = 0.6258 +Query 1/1: Action query time = 2.570 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7336 +t=122: Selected seed 195 with value = 0.7336 +Query 1/1: Action query time = 1.976 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9418 +t=138: Selected seed 195 with value = 0.9418 +Query 1/1: Action query time = 2.167 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9822 +t=154: Selected seed 195 with value = 0.9822 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=19--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=19--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 19 +# successes: 19 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 20... +Query 1/1: Action query time = 1.631 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4208 +t=10: Selected seed 195 with value = 0.4208 +Query 1/1: Action query time = 2.708 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4857 +t=26: Selected seed 195 with value = 0.4857 +Query 1/1: Action query time = 2.639 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5559 +t=42: Selected seed 195 with value = 0.5559 +Query 1/1: Action query time = 2.097 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6388 +t=58: Selected seed 195 with value = 0.6388 +Query 1/1: Action query time = 2.412 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7363 +t=74: Selected seed 195 with value = 0.7363 +Query 1/1: Action query time = 2.544 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8353 +t=90: Selected seed 195 with value = 0.8353 +Query 1/1: Action query time = 2.313 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.857 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=20--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=20--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 20 +# successes: 20 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 21... +Query 1/1: Action query time = 1.433 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3488 +t=10: Selected seed 195 with value = 0.3488 +Query 1/1: Action query time = 2.140 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2898 +t=26: Selected seed 195 with value = 0.2898 +Query 1/1: Action query time = 2.619 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4916 +t=42: Selected seed 195 with value = 0.4916 +Query 1/1: Action query time = 2.470 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5656 +t=58: Selected seed 195 with value = 0.5656 +Query 1/1: Action query time = 1.838 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6559 +t=74: Selected seed 195 with value = 0.6559 +Query 1/1: Action query time = 2.203 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8247 +t=90: Selected seed 195 with value = 0.8247 +Query 1/1: Action query time = 2.583 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9652 +t=106: Selected seed 195 with value = 0.9652 +Query 1/1: Action query time = 2.466 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=21--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=21--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 21 +# successes: 21 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 22... +Query 1/1: Action query time = 2.165 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3595 +t=10: Selected seed 195 with value = 0.3595 +Query 1/1: Action query time = 1.969 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4002 +t=26: Selected seed 195 with value = 0.4002 +Query 1/1: Action query time = 2.121 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4597 +t=42: Selected seed 195 with value = 0.4597 +Query 1/1: Action query time = 2.252 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5330 +t=58: Selected seed 195 with value = 0.5330 +Query 1/1: Action query time = 1.839 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6685 +t=74: Selected seed 195 with value = 0.6685 +Query 1/1: Action query time = 2.467 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7467 +t=90: Selected seed 195 with value = 0.7467 +Query 1/1: Action query time = 1.980 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9214 +t=106: Selected seed 195 with value = 0.9214 +Query 1/1: Action query time = 2.494 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=22--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=22--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 22 +# successes: 22 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 23... +Query 1/1: Action query time = 3.207 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3965 +t=10: Selected seed 195 with value = 0.3965 +Query 1/1: Action query time = 2.867 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4506 +t=26: Selected seed 195 with value = 0.4506 +Query 1/1: Action query time = 2.066 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5173 +t=42: Selected seed 195 with value = 0.5173 +Query 1/1: Action query time = 2.316 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6193 +t=58: Selected seed 195 with value = 0.6193 +Query 1/1: Action query time = 2.906 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7406 +t=74: Selected seed 195 with value = 0.7406 +Query 1/1: Action query time = 3.077 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8697 +t=90: Selected seed 195 with value = 0.8697 +Query 1/1: Action query time = 2.612 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9720 +t=106: Selected seed 195 with value = 0.9720 +Query 1/1: Action query time = 3.374 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=23--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=23--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 23 +# successes: 23 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 24... +Query 1/1: Action query time = 1.848 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3276 +t=10: Selected seed 195 with value = 0.3276 +Query 1/1: Action query time = 2.382 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3513 +t=26: Selected seed 195 with value = 0.3513 +Query 1/1: Action query time = 2.261 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4263 +t=42: Selected seed 195 with value = 0.4263 +Query 1/1: Action query time = 2.201 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5255 +t=58: Selected seed 195 with value = 0.5255 +Query 1/1: Action query time = 2.284 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6229 +t=74: Selected seed 195 with value = 0.6229 +Query 1/1: Action query time = 1.478 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7445 +t=90: Selected seed 195 with value = 0.7445 +Query 1/1: Action query time = 1.848 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9121 +t=106: Selected seed 195 with value = 0.9121 +Query 1/1: Action query time = 3.047 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.503 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=24--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=24--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 24 +# successes: 24 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 25... +Query 1/1: Action query time = 1.901 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3797 +t=10: Selected seed 195 with value = 0.3797 +Query 1/1: Action query time = 2.548 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4045 +t=26: Selected seed 195 with value = 0.4045 +Query 1/1: Action query time = 2.783 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4739 +t=42: Selected seed 195 with value = 0.4739 +Query 1/1: Action query time = 2.956 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5596 +t=58: Selected seed 195 with value = 0.5596 +Query 1/1: Action query time = 3.435 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7105 +t=74: Selected seed 195 with value = 0.7105 +Query 1/1: Action query time = 2.506 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8200 +t=90: Selected seed 195 with value = 0.8200 +Query 1/1: Action query time = 1.945 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9652 +t=106: Selected seed 195 with value = 0.9652 +Query 1/1: Action query time = 2.201 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=25--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=25--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 25 +# successes: 25 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 26... +Query 1/1: Action query time = 1.696 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3717 +t=10: Selected seed 195 with value = 0.3717 +Query 1/1: Action query time = 1.910 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4337 +t=26: Selected seed 195 with value = 0.4337 +Query 1/1: Action query time = 2.228 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4876 +t=42: Selected seed 195 with value = 0.4876 +Query 1/1: Action query time = 2.297 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5507 +t=58: Selected seed 195 with value = 0.5507 +Query 1/1: Action query time = 2.729 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6718 +t=74: Selected seed 195 with value = 0.6718 +Query 1/1: Action query time = 2.672 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7907 +t=90: Selected seed 195 with value = 0.7907 +Query 1/1: Action query time = 1.907 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9155 +t=106: Selected seed 195 with value = 0.9155 +Query 1/1: Action query time = 2.520 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=26--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=26--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 26 +# successes: 26 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 27... +Query 1/1: Action query time = 1.936 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3294 +t=10: Selected seed 195 with value = 0.3294 +Query 1/1: Action query time = 1.962 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3856 +t=26: Selected seed 195 with value = 0.3856 +Query 1/1: Action query time = 2.731 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4461 +t=42: Selected seed 195 with value = 0.4461 +Query 1/1: Action query time = 2.587 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5083 +t=58: Selected seed 195 with value = 0.5083 +Query 1/1: Action query time = 2.770 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5986 +t=74: Selected seed 195 with value = 0.5986 +Query 1/1: Action query time = 2.443 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7154 +t=90: Selected seed 195 with value = 0.7154 +Query 1/1: Action query time = 2.437 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8145 +t=106: Selected seed 195 with value = 0.8145 +Query 1/1: Action query time = 2.008 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9767 +t=122: Selected seed 195 with value = 0.9767 +Query 1/1: Action query time = 1.457 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=27--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=27--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 27 +# successes: 27 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 28... +Query 1/1: Action query time = 2.065 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3686 +t=10: Selected seed 195 with value = 0.3686 +Query 1/1: Action query time = 2.380 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4358 +t=26: Selected seed 195 with value = 0.4358 +Query 1/1: Action query time = 1.871 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5056 +t=42: Selected seed 195 with value = 0.5056 +Query 1/1: Action query time = 1.773 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5920 +t=58: Selected seed 195 with value = 0.5920 +Query 1/1: Action query time = 1.887 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7069 +t=74: Selected seed 195 with value = 0.7069 +Query 1/1: Action query time = 1.935 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8539 +t=90: Selected seed 195 with value = 0.8539 +Query 1/1: Action query time = 1.917 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9584 +t=106: Selected seed 195 with value = 0.9584 +Query 1/1: Action query time = 2.600 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=28--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=28--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 28 +# successes: 28 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 29... +Query 1/1: Action query time = 2.747 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3666 +t=10: Selected seed 195 with value = 0.3666 +Query 1/1: Action query time = 1.807 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3841 +t=26: Selected seed 195 with value = 0.3841 +Query 1/1: Action query time = 1.893 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4157 +t=42: Selected seed 195 with value = 0.4157 +Query 1/1: Action query time = 1.894 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4951 +t=58: Selected seed 195 with value = 0.4951 +Query 1/1: Action query time = 2.210 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5497 +t=74: Selected seed 195 with value = 0.5497 +Query 1/1: Action query time = 2.392 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6267 +t=90: Selected seed 195 with value = 0.6267 +Query 1/1: Action query time = 2.287 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9232 +t=106: Selected seed 195 with value = 0.9232 +Query 1/1: Action query time = 2.508 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=122: Selected seed 195 with value = 0.9955 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=29--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=29--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 29 +# successes: 29 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 30... +Query 1/1: Action query time = 1.205 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3620 +t=10: Selected seed 195 with value = 0.3620 +Query 1/1: Action query time = 1.761 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3915 +t=26: Selected seed 195 with value = 0.3915 +Query 1/1: Action query time = 2.697 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4372 +t=42: Selected seed 195 with value = 0.4372 +Query 1/1: Action query time = 2.702 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5009 +t=58: Selected seed 195 with value = 0.5009 +Query 1/1: Action query time = 2.729 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6366 +t=74: Selected seed 195 with value = 0.6366 +Query 1/1: Action query time = 2.425 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7825 +t=90: Selected seed 195 with value = 0.7825 +Query 1/1: Action query time = 1.969 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9038 +t=106: Selected seed 195 with value = 0.9038 +Query 1/1: Action query time = 2.202 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9883 +t=122: Selected seed 195 with value = 0.9883 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=30--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=30--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 30 +# successes: 30 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 31... +Query 1/1: Action query time = 2.595 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3660 +t=10: Selected seed 195 with value = 0.3660 +Query 1/1: Action query time = 2.586 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4398 +t=26: Selected seed 195 with value = 0.4398 +Query 1/1: Action query time = 2.517 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5127 +t=42: Selected seed 195 with value = 0.5127 +Query 1/1: Action query time = 1.921 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5840 +t=58: Selected seed 195 with value = 0.5840 +Query 1/1: Action query time = 1.744 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7266 +t=74: Selected seed 195 with value = 0.7266 +Query 1/1: Action query time = 2.118 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8757 +t=90: Selected seed 195 with value = 0.8757 +Query 1/1: Action query time = 2.289 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=106: Selected seed 195 with value = 0.9965 +Query 1/1: Action query time = 2.574 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=31--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=31--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 31 +# successes: 31 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 32... +Query 1/1: Action query time = 2.062 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3048 +t=10: Selected seed 195 with value = 0.3048 +Query 1/1: Action query time = 2.173 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3615 +t=26: Selected seed 195 with value = 0.3615 +Query 1/1: Action query time = 2.092 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4023 +t=42: Selected seed 195 with value = 0.4023 +Query 1/1: Action query time = 2.786 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4952 +t=58: Selected seed 195 with value = 0.4952 +Query 1/1: Action query time = 2.497 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5603 +t=74: Selected seed 195 with value = 0.5603 +Query 1/1: Action query time = 2.379 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6187 +t=90: Selected seed 195 with value = 0.6187 +Query 1/1: Action query time = 2.487 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9177 +t=106: Selected seed 195 with value = 0.9177 +Query 1/1: Action query time = 2.519 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=122: Selected seed 195 with value = 0.9991 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=32--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=32--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 32 +# successes: 32 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 33... +Query 1/1: Action query time = 1.904 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3606 +t=10: Selected seed 195 with value = 0.3606 +Query 1/1: Action query time = 3.268 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4210 +t=26: Selected seed 195 with value = 0.4210 +Query 1/1: Action query time = 3.100 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5058 +t=42: Selected seed 195 with value = 0.5058 +Query 1/1: Action query time = 2.592 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4790 +t=58: Selected seed 195 with value = 0.4790 +Query 1/1: Action query time = 2.619 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5494 +t=74: Selected seed 195 with value = 0.5494 +Query 1/1: Action query time = 2.358 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6207 +t=90: Selected seed 195 with value = 0.6207 +Query 1/1: Action query time = 2.420 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8677 +t=106: Selected seed 195 with value = 0.8677 +Query 1/1: Action query time = 2.398 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9904 +t=122: Selected seed 195 with value = 0.9904 +Query 1/1: Action query time = 2.166 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=138: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=33--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=33--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 33 +# successes: 33 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 34... +Query 1/1: Action query time = 2.356 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3700 +t=10: Selected seed 195 with value = 0.3700 +Query 1/1: Action query time = 1.829 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3939 +t=26: Selected seed 195 with value = 0.3939 +Query 1/1: Action query time = 1.813 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4757 +t=42: Selected seed 195 with value = 0.4757 +Query 1/1: Action query time = 2.076 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5484 +t=58: Selected seed 195 with value = 0.5484 +Query 1/1: Action query time = 2.525 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7167 +t=74: Selected seed 195 with value = 0.7167 +Query 1/1: Action query time = 2.497 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8550 +t=90: Selected seed 195 with value = 0.8550 +Query 1/1: Action query time = 2.674 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9717 +t=106: Selected seed 195 with value = 0.9717 +Query 1/1: Action query time = 2.113 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=34--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=34--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 34 +# successes: 34 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 35... +Query 1/1: Action query time = 2.246 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3684 +t=10: Selected seed 195 with value = 0.3684 +Query 1/1: Action query time = 3.203 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4193 +t=26: Selected seed 195 with value = 0.4193 +Query 1/1: Action query time = 2.745 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4511 +t=42: Selected seed 195 with value = 0.4511 +Query 1/1: Action query time = 2.324 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5804 +t=58: Selected seed 195 with value = 0.5804 +Query 1/1: Action query time = 2.245 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6961 +t=74: Selected seed 195 with value = 0.6961 +Query 1/1: Action query time = 1.912 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8201 +t=90: Selected seed 195 with value = 0.8201 +Query 1/1: Action query time = 2.473 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=106: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 2.096 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=35--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=35--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 35 +# successes: 35 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 36... +Query 1/1: Action query time = 2.577 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3598 +t=10: Selected seed 195 with value = 0.3598 +Query 1/1: Action query time = 1.971 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3789 +t=26: Selected seed 195 with value = 0.3789 +Query 1/1: Action query time = 2.092 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4169 +t=42: Selected seed 195 with value = 0.4169 +Query 1/1: Action query time = 2.547 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5208 +t=58: Selected seed 195 with value = 0.5208 +Query 1/1: Action query time = 3.338 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6257 +t=74: Selected seed 195 with value = 0.6257 +Query 1/1: Action query time = 2.418 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7578 +t=90: Selected seed 195 with value = 0.7578 +Query 1/1: Action query time = 2.034 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9069 +t=106: Selected seed 195 with value = 0.9069 +Query 1/1: Action query time = 1.982 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=36--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=36--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 36 +# successes: 36 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 37... +Query 1/1: Action query time = 3.113 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3496 +t=10: Selected seed 195 with value = 0.3496 +Query 1/1: Action query time = 2.299 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2954 +t=26: Selected seed 195 with value = 0.2954 +Query 1/1: Action query time = 2.060 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3490 +t=42: Selected seed 195 with value = 0.3490 +Query 1/1: Action query time = 1.891 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3901 +t=58: Selected seed 195 with value = 0.3901 +Query 1/1: Action query time = 2.423 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4094 +t=74: Selected seed 195 with value = 0.4094 +Query 1/1: Action query time = 2.550 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5447 +t=90: Selected seed 195 with value = 0.5447 +Query 1/1: Action query time = 2.240 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6840 +t=106: Selected seed 195 with value = 0.6840 +Query 1/1: Action query time = 2.375 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7792 +t=122: Selected seed 195 with value = 0.7792 +Query 1/1: Action query time = 2.551 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8151 +t=138: Selected seed 195 with value = 0.8151 +Query 1/1: Action query time = 2.081 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7696 +t=154: Selected seed 195 with value = 0.7696 +Query 1/1: Action query time = 1.405 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6196 +t=170: Selected seed 195 with value = 0.6196 +Query 1/1: Action query time = 1.669 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6812 +t=186: Selected seed 195 with value = 0.6812 +Query 1/1: Action query time = 2.215 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7925 +t=202: Selected seed 195 with value = 0.7925 +Query 1/1: Action query time = 2.296 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8378 +t=218: Selected seed 195 with value = 0.8378 +Query 1/1: Action query time = 2.500 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8118 +t=234: Selected seed 195 with value = 0.8118 +Query 1/1: Action query time = 1.849 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8501 +t=250: Selected seed 195 with value = 0.8501 +Query 1/1: Action query time = 1.856 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8680 +t=266: Selected seed 195 with value = 0.8680 +Query 1/1: Action query time = 1.522 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9019 +t=282: Selected seed 195 with value = 0.9019 +Query 1/1: Action query time = 1.584 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9068 +t=298: Selected seed 195 with value = 0.9068 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=37--success=False--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=37--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 37 +# successes: 36 (97.3%) + +Task: open the middle drawer of the cabinet +Starting episode 38... +Query 1/1: Action query time = 2.547 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3567 +t=10: Selected seed 195 with value = 0.3567 +Query 1/1: Action query time = 2.512 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3796 +t=26: Selected seed 195 with value = 0.3796 +Query 1/1: Action query time = 2.562 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4183 +t=42: Selected seed 195 with value = 0.4183 +Query 1/1: Action query time = 2.040 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5111 +t=58: Selected seed 195 with value = 0.5111 +Query 1/1: Action query time = 2.013 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6386 +t=74: Selected seed 195 with value = 0.6386 +Query 1/1: Action query time = 2.721 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7566 +t=90: Selected seed 195 with value = 0.7566 +Query 1/1: Action query time = 2.218 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9146 +t=106: Selected seed 195 with value = 0.9146 +Query 1/1: Action query time = 2.127 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=38--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=38--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 38 +# successes: 37 (97.4%) + +Task: open the middle drawer of the cabinet +Starting episode 39... +Query 1/1: Action query time = 1.442 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3747 +t=10: Selected seed 195 with value = 0.3747 +Query 1/1: Action query time = 1.015 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4240 +t=26: Selected seed 195 with value = 0.4240 +Query 1/1: Action query time = 1.942 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5253 +t=42: Selected seed 195 with value = 0.5253 +Query 1/1: Action query time = 2.505 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5860 +t=58: Selected seed 195 with value = 0.5860 +Query 1/1: Action query time = 2.520 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7167 +t=74: Selected seed 195 with value = 0.7167 +Query 1/1: Action query time = 2.342 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8546 +t=90: Selected seed 195 with value = 0.8546 +Query 1/1: Action query time = 2.064 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=106: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 2.310 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6492 +t=122: Selected seed 195 with value = 0.6492 +Query 1/1: Action query time = 2.249 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6389 +t=138: Selected seed 195 with value = 0.6389 +Query 1/1: Action query time = 1.412 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6174 +t=154: Selected seed 195 with value = 0.6174 +Query 1/1: Action query time = 1.879 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6108 +t=170: Selected seed 195 with value = 0.6108 +Query 1/1: Action query time = 1.944 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6032 +t=186: Selected seed 195 with value = 0.6032 +Query 1/1: Action query time = 2.237 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6000 +t=202: Selected seed 195 with value = 0.6000 +Query 1/1: Action query time = 1.710 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6047 +t=218: Selected seed 195 with value = 0.6047 +Query 1/1: Action query time = 1.671 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6739 +t=234: Selected seed 195 with value = 0.6739 +Query 1/1: Action query time = 2.035 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7045 +t=250: Selected seed 195 with value = 0.7045 +Query 1/1: Action query time = 1.241 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7177 +t=266: Selected seed 195 with value = 0.7177 +Query 1/1: Action query time = 1.017 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7227 +t=282: Selected seed 195 with value = 0.7227 +Query 1/1: Action query time = 1.028 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7211 +t=298: Selected seed 195 with value = 0.7211 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=39--success=False--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=39--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 39 +# successes: 37 (94.9%) + +Task: open the middle drawer of the cabinet +Starting episode 40... +Query 1/1: Action query time = 2.212 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3730 +t=10: Selected seed 195 with value = 0.3730 +Query 1/1: Action query time = 1.584 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4254 +t=26: Selected seed 195 with value = 0.4254 +Query 1/1: Action query time = 1.473 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4825 +t=42: Selected seed 195 with value = 0.4825 +Query 1/1: Action query time = 1.634 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5691 +t=58: Selected seed 195 with value = 0.5691 +Query 1/1: Action query time = 1.433 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6697 +t=74: Selected seed 195 with value = 0.6697 +Query 1/1: Action query time = 1.453 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7854 +t=90: Selected seed 195 with value = 0.7854 +Query 1/1: Action query time = 1.482 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9354 +t=106: Selected seed 195 with value = 0.9354 +Query 1/1: Action query time = 1.573 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=40--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=40--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 40 +# successes: 38 (95.0%) + +Task: open the middle drawer of the cabinet +Starting episode 41... +Query 1/1: Action query time = 0.998 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3793 +t=10: Selected seed 195 with value = 0.3793 +Query 1/1: Action query time = 0.982 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4317 +t=26: Selected seed 195 with value = 0.4317 +Query 1/1: Action query time = 0.988 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5164 +t=42: Selected seed 195 with value = 0.5164 +Query 1/1: Action query time = 0.978 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6143 +t=58: Selected seed 195 with value = 0.6143 +Query 1/1: Action query time = 1.449 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7193 +t=74: Selected seed 195 with value = 0.7193 +Query 1/1: Action query time = 1.543 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8324 +t=90: Selected seed 195 with value = 0.8324 +Query 1/1: Action query time = 1.740 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9767 +t=106: Selected seed 195 with value = 0.9767 +Query 1/1: Action query time = 1.725 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=41--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=41--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 41 +# successes: 39 (95.1%) + +Task: open the middle drawer of the cabinet +Starting episode 42... +Query 1/1: Action query time = 1.988 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3584 +t=10: Selected seed 195 with value = 0.3584 +Query 1/1: Action query time = 1.617 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4396 +t=26: Selected seed 195 with value = 0.4396 +Query 1/1: Action query time = 1.459 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5108 +t=42: Selected seed 195 with value = 0.5108 +Query 1/1: Action query time = 1.432 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5946 +t=58: Selected seed 195 with value = 0.5946 +Query 1/1: Action query time = 0.980 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7103 +t=74: Selected seed 195 with value = 0.7103 +Query 1/1: Action query time = 0.991 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6833 +t=90: Selected seed 195 with value = 0.6833 +Query 1/1: Action query time = 0.990 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8049 +t=106: Selected seed 195 with value = 0.8049 +Query 1/1: Action query time = 0.989 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9951 +t=122: Selected seed 195 with value = 0.9951 +Query 1/1: Action query time = 1.397 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=42--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=42--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 42 +# successes: 40 (95.2%) + +Task: open the middle drawer of the cabinet +Starting episode 43... +Query 1/1: Action query time = 1.224 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3919 +t=10: Selected seed 195 with value = 0.3919 +Query 1/1: Action query time = 1.219 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4490 +t=26: Selected seed 195 with value = 0.4490 +Query 1/1: Action query time = 1.227 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5202 +t=42: Selected seed 195 with value = 0.5202 +Query 1/1: Action query time = 1.219 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6193 +t=58: Selected seed 195 with value = 0.6193 +Query 1/1: Action query time = 1.330 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6590 +t=74: Selected seed 195 with value = 0.6590 +Query 1/1: Action query time = 1.342 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7690 +t=90: Selected seed 195 with value = 0.7690 +Query 1/1: Action query time = 1.355 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9450 +t=106: Selected seed 195 with value = 0.9450 +Query 1/1: Action query time = 1.328 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=43--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=43--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 43 +# successes: 41 (95.3%) + +Task: open the middle drawer of the cabinet +Starting episode 44... +Query 1/1: Action query time = 1.389 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3512 +t=10: Selected seed 195 with value = 0.3512 +Query 1/1: Action query time = 1.439 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3208 +t=26: Selected seed 195 with value = 0.3208 +Query 1/1: Action query time = 1.483 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5030 +t=42: Selected seed 195 with value = 0.5030 +Query 1/1: Action query time = 1.467 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5810 +t=58: Selected seed 195 with value = 0.5810 +Query 1/1: Action query time = 1.456 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6712 +t=74: Selected seed 195 with value = 0.6712 +Query 1/1: Action query time = 1.479 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7707 +t=90: Selected seed 195 with value = 0.7707 +Query 1/1: Action query time = 1.603 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8423 +t=106: Selected seed 195 with value = 0.8423 +Query 1/1: Action query time = 1.582 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8983 +t=122: Selected seed 195 with value = 0.8983 +Query 1/1: Action query time = 1.361 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9947 +t=138: Selected seed 195 with value = 0.9947 +Query 1/1: Action query time = 1.317 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=154: Selected seed 195 with value = 0.9980 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=44--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=44--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 44 +# successes: 42 (95.5%) + +Task: open the middle drawer of the cabinet +Starting episode 45... +Query 1/1: Action query time = 1.603 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3539 +t=10: Selected seed 195 with value = 0.3539 +Query 1/1: Action query time = 1.553 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4204 +t=26: Selected seed 195 with value = 0.4204 +Query 1/1: Action query time = 1.584 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5128 +t=42: Selected seed 195 with value = 0.5128 +Query 1/1: Action query time = 1.565 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5352 +t=58: Selected seed 195 with value = 0.5352 +Query 1/1: Action query time = 1.492 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6390 +t=74: Selected seed 195 with value = 0.6390 +Query 1/1: Action query time = 1.434 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6896 +t=90: Selected seed 195 with value = 0.6896 +Query 1/1: Action query time = 0.985 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8967 +t=106: Selected seed 195 with value = 0.8967 +Query 1/1: Action query time = 0.999 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.993 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=45--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=45--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 45 +# successes: 43 (95.6%) + +Task: open the middle drawer of the cabinet +Starting episode 46... +Query 1/1: Action query time = 1.334 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3653 +t=10: Selected seed 195 with value = 0.3653 +Query 1/1: Action query time = 1.306 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4023 +t=26: Selected seed 195 with value = 0.4023 +Query 1/1: Action query time = 1.330 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4157 +t=42: Selected seed 195 with value = 0.4157 +Query 1/1: Action query time = 1.319 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4995 +t=58: Selected seed 195 with value = 0.4995 +Query 1/1: Action query time = 1.317 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5525 +t=74: Selected seed 195 with value = 0.5525 +Query 1/1: Action query time = 1.306 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6885 +t=90: Selected seed 195 with value = 0.6885 +Query 1/1: Action query time = 1.301 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8660 +t=106: Selected seed 195 with value = 0.8660 +Query 1/1: Action query time = 1.223 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=122: Selected seed 195 with value = 0.9909 +Query 1/1: Action query time = 1.176 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=46--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=46--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 46 +# successes: 44 (95.7%) + +Task: open the middle drawer of the cabinet +Starting episode 47... +Query 1/1: Action query time = 0.999 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3860 +t=10: Selected seed 195 with value = 0.3860 +Query 1/1: Action query time = 0.989 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4386 +t=26: Selected seed 195 with value = 0.4386 +Query 1/1: Action query time = 0.989 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4956 +t=42: Selected seed 195 with value = 0.4956 +Query 1/1: Action query time = 1.313 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5662 +t=58: Selected seed 195 with value = 0.5662 +Query 1/1: Action query time = 1.323 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6949 +t=74: Selected seed 195 with value = 0.6949 +Query 1/1: Action query time = 1.329 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8128 +t=90: Selected seed 195 with value = 0.8128 +Query 1/1: Action query time = 1.348 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9561 +t=106: Selected seed 195 with value = 0.9561 +Query 1/1: Action query time = 1.355 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=47--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=47--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 47 +# successes: 45 (95.7%) + +Task: open the middle drawer of the cabinet +Starting episode 48... +Query 1/1: Action query time = 1.368 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3981 +t=10: Selected seed 195 with value = 0.3981 +Query 1/1: Action query time = 1.337 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4491 +t=26: Selected seed 195 with value = 0.4491 +Query 1/1: Action query time = 1.303 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5250 +t=42: Selected seed 195 with value = 0.5250 +Query 1/1: Action query time = 1.292 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6190 +t=58: Selected seed 195 with value = 0.6190 +Query 1/1: Action query time = 1.272 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6426 +t=74: Selected seed 195 with value = 0.6426 +Query 1/1: Action query time = 1.274 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7620 +t=90: Selected seed 195 with value = 0.7620 +Query 1/1: Action query time = 1.297 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9031 +t=106: Selected seed 195 with value = 0.9031 +Query 1/1: Action query time = 1.333 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=48--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=48--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 48 +# successes: 46 (95.8%) + +Task: open the middle drawer of the cabinet +Starting episode 49... +Query 1/1: Action query time = 1.002 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3722 +t=10: Selected seed 195 with value = 0.3722 +Query 1/1: Action query time = 0.981 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4042 +t=26: Selected seed 195 with value = 0.4042 +Query 1/1: Action query time = 0.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4628 +t=42: Selected seed 195 with value = 0.4628 +Query 1/1: Action query time = 0.986 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6017 +t=58: Selected seed 195 with value = 0.6017 +Query 1/1: Action query time = 0.991 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6791 +t=74: Selected seed 195 with value = 0.6791 +Query 1/1: Action query time = 1.274 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7876 +t=90: Selected seed 195 with value = 0.7876 +Query 1/1: Action query time = 1.274 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8842 +t=106: Selected seed 195 with value = 0.8842 +Query 1/1: Action query time = 1.276 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.272 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=49--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=49--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 49 +# successes: 47 (95.9%) + +Task: open the middle drawer of the cabinet +Starting episode 50... +Query 1/1: Action query time = 1.338 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3763 +t=10: Selected seed 195 with value = 0.3763 +Query 1/1: Action query time = 1.266 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4117 +t=26: Selected seed 195 with value = 0.4117 +Query 1/1: Action query time = 1.243 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4774 +t=42: Selected seed 195 with value = 0.4774 +Query 1/1: Action query time = 1.228 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5836 +t=58: Selected seed 195 with value = 0.5836 +Query 1/1: Action query time = 1.231 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7096 +t=74: Selected seed 195 with value = 0.7096 +Query 1/1: Action query time = 1.200 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8292 +t=90: Selected seed 195 with value = 0.8292 +Query 1/1: Action query time = 1.203 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9523 +t=106: Selected seed 195 with value = 0.9523 +Query 1/1: Action query time = 1.203 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--episode=50--success=True--task=open_the_middle_drawer_of_the_cabinet.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t0/2026_08_01-11_38_09--with_future_img--episode=50--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 50 +# successes: 48 (96.0%) +Current task success rate: 0.96 +Current total success rate: 0.96 +Final results: +Total episodes: 50 +Total successes: 48 +Overall success rate: 0.9600 (96.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-11_38_09--forget_cl_iter350_par_t1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-11_38_09--forget_cl_iter350_par_t1.txt new file mode 100644 index 0000000000000000000000000000000000000000..87e3020c80260d7c839235591c69462bbeeec213 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-11_38_09--forget_cl_iter350_par_t1.txt @@ -0,0 +1,1720 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_from100_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='forget_cl_iter350_par_t1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=True, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 3.213 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4866 +t=10: Selected seed 195 with value = 0.4866 +Query 1/1: Action query time = 2.399 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6165 +t=26: Selected seed 195 with value = 0.6165 +Query 1/1: Action query time = 2.217 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7374 +t=42: Selected seed 195 with value = 0.7374 +Query 1/1: Action query time = 2.321 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8417 +t=58: Selected seed 195 with value = 0.8417 +Query 1/1: Action query time = 2.613 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9798 +t=74: Selected seed 195 with value = 0.9798 +Query 1/1: Action query time = 2.018 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=90: Selected seed 195 with value = 0.9983 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 2.377 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4975 +t=10: Selected seed 195 with value = 0.4975 +Query 1/1: Action query time = 1.655 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5835 +t=26: Selected seed 195 with value = 0.5835 +Query 1/1: Action query time = 2.468 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7092 +t=42: Selected seed 195 with value = 0.7092 +Query 1/1: Action query time = 3.285 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7995 +t=58: Selected seed 195 with value = 0.7995 +Query 1/1: Action query time = 2.749 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9546 +t=74: Selected seed 195 with value = 0.9546 +Query 1/1: Action query time = 2.290 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=90: Selected seed 195 with value = 0.9999 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 2.745 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5120 +t=10: Selected seed 195 with value = 0.5120 +Query 1/1: Action query time = 1.690 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6267 +t=26: Selected seed 195 with value = 0.6267 +Query 1/1: Action query time = 1.780 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7262 +t=42: Selected seed 195 with value = 0.7262 +Query 1/1: Action query time = 1.919 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8637 +t=58: Selected seed 195 with value = 0.8637 +Query 1/1: Action query time = 1.767 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9855 +t=74: Selected seed 195 with value = 0.9855 +Query 1/1: Action query time = 1.453 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.744 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.397 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.190 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.323 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.793 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.078 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.514 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.228 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.300 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.232 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.519 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.456 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.176 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 2.255 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5165 +t=10: Selected seed 195 with value = 0.5165 +Query 1/1: Action query time = 2.713 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6120 +t=26: Selected seed 195 with value = 0.6120 +Query 1/1: Action query time = 2.742 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6882 +t=42: Selected seed 195 with value = 0.6882 +Query 1/1: Action query time = 2.546 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8353 +t=58: Selected seed 195 with value = 0.8353 +Query 1/1: Action query time = 2.312 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9680 +t=74: Selected seed 195 with value = 0.9680 +Query 1/1: Action query time = 1.403 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=4--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=4--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 3 (75.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 3.039 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5031 +t=10: Selected seed 195 with value = 0.5031 +Query 1/1: Action query time = 1.885 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5942 +t=26: Selected seed 195 with value = 0.5942 +Query 1/1: Action query time = 1.716 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7110 +t=42: Selected seed 195 with value = 0.7110 +Query 1/1: Action query time = 2.389 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8142 +t=58: Selected seed 195 with value = 0.8142 +Query 1/1: Action query time = 2.301 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9153 +t=74: Selected seed 195 with value = 0.9153 +Query 1/1: Action query time = 2.452 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=90: Selected seed 195 with value = 0.9922 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=5--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=5--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 4 (80.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 1.795 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4921 +t=10: Selected seed 195 with value = 0.4921 +Query 1/1: Action query time = 2.387 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5836 +t=26: Selected seed 195 with value = 0.5836 +Query 1/1: Action query time = 2.580 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7277 +t=42: Selected seed 195 with value = 0.7277 +Query 1/1: Action query time = 2.392 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7810 +t=58: Selected seed 195 with value = 0.7810 +Query 1/1: Action query time = 1.915 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9439 +t=74: Selected seed 195 with value = 0.9439 +Query 1/1: Action query time = 1.975 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=90: Selected seed 195 with value = 0.9977 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=6--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=6--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 5 (83.3%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 2.700 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4529 +t=10: Selected seed 195 with value = 0.4529 +Query 1/1: Action query time = 2.514 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5435 +t=26: Selected seed 195 with value = 0.5435 +Query 1/1: Action query time = 2.530 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6875 +t=42: Selected seed 195 with value = 0.6875 +Query 1/1: Action query time = 2.231 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7720 +t=58: Selected seed 195 with value = 0.7720 +Query 1/1: Action query time = 2.758 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9145 +t=74: Selected seed 195 with value = 0.9145 +Query 1/1: Action query time = 2.447 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=7--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=7--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 6 (85.7%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 2.347 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4971 +t=10: Selected seed 195 with value = 0.4971 +Query 1/1: Action query time = 2.450 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6003 +t=26: Selected seed 195 with value = 0.6003 +Query 1/1: Action query time = 2.526 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7054 +t=42: Selected seed 195 with value = 0.7054 +Query 1/1: Action query time = 2.172 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8273 +t=58: Selected seed 195 with value = 0.8273 +Query 1/1: Action query time = 2.205 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9563 +t=74: Selected seed 195 with value = 0.9563 +Query 1/1: Action query time = 2.077 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9890 +t=90: Selected seed 195 with value = 0.9890 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=8--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=8--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 7 (87.5%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 1.247 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5082 +t=10: Selected seed 195 with value = 0.5082 +Query 1/1: Action query time = 1.482 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5825 +t=26: Selected seed 195 with value = 0.5825 +Query 1/1: Action query time = 1.928 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7002 +t=42: Selected seed 195 with value = 0.7002 +Query 1/1: Action query time = 2.341 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7918 +t=58: Selected seed 195 with value = 0.7918 +Query 1/1: Action query time = 1.718 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9120 +t=74: Selected seed 195 with value = 0.9120 +Query 1/1: Action query time = 1.902 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=9--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=9--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 8 (88.9%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 2.837 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5206 +t=10: Selected seed 195 with value = 0.5206 +Query 1/1: Action query time = 2.461 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6075 +t=26: Selected seed 195 with value = 0.6075 +Query 1/1: Action query time = 2.262 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7364 +t=42: Selected seed 195 with value = 0.7364 +Query 1/1: Action query time = 1.654 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8334 +t=58: Selected seed 195 with value = 0.8334 +Query 1/1: Action query time = 1.875 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9446 +t=74: Selected seed 195 with value = 0.9446 +Query 1/1: Action query time = 2.016 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=90: Selected seed 195 with value = 0.9990 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=10--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=10--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 9 (90.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 2.770 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4792 +t=10: Selected seed 195 with value = 0.4792 +Query 1/1: Action query time = 2.346 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5965 +t=26: Selected seed 195 with value = 0.5965 +Query 1/1: Action query time = 2.405 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6660 +t=42: Selected seed 195 with value = 0.6660 +Query 1/1: Action query time = 2.336 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7908 +t=58: Selected seed 195 with value = 0.7908 +Query 1/1: Action query time = 2.025 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8976 +t=74: Selected seed 195 with value = 0.8976 +Query 1/1: Action query time = 1.261 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9852 +t=90: Selected seed 195 with value = 0.9852 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=11--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=11--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 10 (90.9%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 2.771 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5259 +t=10: Selected seed 195 with value = 0.5259 +Query 1/1: Action query time = 2.268 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5937 +t=26: Selected seed 195 with value = 0.5937 +Query 1/1: Action query time = 2.302 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7436 +t=42: Selected seed 195 with value = 0.7436 +Query 1/1: Action query time = 1.982 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8018 +t=58: Selected seed 195 with value = 0.8018 +Query 1/1: Action query time = 2.181 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9223 +t=74: Selected seed 195 with value = 0.9223 +Query 1/1: Action query time = 1.886 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9938 +t=90: Selected seed 195 with value = 0.9938 +Query 1/1: Action query time = 2.639 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=12--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=12--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 11 (91.7%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 2.226 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5281 +t=10: Selected seed 195 with value = 0.5281 +Query 1/1: Action query time = 1.818 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5589 +t=26: Selected seed 195 with value = 0.5589 +Query 1/1: Action query time = 2.471 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6808 +t=42: Selected seed 195 with value = 0.6808 +Query 1/1: Action query time = 1.632 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7842 +t=58: Selected seed 195 with value = 0.7842 +Query 1/1: Action query time = 2.025 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8996 +t=74: Selected seed 195 with value = 0.8996 +Query 1/1: Action query time = 2.353 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=90: Selected seed 195 with value = 0.9981 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=13--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=13--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 13 +# successes: 12 (92.3%) + +Task: put the bowl on the stove +Starting episode 14... +Query 1/1: Action query time = 2.195 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4928 +t=10: Selected seed 195 with value = 0.4928 +Query 1/1: Action query time = 2.210 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5957 +t=26: Selected seed 195 with value = 0.5957 +Query 1/1: Action query time = 2.315 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6816 +t=42: Selected seed 195 with value = 0.6816 +Query 1/1: Action query time = 1.925 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7461 +t=58: Selected seed 195 with value = 0.7461 +Query 1/1: Action query time = 1.171 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9027 +t=74: Selected seed 195 with value = 0.9027 +Query 1/1: Action query time = 1.220 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9897 +t=90: Selected seed 195 with value = 0.9897 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=14--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=14--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 14 +# successes: 13 (92.9%) + +Task: put the bowl on the stove +Starting episode 15... +Query 1/1: Action query time = 2.511 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5308 +t=10: Selected seed 195 with value = 0.5308 +Query 1/1: Action query time = 2.615 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6300 +t=26: Selected seed 195 with value = 0.6300 +Query 1/1: Action query time = 2.180 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7536 +t=42: Selected seed 195 with value = 0.7536 +Query 1/1: Action query time = 2.181 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8638 +t=58: Selected seed 195 with value = 0.8638 +Query 1/1: Action query time = 1.405 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9889 +t=74: Selected seed 195 with value = 0.9889 +Query 1/1: Action query time = 2.313 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=90: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=15--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=15--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 15 +# successes: 14 (93.3%) + +Task: put the bowl on the stove +Starting episode 16... +Query 1/1: Action query time = 2.329 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4503 +t=10: Selected seed 195 with value = 0.4503 +Query 1/1: Action query time = 1.659 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5555 +t=26: Selected seed 195 with value = 0.5555 +Query 1/1: Action query time = 1.588 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6817 +t=42: Selected seed 195 with value = 0.6817 +Query 1/1: Action query time = 1.775 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7748 +t=58: Selected seed 195 with value = 0.7748 +Query 1/1: Action query time = 1.316 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9336 +t=74: Selected seed 195 with value = 0.9336 +Query 1/1: Action query time = 2.395 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=16--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=16--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 16 +# successes: 15 (93.8%) + +Task: put the bowl on the stove +Starting episode 17... +Query 1/1: Action query time = 2.377 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4949 +t=10: Selected seed 195 with value = 0.4949 +Query 1/1: Action query time = 1.944 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6168 +t=26: Selected seed 195 with value = 0.6168 +Query 1/1: Action query time = 2.418 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7166 +t=42: Selected seed 195 with value = 0.7166 +Query 1/1: Action query time = 2.132 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8192 +t=58: Selected seed 195 with value = 0.8192 +Query 1/1: Action query time = 2.599 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9503 +t=74: Selected seed 195 with value = 0.9503 +Query 1/1: Action query time = 1.866 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=17--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=17--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 17 +# successes: 16 (94.1%) + +Task: put the bowl on the stove +Starting episode 18... +Query 1/1: Action query time = 2.783 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5027 +t=10: Selected seed 195 with value = 0.5027 +Query 1/1: Action query time = 2.129 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5992 +t=26: Selected seed 195 with value = 0.5992 +Query 1/1: Action query time = 1.966 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7048 +t=42: Selected seed 195 with value = 0.7048 +Query 1/1: Action query time = 2.070 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8127 +t=58: Selected seed 195 with value = 0.8127 +Query 1/1: Action query time = 2.784 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9017 +t=74: Selected seed 195 with value = 0.9017 +Query 1/1: Action query time = 2.240 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=18--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=18--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 18 +# successes: 17 (94.4%) + +Task: put the bowl on the stove +Starting episode 19... +Query 1/1: Action query time = 2.842 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5005 +t=10: Selected seed 195 with value = 0.5005 +Query 1/1: Action query time = 1.678 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6081 +t=26: Selected seed 195 with value = 0.6081 +Query 1/1: Action query time = 1.798 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6971 +t=42: Selected seed 195 with value = 0.6971 +Query 1/1: Action query time = 1.847 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8101 +t=58: Selected seed 195 with value = 0.8101 +Query 1/1: Action query time = 2.191 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9499 +t=74: Selected seed 195 with value = 0.9499 +Query 1/1: Action query time = 1.655 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=19--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=19--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 19 +# successes: 18 (94.7%) + +Task: put the bowl on the stove +Starting episode 20... +Query 1/1: Action query time = 2.504 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5044 +t=10: Selected seed 195 with value = 0.5044 +Query 1/1: Action query time = 2.460 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6013 +t=26: Selected seed 195 with value = 0.6013 +Query 1/1: Action query time = 2.880 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6935 +t=42: Selected seed 195 with value = 0.6935 +Query 1/1: Action query time = 2.460 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8281 +t=58: Selected seed 195 with value = 0.8281 +Query 1/1: Action query time = 2.188 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9590 +t=74: Selected seed 195 with value = 0.9590 +Query 1/1: Action query time = 2.452 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=20--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=20--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 20 +# successes: 19 (95.0%) + +Task: put the bowl on the stove +Starting episode 21... +Query 1/1: Action query time = 2.373 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4998 +t=10: Selected seed 195 with value = 0.4998 +Query 1/1: Action query time = 2.706 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5964 +t=26: Selected seed 195 with value = 0.5964 +Query 1/1: Action query time = 2.059 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7180 +t=42: Selected seed 195 with value = 0.7180 +Query 1/1: Action query time = 1.401 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8123 +t=58: Selected seed 195 with value = 0.8123 +Query 1/1: Action query time = 1.015 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9555 +t=74: Selected seed 195 with value = 0.9555 +Query 1/1: Action query time = 1.600 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9906 +t=90: Selected seed 195 with value = 0.9906 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=21--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=21--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 21 +# successes: 20 (95.2%) + +Task: put the bowl on the stove +Starting episode 22... +Query 1/1: Action query time = 1.898 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4683 +t=10: Selected seed 195 with value = 0.4683 +Query 1/1: Action query time = 2.188 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5641 +t=26: Selected seed 195 with value = 0.5641 +Query 1/1: Action query time = 2.550 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6671 +t=42: Selected seed 195 with value = 0.6671 +Query 1/1: Action query time = 2.785 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7469 +t=58: Selected seed 195 with value = 0.7469 +Query 1/1: Action query time = 1.730 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9303 +t=74: Selected seed 195 with value = 0.9303 +Query 1/1: Action query time = 1.936 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=22--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=22--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 22 +# successes: 21 (95.5%) + +Task: put the bowl on the stove +Starting episode 23... +Query 1/1: Action query time = 2.830 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4870 +t=10: Selected seed 195 with value = 0.4870 +Query 1/1: Action query time = 2.569 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5921 +t=26: Selected seed 195 with value = 0.5921 +Query 1/1: Action query time = 2.543 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7061 +t=42: Selected seed 195 with value = 0.7061 +Query 1/1: Action query time = 1.944 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7928 +t=58: Selected seed 195 with value = 0.7928 +Query 1/1: Action query time = 1.610 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9339 +t=74: Selected seed 195 with value = 0.9339 +Query 1/1: Action query time = 2.189 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=23--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=23--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 23 +# successes: 22 (95.7%) + +Task: put the bowl on the stove +Starting episode 24... +Query 1/1: Action query time = 2.758 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5171 +t=10: Selected seed 195 with value = 0.5171 +Query 1/1: Action query time = 2.839 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5966 +t=26: Selected seed 195 with value = 0.5966 +Query 1/1: Action query time = 2.184 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6932 +t=42: Selected seed 195 with value = 0.6932 +Query 1/1: Action query time = 1.697 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8029 +t=58: Selected seed 195 with value = 0.8029 +Query 1/1: Action query time = 2.140 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9466 +t=74: Selected seed 195 with value = 0.9466 +Query 1/1: Action query time = 3.438 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=90: Selected seed 195 with value = 0.9928 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=24--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=24--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 24 +# successes: 23 (95.8%) + +Task: put the bowl on the stove +Starting episode 25... +Query 1/1: Action query time = 3.014 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4934 +t=10: Selected seed 195 with value = 0.4934 +Query 1/1: Action query time = 2.545 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5778 +t=26: Selected seed 195 with value = 0.5778 +Query 1/1: Action query time = 2.404 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6804 +t=42: Selected seed 195 with value = 0.6804 +Query 1/1: Action query time = 1.674 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7757 +t=58: Selected seed 195 with value = 0.7757 +Query 1/1: Action query time = 1.355 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9384 +t=74: Selected seed 195 with value = 0.9384 +Query 1/1: Action query time = 1.348 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=90: Selected seed 195 with value = 0.9946 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=25--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=25--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 25 +# successes: 24 (96.0%) + +Task: put the bowl on the stove +Starting episode 26... +Query 1/1: Action query time = 2.520 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5047 +t=10: Selected seed 195 with value = 0.5047 +Query 1/1: Action query time = 2.551 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5805 +t=26: Selected seed 195 with value = 0.5805 +Query 1/1: Action query time = 1.997 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7180 +t=42: Selected seed 195 with value = 0.7180 +Query 1/1: Action query time = 2.260 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8286 +t=58: Selected seed 195 with value = 0.8286 +Query 1/1: Action query time = 2.607 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9722 +t=74: Selected seed 195 with value = 0.9722 +Query 1/1: Action query time = 2.512 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=26--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=26--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 26 +# successes: 25 (96.2%) + +Task: put the bowl on the stove +Starting episode 27... +Query 1/1: Action query time = 2.240 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5118 +t=10: Selected seed 195 with value = 0.5118 +Query 1/1: Action query time = 2.373 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6154 +t=26: Selected seed 195 with value = 0.6154 +Query 1/1: Action query time = 2.513 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7262 +t=42: Selected seed 195 with value = 0.7262 +Query 1/1: Action query time = 2.715 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8554 +t=58: Selected seed 195 with value = 0.8554 +Query 1/1: Action query time = 1.885 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9711 +t=74: Selected seed 195 with value = 0.9711 +Query 1/1: Action query time = 2.372 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=27--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=27--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 27 +# successes: 26 (96.3%) + +Task: put the bowl on the stove +Starting episode 28... +Query 1/1: Action query time = 2.432 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4903 +t=10: Selected seed 195 with value = 0.4903 +Query 1/1: Action query time = 1.454 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6188 +t=26: Selected seed 195 with value = 0.6188 +Query 1/1: Action query time = 1.483 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7236 +t=42: Selected seed 195 with value = 0.7236 +Query 1/1: Action query time = 2.442 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8138 +t=58: Selected seed 195 with value = 0.8138 +Query 1/1: Action query time = 1.834 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9429 +t=74: Selected seed 195 with value = 0.9429 +Query 1/1: Action query time = 2.462 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9829 +t=90: Selected seed 195 with value = 0.9829 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=28--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=28--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 28 +# successes: 27 (96.4%) + +Task: put the bowl on the stove +Starting episode 29... +Query 1/1: Action query time = 2.559 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5021 +t=10: Selected seed 195 with value = 0.5021 +Query 1/1: Action query time = 1.939 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6042 +t=26: Selected seed 195 with value = 0.6042 +Query 1/1: Action query time = 2.083 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7320 +t=42: Selected seed 195 with value = 0.7320 +Query 1/1: Action query time = 2.481 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8200 +t=58: Selected seed 195 with value = 0.8200 +Query 1/1: Action query time = 2.570 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9726 +t=74: Selected seed 195 with value = 0.9726 +Query 1/1: Action query time = 2.132 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=90: Selected seed 195 with value = 0.9985 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=29--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=29--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 29 +# successes: 28 (96.6%) + +Task: put the bowl on the stove +Starting episode 30... +Query 1/1: Action query time = 2.682 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5525 +t=10: Selected seed 195 with value = 0.5525 +Query 1/1: Action query time = 2.477 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6215 +t=26: Selected seed 195 with value = 0.6215 +Query 1/1: Action query time = 1.783 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7135 +t=42: Selected seed 195 with value = 0.7135 +Query 1/1: Action query time = 2.391 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7857 +t=58: Selected seed 195 with value = 0.7857 +Query 1/1: Action query time = 2.193 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9150 +t=74: Selected seed 195 with value = 0.9150 +Query 1/1: Action query time = 2.855 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9706 +t=90: Selected seed 195 with value = 0.9706 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=30--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=30--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 30 +# successes: 29 (96.7%) + +Task: put the bowl on the stove +Starting episode 31... +Query 1/1: Action query time = 1.887 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4774 +t=10: Selected seed 195 with value = 0.4774 +Query 1/1: Action query time = 2.131 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5729 +t=26: Selected seed 195 with value = 0.5729 +Query 1/1: Action query time = 3.043 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6936 +t=42: Selected seed 195 with value = 0.6936 +Query 1/1: Action query time = 2.865 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8021 +t=58: Selected seed 195 with value = 0.8021 +Query 1/1: Action query time = 3.042 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8984 +t=74: Selected seed 195 with value = 0.8984 +Query 1/1: Action query time = 2.899 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=31--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=31--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 31 +# successes: 30 (96.8%) + +Task: put the bowl on the stove +Starting episode 32... +Query 1/1: Action query time = 2.357 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4952 +t=10: Selected seed 195 with value = 0.4952 +Query 1/1: Action query time = 2.705 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5838 +t=26: Selected seed 195 with value = 0.5838 +Query 1/1: Action query time = 2.409 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6979 +t=42: Selected seed 195 with value = 0.6979 +Query 1/1: Action query time = 2.065 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7915 +t=58: Selected seed 195 with value = 0.7915 +Query 1/1: Action query time = 1.662 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9353 +t=74: Selected seed 195 with value = 0.9353 +Query 1/1: Action query time = 1.611 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=32--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=32--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 32 +# successes: 31 (96.9%) + +Task: put the bowl on the stove +Starting episode 33... +Query 1/1: Action query time = 3.110 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5220 +t=10: Selected seed 195 with value = 0.5220 +Query 1/1: Action query time = 2.874 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6036 +t=26: Selected seed 195 with value = 0.6036 +Query 1/1: Action query time = 1.964 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7048 +t=42: Selected seed 195 with value = 0.7048 +Query 1/1: Action query time = 2.103 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8181 +t=58: Selected seed 195 with value = 0.8181 +Query 1/1: Action query time = 1.630 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9398 +t=74: Selected seed 195 with value = 0.9398 +Query 1/1: Action query time = 1.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=33--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=33--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 33 +# successes: 32 (97.0%) + +Task: put the bowl on the stove +Starting episode 34... +Query 1/1: Action query time = 1.759 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4795 +t=10: Selected seed 195 with value = 0.4795 +Query 1/1: Action query time = 2.338 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5923 +t=26: Selected seed 195 with value = 0.5923 +Query 1/1: Action query time = 2.371 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7143 +t=42: Selected seed 195 with value = 0.7143 +Query 1/1: Action query time = 2.502 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8033 +t=58: Selected seed 195 with value = 0.8033 +Query 1/1: Action query time = 2.579 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9521 +t=74: Selected seed 195 with value = 0.9521 +Query 1/1: Action query time = 2.527 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9923 +t=90: Selected seed 195 with value = 0.9923 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=34--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=34--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 34 +# successes: 33 (97.1%) + +Task: put the bowl on the stove +Starting episode 35... +Query 1/1: Action query time = 1.269 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4929 +t=10: Selected seed 195 with value = 0.4929 +Query 1/1: Action query time = 1.683 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5993 +t=26: Selected seed 195 with value = 0.5993 +Query 1/1: Action query time = 2.108 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7095 +t=42: Selected seed 195 with value = 0.7095 +Query 1/1: Action query time = 2.023 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8199 +t=58: Selected seed 195 with value = 0.8199 +Query 1/1: Action query time = 1.448 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9410 +t=74: Selected seed 195 with value = 0.9410 +Query 1/1: Action query time = 1.692 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=35--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=35--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 35 +# successes: 34 (97.1%) + +Task: put the bowl on the stove +Starting episode 36... +Query 1/1: Action query time = 1.799 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5140 +t=10: Selected seed 195 with value = 0.5140 +Query 1/1: Action query time = 1.994 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6211 +t=26: Selected seed 195 with value = 0.6211 +Query 1/1: Action query time = 2.498 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7130 +t=42: Selected seed 195 with value = 0.7130 +Query 1/1: Action query time = 1.607 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8559 +t=58: Selected seed 195 with value = 0.8559 +Query 1/1: Action query time = 1.811 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9687 +t=74: Selected seed 195 with value = 0.9687 +Query 1/1: Action query time = 2.430 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=36--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=36--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 36 +# successes: 35 (97.2%) + +Task: put the bowl on the stove +Starting episode 37... +Query 1/1: Action query time = 1.834 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5116 +t=10: Selected seed 195 with value = 0.5116 +Query 1/1: Action query time = 1.891 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6301 +t=26: Selected seed 195 with value = 0.6301 +Query 1/1: Action query time = 2.348 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7483 +t=42: Selected seed 195 with value = 0.7483 +Query 1/1: Action query time = 2.428 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8539 +t=58: Selected seed 195 with value = 0.8539 +Query 1/1: Action query time = 2.678 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9831 +t=74: Selected seed 195 with value = 0.9831 +Query 1/1: Action query time = 1.989 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9996 +t=90: Selected seed 195 with value = 0.9996 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=37--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=37--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 37 +# successes: 36 (97.3%) + +Task: put the bowl on the stove +Starting episode 38... +Query 1/1: Action query time = 1.476 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5219 +t=10: Selected seed 195 with value = 0.5219 +Query 1/1: Action query time = 1.775 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6133 +t=26: Selected seed 195 with value = 0.6133 +Query 1/1: Action query time = 2.301 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7133 +t=42: Selected seed 195 with value = 0.7133 +Query 1/1: Action query time = 2.524 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8213 +t=58: Selected seed 195 with value = 0.8213 +Query 1/1: Action query time = 2.500 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9600 +t=74: Selected seed 195 with value = 0.9600 +Query 1/1: Action query time = 2.348 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=90: Selected seed 195 with value = 0.9955 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=38--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=38--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 38 +# successes: 37 (97.4%) + +Task: put the bowl on the stove +Starting episode 39... +Query 1/1: Action query time = 1.638 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5017 +t=10: Selected seed 195 with value = 0.5017 +Query 1/1: Action query time = 1.804 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5774 +t=26: Selected seed 195 with value = 0.5774 +Query 1/1: Action query time = 2.069 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6970 +t=42: Selected seed 195 with value = 0.6970 +Query 1/1: Action query time = 2.265 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7935 +t=58: Selected seed 195 with value = 0.7935 +Query 1/1: Action query time = 2.342 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9078 +t=74: Selected seed 195 with value = 0.9078 +Query 1/1: Action query time = 2.204 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9818 +t=90: Selected seed 195 with value = 0.9818 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=39--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=39--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 39 +# successes: 38 (97.4%) + +Task: put the bowl on the stove +Starting episode 40... +Query 1/1: Action query time = 1.944 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5405 +t=10: Selected seed 195 with value = 0.5405 +Query 1/1: Action query time = 2.816 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6010 +t=26: Selected seed 195 with value = 0.6010 +Query 1/1: Action query time = 2.486 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7116 +t=42: Selected seed 195 with value = 0.7116 +Query 1/1: Action query time = 1.900 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8275 +t=58: Selected seed 195 with value = 0.8275 +Query 1/1: Action query time = 1.750 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9517 +t=74: Selected seed 195 with value = 0.9517 +Query 1/1: Action query time = 2.993 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=90: Selected seed 195 with value = 0.9809 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=40--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=40--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 40 +# successes: 39 (97.5%) + +Task: put the bowl on the stove +Starting episode 41... +Query 1/1: Action query time = 2.520 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4933 +t=10: Selected seed 195 with value = 0.4933 +Query 1/1: Action query time = 2.612 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5693 +t=26: Selected seed 195 with value = 0.5693 +Query 1/1: Action query time = 2.401 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6723 +t=42: Selected seed 195 with value = 0.6723 +Query 1/1: Action query time = 2.444 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7688 +t=58: Selected seed 195 with value = 0.7688 +Query 1/1: Action query time = 2.455 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9016 +t=74: Selected seed 195 with value = 0.9016 +Query 1/1: Action query time = 1.296 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=41--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=41--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 41 +# successes: 40 (97.6%) + +Task: put the bowl on the stove +Starting episode 42... +Query 1/1: Action query time = 1.808 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4755 +t=10: Selected seed 195 with value = 0.4755 +Query 1/1: Action query time = 2.168 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6055 +t=26: Selected seed 195 with value = 0.6055 +Query 1/1: Action query time = 2.783 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6985 +t=42: Selected seed 195 with value = 0.6985 +Query 1/1: Action query time = 2.501 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7783 +t=58: Selected seed 195 with value = 0.7783 +Query 1/1: Action query time = 2.569 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9067 +t=74: Selected seed 195 with value = 0.9067 +Query 1/1: Action query time = 2.680 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=90: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 2.412 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.718 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.650 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.228 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.506 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.830 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.446 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.305 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.099 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.581 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.601 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.546 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9765 +t=282: Selected seed 195 with value = 0.9765 +Query 1/1: Action query time = 2.016 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9282 +t=298: Selected seed 195 with value = 0.9282 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=42--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=42--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 42 +# successes: 40 (95.2%) + +Task: put the bowl on the stove +Starting episode 43... +Query 1/1: Action query time = 3.222 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4993 +t=10: Selected seed 195 with value = 0.4993 +Query 1/1: Action query time = 2.987 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6097 +t=26: Selected seed 195 with value = 0.6097 +Query 1/1: Action query time = 2.386 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7102 +t=42: Selected seed 195 with value = 0.7102 +Query 1/1: Action query time = 2.321 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8400 +t=58: Selected seed 195 with value = 0.8400 +Query 1/1: Action query time = 2.190 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9722 +t=74: Selected seed 195 with value = 0.9722 +Query 1/1: Action query time = 2.484 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=43--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=43--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 43 +# successes: 41 (95.3%) + +Task: put the bowl on the stove +Starting episode 44... +Query 1/1: Action query time = 1.917 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5165 +t=10: Selected seed 195 with value = 0.5165 +Query 1/1: Action query time = 2.620 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5961 +t=26: Selected seed 195 with value = 0.5961 +Query 1/1: Action query time = 2.104 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7119 +t=42: Selected seed 195 with value = 0.7119 +Query 1/1: Action query time = 2.083 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8018 +t=58: Selected seed 195 with value = 0.8018 +Query 1/1: Action query time = 2.393 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9475 +t=74: Selected seed 195 with value = 0.9475 +Query 1/1: Action query time = 2.640 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=90: Selected seed 195 with value = 0.9994 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=44--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=44--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 44 +# successes: 42 (95.5%) + +Task: put the bowl on the stove +Starting episode 45... +Query 1/1: Action query time = 3.240 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5093 +t=10: Selected seed 195 with value = 0.5093 +Query 1/1: Action query time = 2.100 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5903 +t=26: Selected seed 195 with value = 0.5903 +Query 1/1: Action query time = 1.497 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7129 +t=42: Selected seed 195 with value = 0.7129 +Query 1/1: Action query time = 1.680 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7630 +t=58: Selected seed 195 with value = 0.7630 +Query 1/1: Action query time = 2.374 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9177 +t=74: Selected seed 195 with value = 0.9177 +Query 1/1: Action query time = 2.170 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9483 +t=90: Selected seed 195 with value = 0.9483 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=45--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=45--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 45 +# successes: 43 (95.6%) + +Task: put the bowl on the stove +Starting episode 46... +Query 1/1: Action query time = 2.415 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4948 +t=10: Selected seed 195 with value = 0.4948 +Query 1/1: Action query time = 2.352 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5751 +t=26: Selected seed 195 with value = 0.5751 +Query 1/1: Action query time = 2.317 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6816 +t=42: Selected seed 195 with value = 0.6816 +Query 1/1: Action query time = 2.425 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7424 +t=58: Selected seed 195 with value = 0.7424 +Query 1/1: Action query time = 2.313 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8461 +t=74: Selected seed 195 with value = 0.8461 +Query 1/1: Action query time = 2.429 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=46--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=46--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 46 +# successes: 44 (95.7%) + +Task: put the bowl on the stove +Starting episode 47... +Query 1/1: Action query time = 2.253 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4826 +t=10: Selected seed 195 with value = 0.4826 +Query 1/1: Action query time = 2.210 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5869 +t=26: Selected seed 195 with value = 0.5869 +Query 1/1: Action query time = 2.460 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6936 +t=42: Selected seed 195 with value = 0.6936 +Query 1/1: Action query time = 1.912 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7948 +t=58: Selected seed 195 with value = 0.7948 +Query 1/1: Action query time = 2.031 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9263 +t=74: Selected seed 195 with value = 0.9263 +Query 1/1: Action query time = 1.488 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=90: Selected seed 195 with value = 0.9961 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=47--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=47--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 47 +# successes: 45 (95.7%) + +Task: put the bowl on the stove +Starting episode 48... +Query 1/1: Action query time = 2.287 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5151 +t=10: Selected seed 195 with value = 0.5151 +Query 1/1: Action query time = 2.113 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6227 +t=26: Selected seed 195 with value = 0.6227 +Query 1/1: Action query time = 2.164 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7146 +t=42: Selected seed 195 with value = 0.7146 +Query 1/1: Action query time = 2.649 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8529 +t=58: Selected seed 195 with value = 0.8529 +Query 1/1: Action query time = 2.483 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9750 +t=74: Selected seed 195 with value = 0.9750 +Query 1/1: Action query time = 1.850 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=48--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=48--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 48 +# successes: 46 (95.8%) + +Task: put the bowl on the stove +Starting episode 49... +Query 1/1: Action query time = 3.221 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5109 +t=10: Selected seed 195 with value = 0.5109 +Query 1/1: Action query time = 2.340 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5900 +t=26: Selected seed 195 with value = 0.5900 +Query 1/1: Action query time = 2.162 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6996 +t=42: Selected seed 195 with value = 0.6996 +Query 1/1: Action query time = 1.464 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8264 +t=58: Selected seed 195 with value = 0.8264 +Query 1/1: Action query time = 1.253 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9639 +t=74: Selected seed 195 with value = 0.9639 +Query 1/1: Action query time = 1.405 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=49--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=49--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 49 +# successes: 47 (95.9%) + +Task: put the bowl on the stove +Starting episode 50... +Query 1/1: Action query time = 2.648 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5519 +t=10: Selected seed 195 with value = 0.5519 +Query 1/1: Action query time = 2.592 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6428 +t=26: Selected seed 195 with value = 0.6428 +Query 1/1: Action query time = 2.506 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7268 +t=42: Selected seed 195 with value = 0.7268 +Query 1/1: Action query time = 2.651 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8296 +t=58: Selected seed 195 with value = 0.8296 +Query 1/1: Action query time = 2.291 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9831 +t=74: Selected seed 195 with value = 0.9831 +Query 1/1: Action query time = 1.772 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9906 +t=90: Selected seed 195 with value = 0.9906 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--episode=50--success=True--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t1/2026_08_01-11_38_09--with_future_img--episode=50--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 50 +# successes: 48 (96.0%) +Current task success rate: 0.96 +Current total success rate: 0.96 +Final results: +Total episodes: 50 +Total successes: 48 +Overall success rate: 0.9600 (96.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-11_38_10--forget_cl_iter350_par_t6.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-11_38_10--forget_cl_iter350_par_t6.txt new file mode 100644 index 0000000000000000000000000000000000000000..bc4dde685db23497b2daf924e24ca773f330b496 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-11_38_10--forget_cl_iter350_par_t6.txt @@ -0,0 +1,2112 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_from100_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='forget_cl_iter350_par_t6', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=True, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 2.954 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4316 +t=10: Selected seed 195 with value = 0.4316 +Query 1/1: Action query time = 2.591 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5198 +t=26: Selected seed 195 with value = 0.5198 +Query 1/1: Action query time = 2.350 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6361 +t=42: Selected seed 195 with value = 0.6361 +Query 1/1: Action query time = 2.220 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5575 +t=58: Selected seed 195 with value = 0.5575 +Query 1/1: Action query time = 2.357 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6377 +t=74: Selected seed 195 with value = 0.6377 +Query 1/1: Action query time = 1.768 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6712 +t=90: Selected seed 195 with value = 0.6712 +Query 1/1: Action query time = 1.672 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7319 +t=106: Selected seed 195 with value = 0.7319 +Query 1/1: Action query time = 2.534 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7370 +t=122: Selected seed 195 with value = 0.7370 +Query 1/1: Action query time = 2.344 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8519 +t=138: Selected seed 195 with value = 0.8519 +Query 1/1: Action query time = 1.802 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7649 +t=154: Selected seed 195 with value = 0.7649 +Query 1/1: Action query time = 1.889 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8599 +t=170: Selected seed 195 with value = 0.8599 +Query 1/1: Action query time = 2.279 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9573 +t=186: Selected seed 195 with value = 0.9573 +Query 1/1: Action query time = 0.966 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7902 +t=202: Selected seed 195 with value = 0.7902 +Query 1/1: Action query time = 1.478 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9037 +t=218: Selected seed 195 with value = 0.9037 +Query 1/1: Action query time = 1.225 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8978 +t=234: Selected seed 195 with value = 0.8978 +Query 1/1: Action query time = 1.544 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9136 +t=250: Selected seed 195 with value = 0.9136 +Query 1/1: Action query time = 1.807 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9178 +t=266: Selected seed 195 with value = 0.9178 +Query 1/1: Action query time = 1.882 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9041 +t=282: Selected seed 195 with value = 0.9041 +Query 1/1: Action query time = 1.917 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8423 +t=298: Selected seed 195 with value = 0.8423 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 1.538 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4498 +t=10: Selected seed 195 with value = 0.4498 +Query 1/1: Action query time = 1.502 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5516 +t=26: Selected seed 195 with value = 0.5516 +Query 1/1: Action query time = 1.330 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6843 +t=42: Selected seed 195 with value = 0.6843 +Query 1/1: Action query time = 1.433 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7549 +t=58: Selected seed 195 with value = 0.7549 +Query 1/1: Action query time = 1.984 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9072 +t=74: Selected seed 195 with value = 0.9072 +Query 1/1: Action query time = 1.993 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9695 +t=90: Selected seed 195 with value = 0.9695 +Query 1/1: Action query time = 1.851 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9786 +t=106: Selected seed 195 with value = 0.9786 +Query 1/1: Action query time = 2.540 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9511 +t=122: Selected seed 195 with value = 0.9511 +Query 1/1: Action query time = 2.039 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9144 +t=138: Selected seed 195 with value = 0.9144 +Query 1/1: Action query time = 1.436 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9131 +t=154: Selected seed 195 with value = 0.9131 +Query 1/1: Action query time = 1.589 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9538 +t=170: Selected seed 195 with value = 0.9538 +Query 1/1: Action query time = 1.748 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.021 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.432 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.021 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.099 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.414 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.641 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9883 +t=282: Selected seed 195 with value = 0.9883 +Query 1/1: Action query time = 2.318 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9508 +t=298: Selected seed 195 with value = 0.9508 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=2--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=2--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 1.685 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4413 +t=10: Selected seed 195 with value = 0.4413 +Query 1/1: Action query time = 1.936 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4796 +t=26: Selected seed 195 with value = 0.4796 +Query 1/1: Action query time = 1.810 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6485 +t=42: Selected seed 195 with value = 0.6485 +Query 1/1: Action query time = 1.773 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7260 +t=58: Selected seed 195 with value = 0.7260 +Query 1/1: Action query time = 1.816 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8197 +t=74: Selected seed 195 with value = 0.8197 +Query 1/1: Action query time = 1.726 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9848 +t=90: Selected seed 195 with value = 0.9848 +Query 1/1: Action query time = 1.800 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9855 +t=106: Selected seed 195 with value = 0.9855 +Query 1/1: Action query time = 1.462 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9884 +t=122: Selected seed 195 with value = 0.9884 +Query 1/1: Action query time = 1.440 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9851 +t=138: Selected seed 195 with value = 0.9851 +Query 1/1: Action query time = 1.604 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9786 +t=154: Selected seed 195 with value = 0.9786 +Query 1/1: Action query time = 1.533 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=170: Selected seed 195 with value = 0.9809 +Query 1/1: Action query time = 1.592 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9818 +t=186: Selected seed 195 with value = 0.9818 +Query 1/1: Action query time = 1.660 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9811 +t=202: Selected seed 195 with value = 0.9811 +Query 1/1: Action query time = 1.539 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9798 +t=218: Selected seed 195 with value = 0.9798 +Query 1/1: Action query time = 2.225 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9711 +t=234: Selected seed 195 with value = 0.9711 +Query 1/1: Action query time = 1.765 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9613 +t=250: Selected seed 195 with value = 0.9613 +Query 1/1: Action query time = 1.816 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9375 +t=266: Selected seed 195 with value = 0.9375 +Query 1/1: Action query time = 1.905 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9317 +t=282: Selected seed 195 with value = 0.9317 +Query 1/1: Action query time = 1.667 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9416 +t=298: Selected seed 195 with value = 0.9416 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 4... +Query 1/1: Action query time = 1.981 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4089 +t=10: Selected seed 195 with value = 0.4089 +Query 1/1: Action query time = 1.947 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5082 +t=26: Selected seed 195 with value = 0.5082 +Query 1/1: Action query time = 2.134 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5600 +t=42: Selected seed 195 with value = 0.5600 +Query 1/1: Action query time = 1.818 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6985 +t=58: Selected seed 195 with value = 0.6985 +Query 1/1: Action query time = 1.630 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8024 +t=74: Selected seed 195 with value = 0.8024 +Query 1/1: Action query time = 1.670 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9423 +t=90: Selected seed 195 with value = 0.9423 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=4--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=4--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 4 +# successes: 1 (25.0%) + +Task: put the cream cheese in the bowl +Starting episode 5... +Query 1/1: Action query time = 1.565 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4244 +t=10: Selected seed 195 with value = 0.4244 +Query 1/1: Action query time = 1.676 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5124 +t=26: Selected seed 195 with value = 0.5124 +Query 1/1: Action query time = 2.045 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5835 +t=42: Selected seed 195 with value = 0.5835 +Query 1/1: Action query time = 1.390 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6841 +t=58: Selected seed 195 with value = 0.6841 +Query 1/1: Action query time = 1.459 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8192 +t=74: Selected seed 195 with value = 0.8192 +Query 1/1: Action query time = 2.068 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9346 +t=90: Selected seed 195 with value = 0.9346 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=5--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=5--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 5 +# successes: 2 (40.0%) + +Task: put the cream cheese in the bowl +Starting episode 6... +Query 1/1: Action query time = 2.350 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4730 +t=10: Selected seed 195 with value = 0.4730 +Query 1/1: Action query time = 2.006 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5070 +t=26: Selected seed 195 with value = 0.5070 +Query 1/1: Action query time = 1.728 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6006 +t=42: Selected seed 195 with value = 0.6006 +Query 1/1: Action query time = 0.992 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5887 +t=58: Selected seed 195 with value = 0.5887 +Query 1/1: Action query time = 1.427 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6255 +t=74: Selected seed 195 with value = 0.6255 +Query 1/1: Action query time = 1.740 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6753 +t=90: Selected seed 195 with value = 0.6753 +Query 1/1: Action query time = 1.826 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8110 +t=106: Selected seed 195 with value = 0.8110 +Query 1/1: Action query time = 1.708 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9260 +t=122: Selected seed 195 with value = 0.9260 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=6--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=6--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 6 +# successes: 3 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 7... +Query 1/1: Action query time = 2.020 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4915 +t=10: Selected seed 195 with value = 0.4915 +Query 1/1: Action query time = 1.108 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5675 +t=26: Selected seed 195 with value = 0.5675 +Query 1/1: Action query time = 1.371 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6082 +t=42: Selected seed 195 with value = 0.6082 +Query 1/1: Action query time = 1.802 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7582 +t=58: Selected seed 195 with value = 0.7582 +Query 1/1: Action query time = 2.193 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8947 +t=74: Selected seed 195 with value = 0.8947 +Query 1/1: Action query time = 2.422 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=90: Selected seed 195 with value = 0.9981 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=7--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=7--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 7 +# successes: 4 (57.1%) + +Task: put the cream cheese in the bowl +Starting episode 8... +Query 1/1: Action query time = 1.886 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4231 +t=10: Selected seed 195 with value = 0.4231 +Query 1/1: Action query time = 1.781 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5341 +t=26: Selected seed 195 with value = 0.5341 +Query 1/1: Action query time = 1.863 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6273 +t=42: Selected seed 195 with value = 0.6273 +Query 1/1: Action query time = 1.644 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7036 +t=58: Selected seed 195 with value = 0.7036 +Query 1/1: Action query time = 1.562 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8317 +t=74: Selected seed 195 with value = 0.8317 +Query 1/1: Action query time = 1.200 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=90: Selected seed 195 with value = 0.9914 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=8--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=8--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 8 +# successes: 5 (62.5%) + +Task: put the cream cheese in the bowl +Starting episode 9... +Query 1/1: Action query time = 2.003 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4576 +t=10: Selected seed 195 with value = 0.4576 +Query 1/1: Action query time = 2.098 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5483 +t=26: Selected seed 195 with value = 0.5483 +Query 1/1: Action query time = 1.660 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6327 +t=42: Selected seed 195 with value = 0.6327 +Query 1/1: Action query time = 1.486 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7285 +t=58: Selected seed 195 with value = 0.7285 +Query 1/1: Action query time = 1.787 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8395 +t=74: Selected seed 195 with value = 0.8395 +Query 1/1: Action query time = 1.699 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=90: Selected seed 195 with value = 0.9963 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=9--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=9--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 9 +# successes: 6 (66.7%) + +Task: put the cream cheese in the bowl +Starting episode 10... +Query 1/1: Action query time = 0.982 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4367 +t=10: Selected seed 195 with value = 0.4367 +Query 1/1: Action query time = 0.945 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5234 +t=26: Selected seed 195 with value = 0.5234 +Query 1/1: Action query time = 1.308 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5765 +t=42: Selected seed 195 with value = 0.5765 +Query 1/1: Action query time = 1.541 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6989 +t=58: Selected seed 195 with value = 0.6989 +Query 1/1: Action query time = 1.761 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8170 +t=74: Selected seed 195 with value = 0.8170 +Query 1/1: Action query time = 1.989 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9801 +t=90: Selected seed 195 with value = 0.9801 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=10--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=10--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 10 +# successes: 7 (70.0%) + +Task: put the cream cheese in the bowl +Starting episode 11... +Query 1/1: Action query time = 1.689 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4444 +t=10: Selected seed 195 with value = 0.4444 +Query 1/1: Action query time = 1.676 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5036 +t=26: Selected seed 195 with value = 0.5036 +Query 1/1: Action query time = 1.574 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5931 +t=42: Selected seed 195 with value = 0.5931 +Query 1/1: Action query time = 1.649 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6948 +t=58: Selected seed 195 with value = 0.6948 +Query 1/1: Action query time = 1.324 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8228 +t=74: Selected seed 195 with value = 0.8228 +Query 1/1: Action query time = 1.291 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9667 +t=90: Selected seed 195 with value = 0.9667 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=11--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=11--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 11 +# successes: 8 (72.7%) + +Task: put the cream cheese in the bowl +Starting episode 12... +Query 1/1: Action query time = 1.738 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4912 +t=10: Selected seed 195 with value = 0.4912 +Query 1/1: Action query time = 1.744 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5245 +t=26: Selected seed 195 with value = 0.5245 +Query 1/1: Action query time = 1.344 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6322 +t=42: Selected seed 195 with value = 0.6322 +Query 1/1: Action query time = 1.450 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7209 +t=58: Selected seed 195 with value = 0.7209 +Query 1/1: Action query time = 1.469 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8416 +t=74: Selected seed 195 with value = 0.8416 +Query 1/1: Action query time = 1.661 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9918 +t=90: Selected seed 195 with value = 0.9918 +Query 1/1: Action query time = 1.712 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=106: Selected seed 195 with value = 0.9997 +Query 1/1: Action query time = 1.471 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=122: Selected seed 195 with value = 0.9980 +Query 1/1: Action query time = 1.323 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=138: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 2.108 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=154: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 1.472 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=170: Selected seed 195 with value = 0.9909 +Query 1/1: Action query time = 1.400 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9793 +t=186: Selected seed 195 with value = 0.9793 +Query 1/1: Action query time = 2.203 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9701 +t=202: Selected seed 195 with value = 0.9701 +Query 1/1: Action query time = 2.107 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9345 +t=218: Selected seed 195 with value = 0.9345 +Query 1/1: Action query time = 2.099 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7838 +t=234: Selected seed 195 with value = 0.7838 +Query 1/1: Action query time = 2.253 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9004 +t=250: Selected seed 195 with value = 0.9004 +Query 1/1: Action query time = 1.745 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9113 +t=266: Selected seed 195 with value = 0.9113 +Query 1/1: Action query time = 1.566 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9102 +t=282: Selected seed 195 with value = 0.9102 +Query 1/1: Action query time = 1.177 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9107 +t=298: Selected seed 195 with value = 0.9107 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=12--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=12--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 12 +# successes: 8 (66.7%) + +Task: put the cream cheese in the bowl +Starting episode 13... +Query 1/1: Action query time = 1.850 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4621 +t=10: Selected seed 195 with value = 0.4621 +Query 1/1: Action query time = 1.662 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5277 +t=26: Selected seed 195 with value = 0.5277 +Query 1/1: Action query time = 1.788 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6208 +t=42: Selected seed 195 with value = 0.6208 +Query 1/1: Action query time = 1.388 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7290 +t=58: Selected seed 195 with value = 0.7290 +Query 1/1: Action query time = 1.439 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8375 +t=74: Selected seed 195 with value = 0.8375 +Query 1/1: Action query time = 1.311 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=90: Selected seed 195 with value = 0.9867 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=13--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=13--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 13 +# successes: 9 (69.2%) + +Task: put the cream cheese in the bowl +Starting episode 14... +Query 1/1: Action query time = 2.189 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4369 +t=10: Selected seed 195 with value = 0.4369 +Query 1/1: Action query time = 2.017 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5384 +t=26: Selected seed 195 with value = 0.5384 +Query 1/1: Action query time = 1.378 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6643 +t=42: Selected seed 195 with value = 0.6643 +Query 1/1: Action query time = 1.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7586 +t=58: Selected seed 195 with value = 0.7586 +Query 1/1: Action query time = 1.319 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9018 +t=74: Selected seed 195 with value = 0.9018 +Query 1/1: Action query time = 1.636 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=90: Selected seed 195 with value = 0.9993 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=14--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=14--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 14 +# successes: 10 (71.4%) + +Task: put the cream cheese in the bowl +Starting episode 15... +Query 1/1: Action query time = 1.803 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4546 +t=10: Selected seed 195 with value = 0.4546 +Query 1/1: Action query time = 1.396 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5898 +t=26: Selected seed 195 with value = 0.5898 +Query 1/1: Action query time = 1.870 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6645 +t=42: Selected seed 195 with value = 0.6645 +Query 1/1: Action query time = 1.504 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7545 +t=58: Selected seed 195 with value = 0.7545 +Query 1/1: Action query time = 1.567 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8992 +t=74: Selected seed 195 with value = 0.8992 +Query 1/1: Action query time = 1.937 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9781 +t=90: Selected seed 195 with value = 0.9781 +Query 1/1: Action query time = 1.913 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9632 +t=106: Selected seed 195 with value = 0.9632 +Query 1/1: Action query time = 1.580 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9820 +t=122: Selected seed 195 with value = 0.9820 +Query 1/1: Action query time = 1.776 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9733 +t=138: Selected seed 195 with value = 0.9733 +Query 1/1: Action query time = 2.071 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9497 +t=154: Selected seed 195 with value = 0.9497 +Query 1/1: Action query time = 2.348 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9323 +t=170: Selected seed 195 with value = 0.9323 +Query 1/1: Action query time = 1.716 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9267 +t=186: Selected seed 195 with value = 0.9267 +Query 1/1: Action query time = 1.021 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9189 +t=202: Selected seed 195 with value = 0.9189 +Query 1/1: Action query time = 1.460 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9053 +t=218: Selected seed 195 with value = 0.9053 +Query 1/1: Action query time = 2.212 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8931 +t=234: Selected seed 195 with value = 0.8931 +Query 1/1: Action query time = 2.012 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8911 +t=250: Selected seed 195 with value = 0.8911 +Query 1/1: Action query time = 1.968 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8921 +t=266: Selected seed 195 with value = 0.8921 +Query 1/1: Action query time = 1.946 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8799 +t=282: Selected seed 195 with value = 0.8799 +Query 1/1: Action query time = 2.234 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9865 +t=298: Selected seed 195 with value = 0.9865 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=15--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=15--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 15 +# successes: 10 (66.7%) + +Task: put the cream cheese in the bowl +Starting episode 16... +Query 1/1: Action query time = 2.052 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4372 +t=10: Selected seed 195 with value = 0.4372 +Query 1/1: Action query time = 1.727 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4914 +t=26: Selected seed 195 with value = 0.4914 +Query 1/1: Action query time = 1.639 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5927 +t=42: Selected seed 195 with value = 0.5927 +Query 1/1: Action query time = 1.507 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7042 +t=58: Selected seed 195 with value = 0.7042 +Query 1/1: Action query time = 1.322 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8300 +t=74: Selected seed 195 with value = 0.8300 +Query 1/1: Action query time = 1.744 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9636 +t=90: Selected seed 195 with value = 0.9636 +Query 1/1: Action query time = 1.900 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=16--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=16--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 16 +# successes: 11 (68.8%) + +Task: put the cream cheese in the bowl +Starting episode 17... +Query 1/1: Action query time = 1.569 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4597 +t=10: Selected seed 195 with value = 0.4597 +Query 1/1: Action query time = 1.636 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5605 +t=26: Selected seed 195 with value = 0.5605 +Query 1/1: Action query time = 1.693 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5945 +t=42: Selected seed 195 with value = 0.5945 +Query 1/1: Action query time = 2.217 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7378 +t=58: Selected seed 195 with value = 0.7378 +Query 1/1: Action query time = 2.237 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8842 +t=74: Selected seed 195 with value = 0.8842 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=17--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=17--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 17 +# successes: 12 (70.6%) + +Task: put the cream cheese in the bowl +Starting episode 18... +Query 1/1: Action query time = 2.266 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4438 +t=10: Selected seed 195 with value = 0.4438 +Query 1/1: Action query time = 1.338 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5168 +t=26: Selected seed 195 with value = 0.5168 +Query 1/1: Action query time = 0.970 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6212 +t=42: Selected seed 195 with value = 0.6212 +Query 1/1: Action query time = 1.018 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6680 +t=58: Selected seed 195 with value = 0.6680 +Query 1/1: Action query time = 1.470 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8333 +t=74: Selected seed 195 with value = 0.8333 +Query 1/1: Action query time = 1.837 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9486 +t=90: Selected seed 195 with value = 0.9486 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=18--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=18--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 18 +# successes: 13 (72.2%) + +Task: put the cream cheese in the bowl +Starting episode 19... +Query 1/1: Action query time = 1.737 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4737 +t=10: Selected seed 195 with value = 0.4737 +Query 1/1: Action query time = 1.729 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5255 +t=26: Selected seed 195 with value = 0.5255 +Query 1/1: Action query time = 1.409 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5739 +t=42: Selected seed 195 with value = 0.5739 +Query 1/1: Action query time = 1.377 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7266 +t=58: Selected seed 195 with value = 0.7266 +Query 1/1: Action query time = 1.507 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8620 +t=74: Selected seed 195 with value = 0.8620 +Query 1/1: Action query time = 1.657 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9931 +t=90: Selected seed 195 with value = 0.9931 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=19--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=19--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 19 +# successes: 14 (73.7%) + +Task: put the cream cheese in the bowl +Starting episode 20... +Query 1/1: Action query time = 2.590 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4482 +t=10: Selected seed 195 with value = 0.4482 +Query 1/1: Action query time = 1.779 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4935 +t=26: Selected seed 195 with value = 0.4935 +Query 1/1: Action query time = 1.507 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5706 +t=42: Selected seed 195 with value = 0.5706 +Query 1/1: Action query time = 2.355 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6741 +t=58: Selected seed 195 with value = 0.6741 +Query 1/1: Action query time = 2.324 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8005 +t=74: Selected seed 195 with value = 0.8005 +Query 1/1: Action query time = 2.458 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9336 +t=90: Selected seed 195 with value = 0.9336 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=20--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=20--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 20 +# successes: 15 (75.0%) + +Task: put the cream cheese in the bowl +Starting episode 21... +Query 1/1: Action query time = 1.642 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4525 +t=10: Selected seed 195 with value = 0.4525 +Query 1/1: Action query time = 1.219 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5136 +t=26: Selected seed 195 with value = 0.5136 +Query 1/1: Action query time = 2.016 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5723 +t=42: Selected seed 195 with value = 0.5723 +Query 1/1: Action query time = 1.786 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6783 +t=58: Selected seed 195 with value = 0.6783 +Query 1/1: Action query time = 2.029 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8281 +t=74: Selected seed 195 with value = 0.8281 +Query 1/1: Action query time = 1.411 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9935 +t=90: Selected seed 195 with value = 0.9935 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=21--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=21--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 21 +# successes: 16 (76.2%) + +Task: put the cream cheese in the bowl +Starting episode 22... +Query 1/1: Action query time = 1.456 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4620 +t=10: Selected seed 195 with value = 0.4620 +Query 1/1: Action query time = 1.204 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4190 +t=26: Selected seed 195 with value = 0.4190 +Query 1/1: Action query time = 1.275 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5617 +t=42: Selected seed 195 with value = 0.5617 +Query 1/1: Action query time = 1.265 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6550 +t=58: Selected seed 195 with value = 0.6550 +Query 1/1: Action query time = 1.297 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7894 +t=74: Selected seed 195 with value = 0.7894 +Query 1/1: Action query time = 1.427 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9927 +t=90: Selected seed 195 with value = 0.9927 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=22--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=22--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 22 +# successes: 17 (77.3%) + +Task: put the cream cheese in the bowl +Starting episode 23... +Query 1/1: Action query time = 1.630 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4793 +t=10: Selected seed 195 with value = 0.4793 +Query 1/1: Action query time = 1.236 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4885 +t=26: Selected seed 195 with value = 0.4885 +Query 1/1: Action query time = 1.731 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5519 +t=42: Selected seed 195 with value = 0.5519 +Query 1/1: Action query time = 1.738 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6730 +t=58: Selected seed 195 with value = 0.6730 +Query 1/1: Action query time = 1.823 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8216 +t=74: Selected seed 195 with value = 0.8216 +Query 1/1: Action query time = 1.693 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9816 +t=90: Selected seed 195 with value = 0.9816 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=23--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=23--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 23 +# successes: 18 (78.3%) + +Task: put the cream cheese in the bowl +Starting episode 24... +Query 1/1: Action query time = 1.359 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4618 +t=10: Selected seed 195 with value = 0.4618 +Query 1/1: Action query time = 1.719 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3572 +t=26: Selected seed 195 with value = 0.3572 +Query 1/1: Action query time = 2.072 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=42: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 2.254 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6639 +t=58: Selected seed 195 with value = 0.6639 +Query 1/1: Action query time = 2.111 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7933 +t=74: Selected seed 195 with value = 0.7933 +Query 1/1: Action query time = 1.923 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9419 +t=90: Selected seed 195 with value = 0.9419 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=24--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=24--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 24 +# successes: 19 (79.2%) + +Task: put the cream cheese in the bowl +Starting episode 25... +Query 1/1: Action query time = 1.017 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4450 +t=10: Selected seed 195 with value = 0.4450 +Query 1/1: Action query time = 1.258 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5374 +t=26: Selected seed 195 with value = 0.5374 +Query 1/1: Action query time = 1.828 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6155 +t=42: Selected seed 195 with value = 0.6155 +Query 1/1: Action query time = 2.653 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7234 +t=58: Selected seed 195 with value = 0.7234 +Query 1/1: Action query time = 2.265 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8534 +t=74: Selected seed 195 with value = 0.8534 +Query 1/1: Action query time = 1.992 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9898 +t=90: Selected seed 195 with value = 0.9898 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=25--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=25--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 25 +# successes: 20 (80.0%) + +Task: put the cream cheese in the bowl +Starting episode 26... +Query 1/1: Action query time = 1.224 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4250 +t=10: Selected seed 195 with value = 0.4250 +Query 1/1: Action query time = 1.787 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4392 +t=26: Selected seed 195 with value = 0.4392 +Query 1/1: Action query time = 1.801 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5954 +t=42: Selected seed 195 with value = 0.5954 +Query 1/1: Action query time = 1.774 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7250 +t=58: Selected seed 195 with value = 0.7250 +Query 1/1: Action query time = 2.673 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8345 +t=74: Selected seed 195 with value = 0.8345 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=26--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=26--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 26 +# successes: 21 (80.8%) + +Task: put the cream cheese in the bowl +Starting episode 27... +Query 1/1: Action query time = 2.882 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4421 +t=10: Selected seed 195 with value = 0.4421 +Query 1/1: Action query time = 1.486 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5256 +t=26: Selected seed 195 with value = 0.5256 +Query 1/1: Action query time = 1.492 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6047 +t=42: Selected seed 195 with value = 0.6047 +Query 1/1: Action query time = 1.818 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7160 +t=58: Selected seed 195 with value = 0.7160 +Query 1/1: Action query time = 1.866 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8228 +t=74: Selected seed 195 with value = 0.8228 +Query 1/1: Action query time = 1.747 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9792 +t=90: Selected seed 195 with value = 0.9792 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=27--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=27--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 27 +# successes: 22 (81.5%) + +Task: put the cream cheese in the bowl +Starting episode 28... +Query 1/1: Action query time = 1.111 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4452 +t=10: Selected seed 195 with value = 0.4452 +Query 1/1: Action query time = 2.050 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5582 +t=26: Selected seed 195 with value = 0.5582 +Query 1/1: Action query time = 2.555 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6205 +t=42: Selected seed 195 with value = 0.6205 +Query 1/1: Action query time = 2.119 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7419 +t=58: Selected seed 195 with value = 0.7419 +Query 1/1: Action query time = 1.788 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8485 +t=74: Selected seed 195 with value = 0.8485 +Query 1/1: Action query time = 1.534 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.481 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=28--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=28--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 28 +# successes: 23 (82.1%) + +Task: put the cream cheese in the bowl +Starting episode 29... +Query 1/1: Action query time = 1.379 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4467 +t=10: Selected seed 195 with value = 0.4467 +Query 1/1: Action query time = 1.420 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5261 +t=26: Selected seed 195 with value = 0.5261 +Query 1/1: Action query time = 1.490 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6296 +t=42: Selected seed 195 with value = 0.6296 +Query 1/1: Action query time = 1.513 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6815 +t=58: Selected seed 195 with value = 0.6815 +Query 1/1: Action query time = 1.583 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7386 +t=74: Selected seed 195 with value = 0.7386 +Query 1/1: Action query time = 1.899 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9385 +t=90: Selected seed 195 with value = 0.9385 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=29--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=29--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 29 +# successes: 24 (82.8%) + +Task: put the cream cheese in the bowl +Starting episode 30... +Query 1/1: Action query time = 1.289 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4618 +t=10: Selected seed 195 with value = 0.4618 +Query 1/1: Action query time = 1.293 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5203 +t=26: Selected seed 195 with value = 0.5203 +Query 1/1: Action query time = 1.906 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6317 +t=42: Selected seed 195 with value = 0.6317 +Query 1/1: Action query time = 1.814 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7362 +t=58: Selected seed 195 with value = 0.7362 +Query 1/1: Action query time = 1.830 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9162 +t=74: Selected seed 195 with value = 0.9162 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=30--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=30--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 30 +# successes: 25 (83.3%) + +Task: put the cream cheese in the bowl +Starting episode 31... +Query 1/1: Action query time = 1.019 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4459 +t=10: Selected seed 195 with value = 0.4459 +Query 1/1: Action query time = 1.892 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5619 +t=26: Selected seed 195 with value = 0.5619 +Query 1/1: Action query time = 2.258 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6278 +t=42: Selected seed 195 with value = 0.6278 +Query 1/1: Action query time = 2.621 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7510 +t=58: Selected seed 195 with value = 0.7510 +Query 1/1: Action query time = 2.058 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8416 +t=74: Selected seed 195 with value = 0.8416 +Query 1/1: Action query time = 1.491 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.881 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.599 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.870 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.727 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.076 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9775 +t=170: Selected seed 195 with value = 0.9775 +Query 1/1: Action query time = 1.397 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6839 +t=186: Selected seed 195 with value = 0.6839 +Query 1/1: Action query time = 1.190 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8113 +t=202: Selected seed 195 with value = 0.8113 +Query 1/1: Action query time = 1.545 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9121 +t=218: Selected seed 195 with value = 0.9121 +Query 1/1: Action query time = 1.670 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9895 +t=234: Selected seed 195 with value = 0.9895 +Query 1/1: Action query time = 1.598 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9772 +t=250: Selected seed 195 with value = 0.9772 +Query 1/1: Action query time = 1.820 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.684 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.676 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=31--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=31--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 31 +# successes: 25 (80.6%) + +Task: put the cream cheese in the bowl +Starting episode 32... +Query 1/1: Action query time = 1.344 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4692 +t=10: Selected seed 195 with value = 0.4692 +Query 1/1: Action query time = 0.975 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4847 +t=26: Selected seed 195 with value = 0.4847 +Query 1/1: Action query time = 1.501 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5624 +t=42: Selected seed 195 with value = 0.5624 +Query 1/1: Action query time = 1.697 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6476 +t=58: Selected seed 195 with value = 0.6476 +Query 1/1: Action query time = 1.684 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8222 +t=74: Selected seed 195 with value = 0.8222 +Query 1/1: Action query time = 1.728 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9848 +t=90: Selected seed 195 with value = 0.9848 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=32--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=32--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 32 +# successes: 26 (81.2%) + +Task: put the cream cheese in the bowl +Starting episode 33... +Query 1/1: Action query time = 1.995 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4411 +t=10: Selected seed 195 with value = 0.4411 +Query 1/1: Action query time = 1.803 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4913 +t=26: Selected seed 195 with value = 0.4913 +Query 1/1: Action query time = 1.038 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5734 +t=42: Selected seed 195 with value = 0.5734 +Query 1/1: Action query time = 0.994 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6665 +t=58: Selected seed 195 with value = 0.6665 +Query 1/1: Action query time = 1.444 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7861 +t=74: Selected seed 195 with value = 0.7861 +Query 1/1: Action query time = 2.283 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9731 +t=90: Selected seed 195 with value = 0.9731 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=33--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=33--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 33 +# successes: 27 (81.8%) + +Task: put the cream cheese in the bowl +Starting episode 34... +Query 1/1: Action query time = 1.679 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4287 +t=10: Selected seed 195 with value = 0.4287 +Query 1/1: Action query time = 1.737 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5119 +t=26: Selected seed 195 with value = 0.5119 +Query 1/1: Action query time = 1.630 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6121 +t=42: Selected seed 195 with value = 0.6121 +Query 1/1: Action query time = 1.209 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7477 +t=58: Selected seed 195 with value = 0.7477 +Query 1/1: Action query time = 1.226 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8675 +t=74: Selected seed 195 with value = 0.8675 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=34--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=34--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 34 +# successes: 28 (82.4%) + +Task: put the cream cheese in the bowl +Starting episode 35... +Query 1/1: Action query time = 1.493 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4586 +t=10: Selected seed 195 with value = 0.4586 +Query 1/1: Action query time = 1.903 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4814 +t=26: Selected seed 195 with value = 0.4814 +Query 1/1: Action query time = 2.138 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5622 +t=42: Selected seed 195 with value = 0.5622 +Query 1/1: Action query time = 1.408 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6651 +t=58: Selected seed 195 with value = 0.6651 +Query 1/1: Action query time = 1.319 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8073 +t=74: Selected seed 195 with value = 0.8073 +Query 1/1: Action query time = 1.710 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9466 +t=90: Selected seed 195 with value = 0.9466 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=35--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=35--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 35 +# successes: 29 (82.9%) + +Task: put the cream cheese in the bowl +Starting episode 36... +Query 1/1: Action query time = 1.791 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4277 +t=10: Selected seed 195 with value = 0.4277 +Query 1/1: Action query time = 1.615 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5323 +t=26: Selected seed 195 with value = 0.5323 +Query 1/1: Action query time = 1.759 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5976 +t=42: Selected seed 195 with value = 0.5976 +Query 1/1: Action query time = 1.197 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7446 +t=58: Selected seed 195 with value = 0.7446 +Query 1/1: Action query time = 1.264 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8691 +t=74: Selected seed 195 with value = 0.8691 +Query 1/1: Action query time = 1.026 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=90: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 2.427 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.015 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.062 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=138: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 1.840 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.608 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.732 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7721 +t=186: Selected seed 195 with value = 0.7721 +Query 1/1: Action query time = 1.393 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8322 +t=202: Selected seed 195 with value = 0.8322 +Query 1/1: Action query time = 1.469 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8957 +t=218: Selected seed 195 with value = 0.8957 +Query 1/1: Action query time = 1.155 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=234: Selected seed 195 with value = 0.9963 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=36--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=36--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 36 +# successes: 30 (83.3%) + +Task: put the cream cheese in the bowl +Starting episode 37... +Query 1/1: Action query time = 1.809 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4374 +t=10: Selected seed 195 with value = 0.4374 +Query 1/1: Action query time = 1.681 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5263 +t=26: Selected seed 195 with value = 0.5263 +Query 1/1: Action query time = 1.554 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6364 +t=42: Selected seed 195 with value = 0.6364 +Query 1/1: Action query time = 1.366 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6886 +t=58: Selected seed 195 with value = 0.6886 +Query 1/1: Action query time = 1.587 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7727 +t=74: Selected seed 195 with value = 0.7727 +Query 1/1: Action query time = 1.855 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9390 +t=90: Selected seed 195 with value = 0.9390 +Query 1/1: Action query time = 1.835 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.680 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.876 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.979 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.802 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.611 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6852 +t=186: Selected seed 195 with value = 0.6852 +Query 1/1: Action query time = 1.349 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8505 +t=202: Selected seed 195 with value = 0.8505 +Query 1/1: Action query time = 1.502 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9152 +t=218: Selected seed 195 with value = 0.9152 +Query 1/1: Action query time = 1.842 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9802 +t=234: Selected seed 195 with value = 0.9802 +Query 1/1: Action query time = 1.839 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.352 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.498 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9942 +t=282: Selected seed 195 with value = 0.9942 +Query 1/1: Action query time = 1.787 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9865 +t=298: Selected seed 195 with value = 0.9865 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=37--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=37--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 37 +# successes: 30 (81.1%) + +Task: put the cream cheese in the bowl +Starting episode 38... +Query 1/1: Action query time = 1.000 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4162 +t=10: Selected seed 195 with value = 0.4162 +Query 1/1: Action query time = 0.981 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4939 +t=26: Selected seed 195 with value = 0.4939 +Query 1/1: Action query time = 1.299 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5977 +t=42: Selected seed 195 with value = 0.5977 +Query 1/1: Action query time = 1.455 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7105 +t=58: Selected seed 195 with value = 0.7105 +Query 1/1: Action query time = 1.401 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7862 +t=74: Selected seed 195 with value = 0.7862 +Query 1/1: Action query time = 2.071 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8794 +t=90: Selected seed 195 with value = 0.8794 +Query 1/1: Action query time = 1.273 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9156 +t=106: Selected seed 195 with value = 0.9156 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=38--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=38--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 38 +# successes: 31 (81.6%) + +Task: put the cream cheese in the bowl +Starting episode 39... +Query 1/1: Action query time = 1.792 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4356 +t=10: Selected seed 195 with value = 0.4356 +Query 1/1: Action query time = 2.068 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5478 +t=26: Selected seed 195 with value = 0.5478 +Query 1/1: Action query time = 1.432 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6426 +t=42: Selected seed 195 with value = 0.6426 +Query 1/1: Action query time = 1.174 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7524 +t=58: Selected seed 195 with value = 0.7524 +Query 1/1: Action query time = 0.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8551 +t=74: Selected seed 195 with value = 0.8551 +Query 1/1: Action query time = 1.440 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=39--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=39--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 39 +# successes: 32 (82.1%) + +Task: put the cream cheese in the bowl +Starting episode 40... +Query 1/1: Action query time = 1.506 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4794 +t=10: Selected seed 195 with value = 0.4794 +Query 1/1: Action query time = 1.718 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5560 +t=26: Selected seed 195 with value = 0.5560 +Query 1/1: Action query time = 1.354 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6010 +t=42: Selected seed 195 with value = 0.6010 +Query 1/1: Action query time = 1.610 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6775 +t=58: Selected seed 195 with value = 0.6775 +Query 1/1: Action query time = 1.253 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7985 +t=74: Selected seed 195 with value = 0.7985 +Query 1/1: Action query time = 1.031 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9240 +t=90: Selected seed 195 with value = 0.9240 +Query 1/1: Action query time = 0.989 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.166 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6145 +t=122: Selected seed 195 with value = 0.6145 +Query 1/1: Action query time = 1.268 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7097 +t=138: Selected seed 195 with value = 0.7097 +Query 1/1: Action query time = 1.825 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8599 +t=154: Selected seed 195 with value = 0.8599 +Query 1/1: Action query time = 1.283 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9250 +t=170: Selected seed 195 with value = 0.9250 +Query 1/1: Action query time = 2.116 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9254 +t=186: Selected seed 195 with value = 0.9254 +Query 1/1: Action query time = 1.307 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9061 +t=202: Selected seed 195 with value = 0.9061 +Query 1/1: Action query time = 1.307 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8902 +t=218: Selected seed 195 with value = 0.8902 +Query 1/1: Action query time = 1.751 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9144 +t=234: Selected seed 195 with value = 0.9144 +Query 1/1: Action query time = 1.263 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9172 +t=250: Selected seed 195 with value = 0.9172 +Query 1/1: Action query time = 1.284 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9224 +t=266: Selected seed 195 with value = 0.9224 +Query 1/1: Action query time = 1.471 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9325 +t=282: Selected seed 195 with value = 0.9325 +Query 1/1: Action query time = 1.197 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8915 +t=298: Selected seed 195 with value = 0.8915 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=40--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=40--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 40 +# successes: 32 (80.0%) + +Task: put the cream cheese in the bowl +Starting episode 41... +Query 1/1: Action query time = 0.978 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4735 +t=10: Selected seed 195 with value = 0.4735 +Query 1/1: Action query time = 1.025 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5359 +t=26: Selected seed 195 with value = 0.5359 +Query 1/1: Action query time = 0.976 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6410 +t=42: Selected seed 195 with value = 0.6410 +Query 1/1: Action query time = 1.727 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7354 +t=58: Selected seed 195 with value = 0.7354 +Query 1/1: Action query time = 1.444 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8680 +t=74: Selected seed 195 with value = 0.8680 +Query 1/1: Action query time = 1.225 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8993 +t=90: Selected seed 195 with value = 0.8993 +Query 1/1: Action query time = 1.464 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9309 +t=106: Selected seed 195 with value = 0.9309 +Query 1/1: Action query time = 1.174 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9659 +t=122: Selected seed 195 with value = 0.9659 +Query 1/1: Action query time = 1.700 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9662 +t=138: Selected seed 195 with value = 0.9662 +Query 1/1: Action query time = 1.267 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9577 +t=154: Selected seed 195 with value = 0.9577 +Query 1/1: Action query time = 1.263 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9595 +t=170: Selected seed 195 with value = 0.9595 +Query 1/1: Action query time = 1.020 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9612 +t=186: Selected seed 195 with value = 0.9612 +Query 1/1: Action query time = 1.523 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9627 +t=202: Selected seed 195 with value = 0.9627 +Query 1/1: Action query time = 1.256 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9645 +t=218: Selected seed 195 with value = 0.9645 +Query 1/1: Action query time = 1.148 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9686 +t=234: Selected seed 195 with value = 0.9686 +Query 1/1: Action query time = 1.460 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9699 +t=250: Selected seed 195 with value = 0.9699 +Query 1/1: Action query time = 1.220 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9711 +t=266: Selected seed 195 with value = 0.9711 +Query 1/1: Action query time = 1.444 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9523 +t=282: Selected seed 195 with value = 0.9523 +Query 1/1: Action query time = 1.172 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9612 +t=298: Selected seed 195 with value = 0.9612 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=41--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=41--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 41 +# successes: 32 (78.0%) + +Task: put the cream cheese in the bowl +Starting episode 42... +Query 1/1: Action query time = 1.368 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4197 +t=10: Selected seed 195 with value = 0.4197 +Query 1/1: Action query time = 1.205 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4944 +t=26: Selected seed 195 with value = 0.4944 +Query 1/1: Action query time = 1.534 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5815 +t=42: Selected seed 195 with value = 0.5815 +Query 1/1: Action query time = 1.341 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6490 +t=58: Selected seed 195 with value = 0.6490 +Query 1/1: Action query time = 1.360 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7828 +t=74: Selected seed 195 with value = 0.7828 +Query 1/1: Action query time = 1.295 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9144 +t=90: Selected seed 195 with value = 0.9144 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=42--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=42--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 42 +# successes: 33 (78.6%) + +Task: put the cream cheese in the bowl +Starting episode 43... +Query 1/1: Action query time = 1.396 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4144 +t=10: Selected seed 195 with value = 0.4144 +Query 1/1: Action query time = 1.068 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5133 +t=26: Selected seed 195 with value = 0.5133 +Query 1/1: Action query time = 0.984 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6234 +t=42: Selected seed 195 with value = 0.6234 +Query 1/1: Action query time = 1.018 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6992 +t=58: Selected seed 195 with value = 0.6992 +Query 1/1: Action query time = 1.399 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8545 +t=74: Selected seed 195 with value = 0.8545 +Query 1/1: Action query time = 1.502 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9950 +t=90: Selected seed 195 with value = 0.9950 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=43--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=43--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 43 +# successes: 34 (79.1%) + +Task: put the cream cheese in the bowl +Starting episode 44... +Query 1/1: Action query time = 1.204 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4513 +t=10: Selected seed 195 with value = 0.4513 +Query 1/1: Action query time = 1.396 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3540 +t=26: Selected seed 195 with value = 0.3540 +Query 1/1: Action query time = 1.155 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5602 +t=42: Selected seed 195 with value = 0.5602 +Query 1/1: Action query time = 1.185 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6704 +t=58: Selected seed 195 with value = 0.6704 +Query 1/1: Action query time = 1.238 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8496 +t=74: Selected seed 195 with value = 0.8496 +Query 1/1: Action query time = 1.316 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9791 +t=90: Selected seed 195 with value = 0.9791 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=44--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=44--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 44 +# successes: 35 (79.5%) + +Task: put the cream cheese in the bowl +Starting episode 45... +Query 1/1: Action query time = 1.212 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4406 +t=10: Selected seed 195 with value = 0.4406 +Query 1/1: Action query time = 1.679 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5252 +t=26: Selected seed 195 with value = 0.5252 +Query 1/1: Action query time = 0.995 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5868 +t=42: Selected seed 195 with value = 0.5868 +Query 1/1: Action query time = 1.483 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6663 +t=58: Selected seed 195 with value = 0.6663 +Query 1/1: Action query time = 1.479 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7778 +t=74: Selected seed 195 with value = 0.7778 +Query 1/1: Action query time = 1.203 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9316 +t=90: Selected seed 195 with value = 0.9316 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=45--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=45--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 45 +# successes: 36 (80.0%) + +Task: put the cream cheese in the bowl +Starting episode 46... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4664 +t=10: Selected seed 195 with value = 0.4664 +Query 1/1: Action query time = 1.611 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5039 +t=26: Selected seed 195 with value = 0.5039 +Query 1/1: Action query time = 0.986 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6142 +t=42: Selected seed 195 with value = 0.6142 +Query 1/1: Action query time = 1.576 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6611 +t=58: Selected seed 195 with value = 0.6611 +Query 1/1: Action query time = 1.247 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7786 +t=74: Selected seed 195 with value = 0.7786 +Query 1/1: Action query time = 1.288 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9292 +t=90: Selected seed 195 with value = 0.9292 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=46--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=46--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 46 +# successes: 37 (80.4%) + +Task: put the cream cheese in the bowl +Starting episode 47... +Query 1/1: Action query time = 1.611 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4551 +t=10: Selected seed 195 with value = 0.4551 +Query 1/1: Action query time = 1.583 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5270 +t=26: Selected seed 195 with value = 0.5270 +Query 1/1: Action query time = 1.266 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6255 +t=42: Selected seed 195 with value = 0.6255 +Query 1/1: Action query time = 1.115 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6568 +t=58: Selected seed 195 with value = 0.6568 +Query 1/1: Action query time = 1.233 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8165 +t=74: Selected seed 195 with value = 0.8165 +Query 1/1: Action query time = 1.305 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9496 +t=90: Selected seed 195 with value = 0.9496 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=47--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=47--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 47 +# successes: 38 (80.9%) + +Task: put the cream cheese in the bowl +Starting episode 48... +Query 1/1: Action query time = 1.190 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4680 +t=10: Selected seed 195 with value = 0.4680 +Query 1/1: Action query time = 1.346 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6057 +t=26: Selected seed 195 with value = 0.6057 +Query 1/1: Action query time = 1.201 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6227 +t=42: Selected seed 195 with value = 0.6227 +Query 1/1: Action query time = 1.533 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7300 +t=58: Selected seed 195 with value = 0.7300 +Query 1/1: Action query time = 1.042 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8504 +t=74: Selected seed 195 with value = 0.8504 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=48--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=48--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 48 +# successes: 39 (81.2%) + +Task: put the cream cheese in the bowl +Starting episode 49... +Query 1/1: Action query time = 1.440 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4278 +t=10: Selected seed 195 with value = 0.4278 +Query 1/1: Action query time = 1.772 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5153 +t=26: Selected seed 195 with value = 0.5153 +Query 1/1: Action query time = 1.368 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6342 +t=42: Selected seed 195 with value = 0.6342 +Query 1/1: Action query time = 1.253 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7210 +t=58: Selected seed 195 with value = 0.7210 +Query 1/1: Action query time = 1.433 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8817 +t=74: Selected seed 195 with value = 0.8817 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=49--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=49--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 49 +# successes: 40 (81.6%) + +Task: put the cream cheese in the bowl +Starting episode 50... +Query 1/1: Action query time = 1.676 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4234 +t=10: Selected seed 195 with value = 0.4234 +Query 1/1: Action query time = 0.969 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5422 +t=26: Selected seed 195 with value = 0.5422 +Query 1/1: Action query time = 0.969 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6217 +t=42: Selected seed 195 with value = 0.6217 +Query 1/1: Action query time = 1.120 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6930 +t=58: Selected seed 195 with value = 0.6930 +Query 1/1: Action query time = 1.536 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8219 +t=74: Selected seed 195 with value = 0.8219 +Query 1/1: Action query time = 1.560 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9649 +t=90: Selected seed 195 with value = 0.9649 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--episode=50--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Saved rollout MP4 at path ./rollouts/forget_cl_iter350_par_t6/2026_08_01-11_38_10--with_future_img--episode=50--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 50 +# successes: 41 (82.0%) +Current task success rate: 0.82 +Current total success rate: 0.82 +Final results: +Total episodes: 50 +Total successes: 41 +Overall success rate: 0.8200 (82.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-12_39_04--base40k_t3_shard08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-12_39_04--base40k_t3_shard08.txt new file mode 100644 index 0000000000000000000000000000000000000000..d769a030964b7a8cca981890347b2a7152ce3022 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-12_39_04--base40k_t3_shard08.txt @@ -0,0 +1,177 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40k_t3_shard08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 2.585 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2110 +t=10: Selected seed 195 with value = 0.2110 +Query 1/1: Action query time = 3.154 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2319 +t=26: Selected seed 195 with value = 0.2319 +Query 1/1: Action query time = 4.545 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2398 +t=42: Selected seed 195 with value = 0.2398 +Query 1/1: Action query time = 5.600 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2850 +t=58: Selected seed 195 with value = 0.2850 +Query 1/1: Action query time = 5.028 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3150 +t=74: Selected seed 195 with value = 0.3150 +Query 1/1: Action query time = 5.109 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2938 +t=90: Selected seed 195 with value = 0.2938 +Query 1/1: Action query time = 5.218 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3046 +t=106: Selected seed 195 with value = 0.3046 +Query 1/1: Action query time = 4.775 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3629 +t=122: Selected seed 195 with value = 0.3629 +Query 1/1: Action query time = 5.372 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3890 +t=138: Selected seed 195 with value = 0.3890 +Query 1/1: Action query time = 4.344 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5419 +t=154: Selected seed 195 with value = 0.5419 +Query 1/1: Action query time = 4.258 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6786 +t=170: Selected seed 195 with value = 0.6786 +Query 1/1: Action query time = 5.137 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8045 +t=186: Selected seed 195 with value = 0.8045 +Query 1/1: Action query time = 4.672 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9411 +t=202: Selected seed 195 with value = 0.9411 +Saved rollout MP4 at path ./rollouts/base40k_t3_shard08/2026_08_01-12_39_04--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 6.126 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1701 +t=10: Selected seed 195 with value = 0.1701 +Query 1/1: Action query time = 5.910 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1949 +t=26: Selected seed 195 with value = 0.1949 +Query 1/1: Action query time = 3.485 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2320 +t=42: Selected seed 195 with value = 0.2320 +Query 1/1: Action query time = 4.377 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2509 +t=58: Selected seed 195 with value = 0.2509 +Query 1/1: Action query time = 5.876 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3449 +t=74: Selected seed 195 with value = 0.3449 +Query 1/1: Action query time = 5.072 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4051 +t=90: Selected seed 195 with value = 0.4051 +Query 1/1: Action query time = 4.066 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4790 +t=106: Selected seed 195 with value = 0.4790 +Query 1/1: Action query time = 4.732 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5432 +t=122: Selected seed 195 with value = 0.5432 +Query 1/1: Action query time = 5.325 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6322 +t=138: Selected seed 195 with value = 0.6322 +Query 1/1: Action query time = 3.632 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7815 +t=154: Selected seed 195 with value = 0.7815 +Query 1/1: Action query time = 4.132 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9404 +t=170: Selected seed 195 with value = 0.9404 +Saved rollout MP4 at path ./rollouts/base40k_t3_shard08/2026_08_01-12_39_04--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.348 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1769 +t=10: Selected seed 195 with value = 0.1769 +Query 1/1: Action query time = 4.781 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1550 +t=26: Selected seed 195 with value = 0.1550 +Query 1/1: Action query time = 4.469 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1635 +t=42: Selected seed 195 with value = 0.1635 +Query 1/1: Action query time = 4.974 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2677 +t=58: Selected seed 195 with value = 0.2677 +Query 1/1: Action query time = 5.061 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2712 +t=74: Selected seed 195 with value = 0.2712 +Query 1/1: Action query time = 4.665 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2368 +t=90: Selected seed 195 with value = 0.2368 +Query 1/1: Action query time = 4.677 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4251 +t=106: Selected seed 195 with value = 0.4251 +Query 1/1: Action query time = 4.879 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4632 +t=122: Selected seed 195 with value = 0.4632 +Query 1/1: Action query time = 4.972 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5786 +t=138: Selected seed 195 with value = 0.5786 +Query 1/1: Action query time = 5.175 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6628 +t=154: Selected seed 195 with value = 0.6628 +Query 1/1: Action query time = 3.646 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8136 +t=170: Selected seed 195 with value = 0.8136 +Query 1/1: Action query time = 3.727 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9571 +t=186: Selected seed 195 with value = 0.9571 +Saved rollout MP4 at path ./rollouts/base40k_t3_shard08/2026_08_01-12_39_04--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-12_39_04--base40k_t3_shard09.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-12_39_04--base40k_t3_shard09.txt new file mode 100644 index 0000000000000000000000000000000000000000..744f39ee72a8cfaacd62be9b1e30ec786cc3a74c --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-12_39_04--base40k_t3_shard09.txt @@ -0,0 +1,205 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40k_t3_shard09', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='9,25,41', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 6.124 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1059 +t=10: Selected seed 195 with value = 0.1059 +Query 1/1: Action query time = 4.845 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2284 +t=26: Selected seed 195 with value = 0.2284 +Query 1/1: Action query time = 4.168 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2674 +t=42: Selected seed 195 with value = 0.2674 +Query 1/1: Action query time = 4.731 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2979 +t=58: Selected seed 195 with value = 0.2979 +Query 1/1: Action query time = 4.852 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3856 +t=74: Selected seed 195 with value = 0.3856 +Query 1/1: Action query time = 5.457 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4623 +t=90: Selected seed 195 with value = 0.4623 +Query 1/1: Action query time = 4.171 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5271 +t=106: Selected seed 195 with value = 0.5271 +Query 1/1: Action query time = 4.805 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6178 +t=122: Selected seed 195 with value = 0.6178 +Query 1/1: Action query time = 5.240 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7085 +t=138: Selected seed 195 with value = 0.7085 +Query 1/1: Action query time = 5.393 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7663 +t=154: Selected seed 195 with value = 0.7663 +Query 1/1: Action query time = 3.907 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8728 +t=170: Selected seed 195 with value = 0.8728 +Query 1/1: Action query time = 3.648 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7394 +t=186: Selected seed 195 with value = 0.7394 +Query 1/1: Action query time = 3.201 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2852 +t=202: Selected seed 195 with value = 0.2852 +Query 1/1: Action query time = 4.191 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2984 +t=218: Selected seed 195 with value = 0.2984 +Query 1/1: Action query time = 4.236 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2945 +t=234: Selected seed 195 with value = 0.2945 +Query 1/1: Action query time = 4.894 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2929 +t=250: Selected seed 195 with value = 0.2929 +Query 1/1: Action query time = 4.617 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2930 +t=266: Selected seed 195 with value = 0.2930 +Query 1/1: Action query time = 3.278 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2902 +t=282: Selected seed 195 with value = 0.2902 +Query 1/1: Action query time = 5.254 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2912 +t=298: Selected seed 195 with value = 0.2912 +Saved rollout MP4 at path ./rollouts/base40k_t3_shard09/2026_08_01-12_39_04--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.550 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1299 +t=10: Selected seed 195 with value = 0.1299 +Query 1/1: Action query time = 4.850 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1808 +t=26: Selected seed 195 with value = 0.1808 +Query 1/1: Action query time = 4.599 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2050 +t=42: Selected seed 195 with value = 0.2050 +Query 1/1: Action query time = 3.986 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2018 +t=58: Selected seed 195 with value = 0.2018 +Query 1/1: Action query time = 4.671 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2434 +t=74: Selected seed 195 with value = 0.2434 +Query 1/1: Action query time = 3.857 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2981 +t=90: Selected seed 195 with value = 0.2981 +Query 1/1: Action query time = 4.495 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3460 +t=106: Selected seed 195 with value = 0.3460 +Query 1/1: Action query time = 4.473 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4191 +t=122: Selected seed 195 with value = 0.4191 +Query 1/1: Action query time = 4.430 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5996 +t=138: Selected seed 195 with value = 0.5996 +Query 1/1: Action query time = 5.132 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7225 +t=154: Selected seed 195 with value = 0.7225 +Query 1/1: Action query time = 5.164 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8205 +t=170: Selected seed 195 with value = 0.8205 +Query 1/1: Action query time = 5.183 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9500 +t=186: Selected seed 195 with value = 0.9500 +Saved rollout MP4 at path ./rollouts/base40k_t3_shard09/2026_08_01-12_39_04--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 5.131 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1887 +t=10: Selected seed 195 with value = 0.1887 +Query 1/1: Action query time = 5.343 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2153 +t=26: Selected seed 195 with value = 0.2153 +Query 1/1: Action query time = 4.905 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2424 +t=42: Selected seed 195 with value = 0.2424 +Query 1/1: Action query time = 4.043 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2169 +t=58: Selected seed 195 with value = 0.2169 +Query 1/1: Action query time = 3.780 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2501 +t=74: Selected seed 195 with value = 0.2501 +Query 1/1: Action query time = 3.936 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2531 +t=90: Selected seed 195 with value = 0.2531 +Query 1/1: Action query time = 3.614 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3340 +t=106: Selected seed 195 with value = 0.3340 +Query 1/1: Action query time = 2.082 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4191 +t=122: Selected seed 195 with value = 0.4191 +Query 1/1: Action query time = 2.316 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4961 +t=138: Selected seed 195 with value = 0.4961 +Query 1/1: Action query time = 2.753 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6339 +t=154: Selected seed 195 with value = 0.6339 +Query 1/1: Action query time = 2.751 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7805 +t=170: Selected seed 195 with value = 0.7805 +Query 1/1: Action query time = 2.835 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9219 +t=186: Selected seed 195 with value = 0.9219 +Saved rollout MP4 at path ./rollouts/base40k_t3_shard09/2026_08_01-12_39_04--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_06--demochan200_t1_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_06--demochan200_t1_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..e636697f854e7912bc2b9f0550e14de5121ffba5 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_06--demochan200_t1_s01.txt @@ -0,0 +1,136 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t1_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 2.670 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4951 +t=10: Selected seed 195 with value = 0.4951 +Query 1/1: Action query time = 4.678 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5799 +t=26: Selected seed 195 with value = 0.5799 +Query 1/1: Action query time = 4.468 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6991 +t=42: Selected seed 195 with value = 0.6991 +Query 1/1: Action query time = 5.106 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8254 +t=58: Selected seed 195 with value = 0.8254 +Query 1/1: Action query time = 5.859 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8948 +t=74: Selected seed 195 with value = 0.8948 +Query 1/1: Action query time = 4.871 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=90: Selected seed 195 with value = 0.9997 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s01/2026_08_01-13_00_06--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 2.240 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4993 +t=10: Selected seed 195 with value = 0.4993 +Query 1/1: Action query time = 4.568 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5819 +t=26: Selected seed 195 with value = 0.5819 +Query 1/1: Action query time = 3.637 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7046 +t=42: Selected seed 195 with value = 0.7046 +Query 1/1: Action query time = 4.789 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8330 +t=58: Selected seed 195 with value = 0.8330 +Query 1/1: Action query time = 5.284 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9211 +t=74: Selected seed 195 with value = 0.9211 +Query 1/1: Action query time = 4.599 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9938 +t=90: Selected seed 195 with value = 0.9938 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s01/2026_08_01-13_00_06--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 2.998 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4941 +t=10: Selected seed 195 with value = 0.4941 +Query 1/1: Action query time = 4.035 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5744 +t=26: Selected seed 195 with value = 0.5744 +Query 1/1: Action query time = 3.833 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6825 +t=42: Selected seed 195 with value = 0.6825 +Query 1/1: Action query time = 5.142 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8006 +t=58: Selected seed 195 with value = 0.8006 +Query 1/1: Action query time = 6.028 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9134 +t=74: Selected seed 195 with value = 0.9134 +Query 1/1: Action query time = 5.251 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9939 +t=90: Selected seed 195 with value = 0.9939 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s01/2026_08_01-13_00_06--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 3.515 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5295 +t=10: Selected seed 195 with value = 0.5295 +Query 1/1: Action query time = 1.462 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5815 +t=26: Selected seed 195 with value = 0.5815 +Query 1/1: Action query time = 1.351 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6730 +t=42: Selected seed 195 with value = 0.6730 +Query 1/1: Action query time = 1.309 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7796 +t=58: Selected seed 195 with value = 0.7796 +Query 1/1: Action query time = 1.335 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8868 +t=74: Selected seed 195 with value = 0.8868 +Query 1/1: Action query time = 1.332 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9874 +t=90: Selected seed 195 with value = 0.9874 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s01/2026_08_01-13_00_06--with_future_img--episode=4--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 4 +Total successes: 4 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_06--demochan200_t1_s04.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_06--demochan200_t1_s04.txt new file mode 100644 index 0000000000000000000000000000000000000000..cc63f1bd58b11cf41d2877b3d9842cd01aed46cc --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_06--demochan200_t1_s04.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t1_s04', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='4,20,36', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 6.244 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4974 +t=10: Selected seed 195 with value = 0.4974 +Query 1/1: Action query time = 4.574 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5756 +t=26: Selected seed 195 with value = 0.5756 +Query 1/1: Action query time = 4.739 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7176 +t=42: Selected seed 195 with value = 0.7176 +Query 1/1: Action query time = 6.152 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8302 +t=58: Selected seed 195 with value = 0.8302 +Query 1/1: Action query time = 4.985 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9152 +t=74: Selected seed 195 with value = 0.9152 +Query 1/1: Action query time = 2.625 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9974 +t=90: Selected seed 195 with value = 0.9974 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s04/2026_08_01-13_00_06--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 4.806 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4949 +t=10: Selected seed 195 with value = 0.4949 +Query 1/1: Action query time = 6.218 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5917 +t=26: Selected seed 195 with value = 0.5917 +Query 1/1: Action query time = 5.064 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6891 +t=42: Selected seed 195 with value = 0.6891 +Query 1/1: Action query time = 4.927 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8048 +t=58: Selected seed 195 with value = 0.8048 +Query 1/1: Action query time = 4.091 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8753 +t=74: Selected seed 195 with value = 0.8753 +Query 1/1: Action query time = 4.248 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9910 +t=90: Selected seed 195 with value = 0.9910 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s04/2026_08_01-13_00_06--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 4.762 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5108 +t=10: Selected seed 195 with value = 0.5108 +Query 1/1: Action query time = 5.504 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5999 +t=26: Selected seed 195 with value = 0.5999 +Query 1/1: Action query time = 6.262 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7142 +t=42: Selected seed 195 with value = 0.7142 +Query 1/1: Action query time = 5.371 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8173 +t=58: Selected seed 195 with value = 0.8173 +Query 1/1: Action query time = 3.449 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9384 +t=74: Selected seed 195 with value = 0.9384 +Query 1/1: Action query time = 3.394 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=90: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s04/2026_08_01-13_00_06--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_08--demochan200_t1_s13.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_08--demochan200_t1_s13.txt new file mode 100644 index 0000000000000000000000000000000000000000..53a5a9d169b78bb39677fe9044d756d46ef67f68 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_08--demochan200_t1_s13.txt @@ -0,0 +1,109 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t1_s13', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='13,29,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 4.388 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5096 +t=10: Selected seed 195 with value = 0.5096 +Query 1/1: Action query time = 4.428 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5974 +t=26: Selected seed 195 with value = 0.5974 +Query 1/1: Action query time = 5.091 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6784 +t=42: Selected seed 195 with value = 0.6784 +Query 1/1: Action query time = 5.303 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7855 +t=58: Selected seed 195 with value = 0.7855 +Query 1/1: Action query time = 3.684 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8930 +t=74: Selected seed 195 with value = 0.8930 +Query 1/1: Action query time = 4.311 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9952 +t=90: Selected seed 195 with value = 0.9952 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s13/2026_08_01-13_00_08--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 2.202 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5442 +t=10: Selected seed 195 with value = 0.5442 +Query 1/1: Action query time = 2.505 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5939 +t=26: Selected seed 195 with value = 0.5939 +Query 1/1: Action query time = 5.461 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7063 +t=42: Selected seed 195 with value = 0.7063 +Query 1/1: Action query time = 4.392 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8136 +t=58: Selected seed 195 with value = 0.8136 +Query 1/1: Action query time = 4.119 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9305 +t=74: Selected seed 195 with value = 0.9305 +Query 1/1: Action query time = 4.965 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=90: Selected seed 195 with value = 0.9984 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s13/2026_08_01-13_00_08--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.059 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4983 +t=10: Selected seed 195 with value = 0.4983 +Query 1/1: Action query time = 2.475 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5743 +t=26: Selected seed 195 with value = 0.5743 +Query 1/1: Action query time = 3.463 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8213 +t=42: Selected seed 195 with value = 0.8213 +Query 1/1: Action query time = 4.234 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7098 +t=58: Selected seed 195 with value = 0.7098 +Query 1/1: Action query time = 5.308 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7969 +t=74: Selected seed 195 with value = 0.7969 +Query 1/1: Action query time = 4.353 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9417 +t=90: Selected seed 195 with value = 0.9417 +Query 1/1: Action query time = 4.767 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s13/2026_08_01-13_00_08--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_08--demochan200_t1_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_08--demochan200_t1_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..bc099379c31fc0c29d99fadbbb70fc84aa4a0a84 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_00_08--demochan200_t1_s14.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t1_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 6.914 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5093 +t=10: Selected seed 195 with value = 0.5093 +Query 1/1: Action query time = 5.684 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5948 +t=26: Selected seed 195 with value = 0.5948 +Query 1/1: Action query time = 4.371 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7051 +t=42: Selected seed 195 with value = 0.7051 +Query 1/1: Action query time = 4.462 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8098 +t=58: Selected seed 195 with value = 0.8098 +Query 1/1: Action query time = 5.464 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9047 +t=74: Selected seed 195 with value = 0.9047 +Query 1/1: Action query time = 4.194 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9935 +t=90: Selected seed 195 with value = 0.9935 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s14/2026_08_01-13_00_08--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 4.241 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5002 +t=10: Selected seed 195 with value = 0.5002 +Query 1/1: Action query time = 4.854 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5789 +t=26: Selected seed 195 with value = 0.5789 +Query 1/1: Action query time = 5.721 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6773 +t=42: Selected seed 195 with value = 0.6773 +Query 1/1: Action query time = 5.483 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8061 +t=58: Selected seed 195 with value = 0.8061 +Query 1/1: Action query time = 4.508 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9123 +t=74: Selected seed 195 with value = 0.9123 +Query 1/1: Action query time = 5.045 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9930 +t=90: Selected seed 195 with value = 0.9930 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s14/2026_08_01-13_00_08--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.799 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5004 +t=10: Selected seed 195 with value = 0.5004 +Query 1/1: Action query time = 5.392 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5649 +t=26: Selected seed 195 with value = 0.5649 +Query 1/1: Action query time = 5.521 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6352 +t=42: Selected seed 195 with value = 0.6352 +Query 1/1: Action query time = 5.626 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7560 +t=58: Selected seed 195 with value = 0.7560 +Query 1/1: Action query time = 4.072 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8755 +t=74: Selected seed 195 with value = 0.8755 +Query 1/1: Action query time = 2.907 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=90: Selected seed 195 with value = 0.9962 +Saved rollout MP4 at path ./rollouts/demochan200_t1_s14/2026_08_01-13_00_08--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s00.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s00.txt new file mode 100644 index 0000000000000000000000000000000000000000..d0781ee21a485d86dd31311e856ca3a2ef0b1da7 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s00.txt @@ -0,0 +1,344 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t3_s00', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,16,32,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 6.302 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4312 +t=10: Selected seed 195 with value = 0.4312 +Query 1/1: Action query time = 5.268 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5702 +t=26: Selected seed 195 with value = 0.5702 +Query 1/1: Action query time = 5.154 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6488 +t=42: Selected seed 195 with value = 0.6488 +Query 1/1: Action query time = 4.884 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6565 +t=58: Selected seed 195 with value = 0.6565 +Query 1/1: Action query time = 5.406 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6689 +t=74: Selected seed 195 with value = 0.6689 +Query 1/1: Action query time = 5.356 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6746 +t=90: Selected seed 195 with value = 0.6746 +Query 1/1: Action query time = 5.141 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8536 +t=106: Selected seed 195 with value = 0.8536 +Query 1/1: Action query time = 5.141 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9310 +t=122: Selected seed 195 with value = 0.9310 +Query 1/1: Action query time = 5.216 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9390 +t=138: Selected seed 195 with value = 0.9390 +Query 1/1: Action query time = 5.225 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9538 +t=154: Selected seed 195 with value = 0.9538 +Query 1/1: Action query time = 5.311 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9509 +t=170: Selected seed 195 with value = 0.9509 +Query 1/1: Action query time = 5.260 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9768 +t=186: Selected seed 195 with value = 0.9768 +Query 1/1: Action query time = 5.337 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.985 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.343 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.718 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.257 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.075 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.426 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s00/2026_08_01-13_02_59--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 5.478 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3886 +t=10: Selected seed 195 with value = 0.3886 +Query 1/1: Action query time = 4.524 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5614 +t=26: Selected seed 195 with value = 0.5614 +Query 1/1: Action query time = 4.677 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6886 +t=42: Selected seed 195 with value = 0.6886 +Query 1/1: Action query time = 4.760 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6948 +t=58: Selected seed 195 with value = 0.6948 +Query 1/1: Action query time = 4.811 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7647 +t=74: Selected seed 195 with value = 0.7647 +Query 1/1: Action query time = 4.730 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9699 +t=90: Selected seed 195 with value = 0.9699 +Query 1/1: Action query time = 5.260 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9542 +t=106: Selected seed 195 with value = 0.9542 +Query 1/1: Action query time = 5.463 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9770 +t=122: Selected seed 195 with value = 0.9770 +Query 1/1: Action query time = 5.088 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9844 +t=138: Selected seed 195 with value = 0.9844 +Query 1/1: Action query time = 5.362 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.626 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.057 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.716 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.910 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.771 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.385 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.644 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.399 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.684 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s00/2026_08_01-13_02_59--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 6.438 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3924 +t=10: Selected seed 195 with value = 0.3924 +Query 1/1: Action query time = 5.459 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7584 +t=26: Selected seed 195 with value = 0.7584 +Query 1/1: Action query time = 4.884 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6858 +t=42: Selected seed 195 with value = 0.6858 +Query 1/1: Action query time = 5.028 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6830 +t=58: Selected seed 195 with value = 0.6830 +Query 1/1: Action query time = 5.324 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9163 +t=74: Selected seed 195 with value = 0.9163 +Query 1/1: Action query time = 5.247 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9292 +t=90: Selected seed 195 with value = 0.9292 +Query 1/1: Action query time = 4.013 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8196 +t=106: Selected seed 195 with value = 0.8196 +Query 1/1: Action query time = 4.990 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8054 +t=122: Selected seed 195 with value = 0.8054 +Query 1/1: Action query time = 4.900 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7626 +t=138: Selected seed 195 with value = 0.7626 +Query 1/1: Action query time = 4.805 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8205 +t=154: Selected seed 195 with value = 0.8205 +Query 1/1: Action query time = 5.166 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7447 +t=170: Selected seed 195 with value = 0.7447 +Query 1/1: Action query time = 5.396 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8074 +t=186: Selected seed 195 with value = 0.8074 +Query 1/1: Action query time = 5.428 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7264 +t=202: Selected seed 195 with value = 0.7264 +Query 1/1: Action query time = 4.395 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9738 +t=218: Selected seed 195 with value = 0.9738 +Query 1/1: Action query time = 4.537 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8315 +t=234: Selected seed 195 with value = 0.8315 +Query 1/1: Action query time = 2.741 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9848 +t=250: Selected seed 195 with value = 0.9848 +Query 1/1: Action query time = 2.876 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8704 +t=266: Selected seed 195 with value = 0.8704 +Query 1/1: Action query time = 3.949 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9714 +t=282: Selected seed 195 with value = 0.9714 +Query 1/1: Action query time = 2.226 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s00/2026_08_01-13_02_59--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 1.625 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3962 +t=10: Selected seed 195 with value = 0.3962 +Query 1/1: Action query time = 1.623 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5688 +t=26: Selected seed 195 with value = 0.5688 +Query 1/1: Action query time = 1.474 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6459 +t=42: Selected seed 195 with value = 0.6459 +Query 1/1: Action query time = 1.464 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6738 +t=58: Selected seed 195 with value = 0.6738 +Query 1/1: Action query time = 1.474 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6890 +t=74: Selected seed 195 with value = 0.6890 +Query 1/1: Action query time = 1.450 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6893 +t=90: Selected seed 195 with value = 0.6893 +Query 1/1: Action query time = 1.446 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7958 +t=106: Selected seed 195 with value = 0.7958 +Query 1/1: Action query time = 1.416 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9233 +t=122: Selected seed 195 with value = 0.9233 +Query 1/1: Action query time = 1.422 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8614 +t=138: Selected seed 195 with value = 0.8614 +Query 1/1: Action query time = 1.420 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8404 +t=154: Selected seed 195 with value = 0.8404 +Query 1/1: Action query time = 1.374 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7395 +t=170: Selected seed 195 with value = 0.7395 +Query 1/1: Action query time = 1.356 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9052 +t=186: Selected seed 195 with value = 0.9052 +Query 1/1: Action query time = 0.976 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9452 +t=202: Selected seed 195 with value = 0.9452 +Query 1/1: Action query time = 0.982 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9927 +t=218: Selected seed 195 with value = 0.9927 +Query 1/1: Action query time = 0.963 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9878 +t=234: Selected seed 195 with value = 0.9878 +Query 1/1: Action query time = 0.967 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9862 +t=250: Selected seed 195 with value = 0.9862 +Query 1/1: Action query time = 0.959 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=266: Selected seed 195 with value = 0.9864 +Query 1/1: Action query time = 0.960 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9857 +t=282: Selected seed 195 with value = 0.9857 +Query 1/1: Action query time = 0.957 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9847 +t=298: Selected seed 195 with value = 0.9847 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s00/2026_08_01-13_02_59--with_future_img--episode=4--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 4 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s05.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s05.txt new file mode 100644 index 0000000000000000000000000000000000000000..7ce1f2d482b93dc563215e80a10e94b1ec5f18d3 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s05.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t3_s05', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,21,37', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 6.206 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3658 +t=10: Selected seed 195 with value = 0.3658 +Query 1/1: Action query time = 5.233 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5382 +t=26: Selected seed 195 with value = 0.5382 +Query 1/1: Action query time = 4.658 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6450 +t=42: Selected seed 195 with value = 0.6450 +Query 1/1: Action query time = 5.259 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6669 +t=58: Selected seed 195 with value = 0.6669 +Query 1/1: Action query time = 5.254 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6666 +t=74: Selected seed 195 with value = 0.6666 +Query 1/1: Action query time = 5.086 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7891 +t=90: Selected seed 195 with value = 0.7891 +Query 1/1: Action query time = 5.136 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9846 +t=106: Selected seed 195 with value = 0.9846 +Query 1/1: Action query time = 5.054 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=122: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 4.949 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.004 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.059 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.103 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.900 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.995 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.145 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.965 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.705 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.134 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.810 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s05/2026_08_01-13_02_59--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.492 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3914 +t=10: Selected seed 195 with value = 0.3914 +Query 1/1: Action query time = 4.844 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6878 +t=26: Selected seed 195 with value = 0.6878 +Query 1/1: Action query time = 5.153 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6706 +t=42: Selected seed 195 with value = 0.6706 +Query 1/1: Action query time = 5.281 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6843 +t=58: Selected seed 195 with value = 0.6843 +Query 1/1: Action query time = 5.251 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9193 +t=74: Selected seed 195 with value = 0.9193 +Query 1/1: Action query time = 4.793 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9158 +t=90: Selected seed 195 with value = 0.9158 +Query 1/1: Action query time = 5.301 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8203 +t=106: Selected seed 195 with value = 0.8203 +Query 1/1: Action query time = 5.214 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8148 +t=122: Selected seed 195 with value = 0.8148 +Query 1/1: Action query time = 4.812 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7749 +t=138: Selected seed 195 with value = 0.7749 +Query 1/1: Action query time = 5.445 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7283 +t=154: Selected seed 195 with value = 0.7283 +Query 1/1: Action query time = 5.523 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6874 +t=170: Selected seed 195 with value = 0.6874 +Query 1/1: Action query time = 4.607 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9474 +t=186: Selected seed 195 with value = 0.9474 +Query 1/1: Action query time = 4.616 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7518 +t=202: Selected seed 195 with value = 0.7518 +Query 1/1: Action query time = 4.811 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9672 +t=218: Selected seed 195 with value = 0.9672 +Query 1/1: Action query time = 4.859 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8329 +t=234: Selected seed 195 with value = 0.8329 +Query 1/1: Action query time = 4.630 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9932 +t=250: Selected seed 195 with value = 0.9932 +Query 1/1: Action query time = 4.513 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8735 +t=266: Selected seed 195 with value = 0.8735 +Query 1/1: Action query time = 3.281 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9756 +t=282: Selected seed 195 with value = 0.9756 +Query 1/1: Action query time = 3.865 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s05/2026_08_01-13_02_59--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 3.734 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3675 +t=10: Selected seed 195 with value = 0.3675 +Query 1/1: Action query time = 6.000 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5346 +t=26: Selected seed 195 with value = 0.5346 +Query 1/1: Action query time = 5.476 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6504 +t=42: Selected seed 195 with value = 0.6504 +Query 1/1: Action query time = 4.783 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6573 +t=58: Selected seed 195 with value = 0.6573 +Query 1/1: Action query time = 5.256 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6982 +t=74: Selected seed 195 with value = 0.6982 +Query 1/1: Action query time = 5.226 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9022 +t=90: Selected seed 195 with value = 0.9022 +Query 1/1: Action query time = 5.622 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9595 +t=106: Selected seed 195 with value = 0.9595 +Query 1/1: Action query time = 5.716 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6947 +t=122: Selected seed 195 with value = 0.6947 +Query 1/1: Action query time = 5.225 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9310 +t=138: Selected seed 195 with value = 0.9310 +Query 1/1: Action query time = 4.854 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9863 +t=154: Selected seed 195 with value = 0.9863 +Query 1/1: Action query time = 4.847 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.117 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.913 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.785 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.484 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.311 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.572 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.702 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.916 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s05/2026_08_01-13_02_59--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..15864ed3e8bfb374f816f1ae0cb0657b98a0b083 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s07.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t3_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 2.342 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3954 +t=10: Selected seed 195 with value = 0.3954 +Query 1/1: Action query time = 2.844 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5734 +t=26: Selected seed 195 with value = 0.5734 +Query 1/1: Action query time = 5.573 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6847 +t=42: Selected seed 195 with value = 0.6847 +Query 1/1: Action query time = 5.080 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7205 +t=58: Selected seed 195 with value = 0.7205 +Query 1/1: Action query time = 4.720 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9803 +t=74: Selected seed 195 with value = 0.9803 +Query 1/1: Action query time = 4.854 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.321 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9189 +t=106: Selected seed 195 with value = 0.9189 +Query 1/1: Action query time = 5.238 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8261 +t=122: Selected seed 195 with value = 0.8261 +Query 1/1: Action query time = 5.271 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9619 +t=138: Selected seed 195 with value = 0.9619 +Query 1/1: Action query time = 5.214 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9473 +t=154: Selected seed 195 with value = 0.9473 +Query 1/1: Action query time = 5.106 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9872 +t=170: Selected seed 195 with value = 0.9872 +Query 1/1: Action query time = 5.105 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.117 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.223 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.463 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.193 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.217 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.044 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.856 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s07/2026_08_01-13_02_59--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 1.520 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3926 +t=10: Selected seed 195 with value = 0.3926 +Query 1/1: Action query time = 2.103 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7706 +t=26: Selected seed 195 with value = 0.7706 +Query 1/1: Action query time = 1.844 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6902 +t=42: Selected seed 195 with value = 0.6902 +Query 1/1: Action query time = 3.695 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7100 +t=58: Selected seed 195 with value = 0.7100 +Query 1/1: Action query time = 4.961 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9244 +t=74: Selected seed 195 with value = 0.9244 +Query 1/1: Action query time = 5.500 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.317 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.163 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9699 +t=122: Selected seed 195 with value = 0.9699 +Query 1/1: Action query time = 5.038 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8954 +t=138: Selected seed 195 with value = 0.8954 +Query 1/1: Action query time = 5.352 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9610 +t=154: Selected seed 195 with value = 0.9610 +Query 1/1: Action query time = 4.918 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9708 +t=170: Selected seed 195 with value = 0.9708 +Query 1/1: Action query time = 5.161 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9805 +t=186: Selected seed 195 with value = 0.9805 +Query 1/1: Action query time = 5.216 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.155 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.076 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.377 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.426 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=266: Selected seed 195 with value = 0.9982 +Query 1/1: Action query time = 5.467 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9975 +t=282: Selected seed 195 with value = 0.9975 +Query 1/1: Action query time = 4.253 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9975 +t=298: Selected seed 195 with value = 0.9975 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s07/2026_08_01-13_02_59--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.725 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3931 +t=10: Selected seed 195 with value = 0.3931 +Query 1/1: Action query time = 3.982 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5110 +t=26: Selected seed 195 with value = 0.5110 +Query 1/1: Action query time = 2.905 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6417 +t=42: Selected seed 195 with value = 0.6417 +Query 1/1: Action query time = 4.330 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6495 +t=58: Selected seed 195 with value = 0.6495 +Query 1/1: Action query time = 4.823 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6444 +t=74: Selected seed 195 with value = 0.6444 +Query 1/1: Action query time = 3.267 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6724 +t=90: Selected seed 195 with value = 0.6724 +Query 1/1: Action query time = 5.745 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6887 +t=106: Selected seed 195 with value = 0.6887 +Query 1/1: Action query time = 5.446 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7074 +t=122: Selected seed 195 with value = 0.7074 +Query 1/1: Action query time = 4.858 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7177 +t=138: Selected seed 195 with value = 0.7177 +Query 1/1: Action query time = 4.329 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7327 +t=154: Selected seed 195 with value = 0.7327 +Query 1/1: Action query time = 5.041 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7650 +t=170: Selected seed 195 with value = 0.7650 +Query 1/1: Action query time = 5.909 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7937 +t=186: Selected seed 195 with value = 0.7937 +Query 1/1: Action query time = 5.721 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8276 +t=202: Selected seed 195 with value = 0.8276 +Query 1/1: Action query time = 5.489 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8775 +t=218: Selected seed 195 with value = 0.8775 +Query 1/1: Action query time = 5.410 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9324 +t=234: Selected seed 195 with value = 0.9324 +Query 1/1: Action query time = 5.196 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9580 +t=250: Selected seed 195 with value = 0.9580 +Query 1/1: Action query time = 5.347 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9887 +t=266: Selected seed 195 with value = 0.9887 +Query 1/1: Action query time = 4.659 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.396 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s07/2026_08_01-13_02_59--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..72702abd1e6cd253e1edd4b42cdb2ff711aa2001 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_02_59--demochan200_t3_s14.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t3_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 6.685 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4155 +t=10: Selected seed 195 with value = 0.4155 +Query 1/1: Action query time = 5.798 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7870 +t=26: Selected seed 195 with value = 0.7870 +Query 1/1: Action query time = 5.466 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6854 +t=42: Selected seed 195 with value = 0.6854 +Query 1/1: Action query time = 5.006 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7201 +t=58: Selected seed 195 with value = 0.7201 +Query 1/1: Action query time = 5.090 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9492 +t=74: Selected seed 195 with value = 0.9492 +Query 1/1: Action query time = 5.116 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.125 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9125 +t=106: Selected seed 195 with value = 0.9125 +Query 1/1: Action query time = 5.010 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8026 +t=122: Selected seed 195 with value = 0.8026 +Query 1/1: Action query time = 4.906 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9197 +t=138: Selected seed 195 with value = 0.9197 +Query 1/1: Action query time = 4.862 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9136 +t=154: Selected seed 195 with value = 0.9136 +Query 1/1: Action query time = 5.029 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9732 +t=170: Selected seed 195 with value = 0.9732 +Query 1/1: Action query time = 5.233 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9972 +t=186: Selected seed 195 with value = 0.9972 +Query 1/1: Action query time = 5.298 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.894 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=218: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 5.308 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=234: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 5.338 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.628 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.892 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.892 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s14/2026_08_01-13_02_59--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 5.782 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3918 +t=10: Selected seed 195 with value = 0.3918 +Query 1/1: Action query time = 4.848 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5178 +t=26: Selected seed 195 with value = 0.5178 +Query 1/1: Action query time = 5.042 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6395 +t=42: Selected seed 195 with value = 0.6395 +Query 1/1: Action query time = 5.248 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6728 +t=58: Selected seed 195 with value = 0.6728 +Query 1/1: Action query time = 5.672 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6992 +t=74: Selected seed 195 with value = 0.6992 +Query 1/1: Action query time = 4.748 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7814 +t=90: Selected seed 195 with value = 0.7814 +Query 1/1: Action query time = 4.965 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8632 +t=106: Selected seed 195 with value = 0.8632 +Query 1/1: Action query time = 4.971 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9050 +t=122: Selected seed 195 with value = 0.9050 +Query 1/1: Action query time = 4.890 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8921 +t=138: Selected seed 195 with value = 0.8921 +Query 1/1: Action query time = 5.069 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9551 +t=154: Selected seed 195 with value = 0.9551 +Query 1/1: Action query time = 5.255 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9206 +t=170: Selected seed 195 with value = 0.9206 +Query 1/1: Action query time = 5.373 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9948 +t=186: Selected seed 195 with value = 0.9948 +Query 1/1: Action query time = 5.081 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=202: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 4.792 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9929 +t=218: Selected seed 195 with value = 0.9929 +Query 1/1: Action query time = 4.864 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9897 +t=234: Selected seed 195 with value = 0.9897 +Query 1/1: Action query time = 3.554 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9847 +t=250: Selected seed 195 with value = 0.9847 +Query 1/1: Action query time = 3.912 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9882 +t=266: Selected seed 195 with value = 0.9882 +Query 1/1: Action query time = 4.942 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=282: Selected seed 195 with value = 0.9958 +Query 1/1: Action query time = 4.345 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=298: Selected seed 195 with value = 0.9997 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s14/2026_08_01-13_02_59--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 3.968 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3653 +t=10: Selected seed 195 with value = 0.3653 +Query 1/1: Action query time = 4.367 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5416 +t=26: Selected seed 195 with value = 0.5416 +Query 1/1: Action query time = 5.968 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6569 +t=42: Selected seed 195 with value = 0.6569 +Query 1/1: Action query time = 5.809 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6628 +t=58: Selected seed 195 with value = 0.6628 +Query 1/1: Action query time = 5.716 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6972 +t=74: Selected seed 195 with value = 0.6972 +Query 1/1: Action query time = 5.242 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7830 +t=90: Selected seed 195 with value = 0.7830 +Query 1/1: Action query time = 4.689 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9199 +t=106: Selected seed 195 with value = 0.9199 +Query 1/1: Action query time = 4.417 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9798 +t=122: Selected seed 195 with value = 0.9798 +Query 1/1: Action query time = 4.906 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.144 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.986 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.764 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.989 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.650 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.735 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.122 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.754 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.705 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.483 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s14/2026_08_01-13_02_59--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_03_00--demochan200_t3_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_03_00--demochan200_t3_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..c685f7ab02d6c03e597c25d4bfa8c9b7d95bc0cb --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_03_00--demochan200_t3_s01.txt @@ -0,0 +1,344 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t3_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 3.336 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3962 +t=10: Selected seed 195 with value = 0.3962 +Query 1/1: Action query time = 4.471 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7724 +t=26: Selected seed 195 with value = 0.7724 +Query 1/1: Action query time = 4.833 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6908 +t=42: Selected seed 195 with value = 0.6908 +Query 1/1: Action query time = 4.905 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7050 +t=58: Selected seed 195 with value = 0.7050 +Query 1/1: Action query time = 5.129 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9582 +t=74: Selected seed 195 with value = 0.9582 +Query 1/1: Action query time = 5.048 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7492 +t=90: Selected seed 195 with value = 0.7492 +Query 1/1: Action query time = 5.109 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7884 +t=106: Selected seed 195 with value = 0.7884 +Query 1/1: Action query time = 5.073 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8058 +t=122: Selected seed 195 with value = 0.8058 +Query 1/1: Action query time = 5.047 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7385 +t=138: Selected seed 195 with value = 0.7385 +Query 1/1: Action query time = 4.962 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7342 +t=154: Selected seed 195 with value = 0.7342 +Query 1/1: Action query time = 5.123 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7400 +t=170: Selected seed 195 with value = 0.7400 +Query 1/1: Action query time = 5.110 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9329 +t=186: Selected seed 195 with value = 0.9329 +Query 1/1: Action query time = 5.102 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8402 +t=202: Selected seed 195 with value = 0.8402 +Query 1/1: Action query time = 4.783 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9875 +t=218: Selected seed 195 with value = 0.9875 +Query 1/1: Action query time = 5.044 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9434 +t=234: Selected seed 195 with value = 0.9434 +Query 1/1: Action query time = 5.021 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9612 +t=250: Selected seed 195 with value = 0.9612 +Query 1/1: Action query time = 5.003 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9852 +t=266: Selected seed 195 with value = 0.9852 +Query 1/1: Action query time = 4.029 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=282: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 3.978 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s01/2026_08_01-13_03_00--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 1.739 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4310 +t=10: Selected seed 195 with value = 0.4310 +Query 1/1: Action query time = 1.800 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9294 +t=26: Selected seed 195 with value = 0.9294 +Query 1/1: Action query time = 3.943 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5786 +t=42: Selected seed 195 with value = 0.5786 +Query 1/1: Action query time = 5.426 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6884 +t=58: Selected seed 195 with value = 0.6884 +Query 1/1: Action query time = 5.721 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5604 +t=74: Selected seed 195 with value = 0.5604 +Query 1/1: Action query time = 5.507 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8436 +t=90: Selected seed 195 with value = 0.8436 +Query 1/1: Action query time = 5.109 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6081 +t=106: Selected seed 195 with value = 0.6081 +Query 1/1: Action query time = 4.997 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6173 +t=122: Selected seed 195 with value = 0.6173 +Query 1/1: Action query time = 5.278 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7484 +t=138: Selected seed 195 with value = 0.7484 +Query 1/1: Action query time = 4.827 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6832 +t=154: Selected seed 195 with value = 0.6832 +Query 1/1: Action query time = 5.136 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6134 +t=170: Selected seed 195 with value = 0.6134 +Query 1/1: Action query time = 5.049 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5837 +t=186: Selected seed 195 with value = 0.5837 +Query 1/1: Action query time = 5.104 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5917 +t=202: Selected seed 195 with value = 0.5917 +Query 1/1: Action query time = 5.210 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5976 +t=218: Selected seed 195 with value = 0.5976 +Query 1/1: Action query time = 5.360 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5868 +t=234: Selected seed 195 with value = 0.5868 +Query 1/1: Action query time = 5.514 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6585 +t=250: Selected seed 195 with value = 0.6585 +Query 1/1: Action query time = 5.379 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6679 +t=266: Selected seed 195 with value = 0.6679 +Query 1/1: Action query time = 4.398 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6164 +t=282: Selected seed 195 with value = 0.6164 +Query 1/1: Action query time = 4.001 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6161 +t=298: Selected seed 195 with value = 0.6161 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s01/2026_08_01-13_03_00--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 3.927 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3674 +t=10: Selected seed 195 with value = 0.3674 +Query 1/1: Action query time = 2.603 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5140 +t=26: Selected seed 195 with value = 0.5140 +Query 1/1: Action query time = 3.377 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6852 +t=42: Selected seed 195 with value = 0.6852 +Query 1/1: Action query time = 3.071 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6919 +t=58: Selected seed 195 with value = 0.6919 +Query 1/1: Action query time = 4.992 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7530 +t=74: Selected seed 195 with value = 0.7530 +Query 1/1: Action query time = 5.550 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9519 +t=90: Selected seed 195 with value = 0.9519 +Query 1/1: Action query time = 5.100 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9947 +t=106: Selected seed 195 with value = 0.9947 +Query 1/1: Action query time = 5.065 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9698 +t=122: Selected seed 195 with value = 0.9698 +Query 1/1: Action query time = 5.079 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8464 +t=138: Selected seed 195 with value = 0.8464 +Query 1/1: Action query time = 5.603 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8674 +t=154: Selected seed 195 with value = 0.8674 +Query 1/1: Action query time = 4.880 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9597 +t=170: Selected seed 195 with value = 0.9597 +Query 1/1: Action query time = 4.398 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9690 +t=186: Selected seed 195 with value = 0.9690 +Query 1/1: Action query time = 5.337 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9579 +t=202: Selected seed 195 with value = 0.9579 +Query 1/1: Action query time = 5.036 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9739 +t=218: Selected seed 195 with value = 0.9739 +Query 1/1: Action query time = 4.517 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9872 +t=234: Selected seed 195 with value = 0.9872 +Query 1/1: Action query time = 4.980 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.319 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.557 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.343 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s01/2026_08_01-13_03_00--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 3.000 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3974 +t=10: Selected seed 195 with value = 0.3974 +Query 1/1: Action query time = 1.841 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5695 +t=26: Selected seed 195 with value = 0.5695 +Query 1/1: Action query time = 0.993 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6476 +t=42: Selected seed 195 with value = 0.6476 +Query 1/1: Action query time = 1.017 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6826 +t=58: Selected seed 195 with value = 0.6826 +Query 1/1: Action query time = 1.012 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7190 +t=74: Selected seed 195 with value = 0.7190 +Query 1/1: Action query time = 0.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9941 +t=90: Selected seed 195 with value = 0.9941 +Query 1/1: Action query time = 1.006 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=106: Selected seed 195 with value = 0.9909 +Query 1/1: Action query time = 1.651 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8560 +t=122: Selected seed 195 with value = 0.8560 +Query 1/1: Action query time = 1.615 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7807 +t=138: Selected seed 195 with value = 0.7807 +Query 1/1: Action query time = 1.525 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9681 +t=154: Selected seed 195 with value = 0.9681 +Query 1/1: Action query time = 1.496 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9546 +t=170: Selected seed 195 with value = 0.9546 +Query 1/1: Action query time = 1.489 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9852 +t=186: Selected seed 195 with value = 0.9852 +Query 1/1: Action query time = 1.486 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.476 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.446 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.464 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.456 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.418 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.408 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s01/2026_08_01-13_03_00--with_future_img--episode=4--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 4 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_03_00--demochan200_t3_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_03_00--demochan200_t3_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..da61cee6390039b4d0335ec2826fe529b655c6e2 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_03_00--demochan200_t3_s11.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t3_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 5.993 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3950 +t=10: Selected seed 195 with value = 0.3950 +Query 1/1: Action query time = 3.666 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=26: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 4.899 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6753 +t=42: Selected seed 195 with value = 0.6753 +Query 1/1: Action query time = 5.606 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6744 +t=58: Selected seed 195 with value = 0.6744 +Query 1/1: Action query time = 5.282 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7035 +t=74: Selected seed 195 with value = 0.7035 +Query 1/1: Action query time = 5.008 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9177 +t=90: Selected seed 195 with value = 0.9177 +Query 1/1: Action query time = 5.067 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9395 +t=106: Selected seed 195 with value = 0.9395 +Query 1/1: Action query time = 5.063 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9792 +t=122: Selected seed 195 with value = 0.9792 +Query 1/1: Action query time = 5.070 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=138: Selected seed 195 with value = 0.9809 +Query 1/1: Action query time = 5.279 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.271 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.011 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.816 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.312 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.214 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.128 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.997 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.328 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.947 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s11/2026_08_01-13_03_00--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.659 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3835 +t=10: Selected seed 195 with value = 0.3835 +Query 1/1: Action query time = 4.848 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6108 +t=26: Selected seed 195 with value = 0.6108 +Query 1/1: Action query time = 5.084 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6832 +t=42: Selected seed 195 with value = 0.6832 +Query 1/1: Action query time = 5.042 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7116 +t=58: Selected seed 195 with value = 0.7116 +Query 1/1: Action query time = 4.968 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9427 +t=74: Selected seed 195 with value = 0.9427 +Query 1/1: Action query time = 4.820 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=90: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 5.651 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9539 +t=106: Selected seed 195 with value = 0.9539 +Query 1/1: Action query time = 5.455 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8082 +t=122: Selected seed 195 with value = 0.8082 +Query 1/1: Action query time = 5.356 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9430 +t=138: Selected seed 195 with value = 0.9430 +Query 1/1: Action query time = 5.440 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9467 +t=154: Selected seed 195 with value = 0.9467 +Query 1/1: Action query time = 4.763 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9747 +t=170: Selected seed 195 with value = 0.9747 +Query 1/1: Action query time = 4.625 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9883 +t=186: Selected seed 195 with value = 0.9883 +Query 1/1: Action query time = 4.863 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=202: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 5.056 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.320 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.170 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.043 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=266: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 4.671 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9931 +t=282: Selected seed 195 with value = 0.9931 +Query 1/1: Action query time = 4.705 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9863 +t=298: Selected seed 195 with value = 0.9863 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s11/2026_08_01-13_03_00--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 3.896 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4028 +t=10: Selected seed 195 with value = 0.4028 +Query 1/1: Action query time = 5.028 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6195 +t=26: Selected seed 195 with value = 0.6195 +Query 1/1: Action query time = 6.466 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6941 +t=42: Selected seed 195 with value = 0.6941 +Query 1/1: Action query time = 5.730 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6795 +t=58: Selected seed 195 with value = 0.6795 +Query 1/1: Action query time = 4.386 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7299 +t=74: Selected seed 195 with value = 0.7299 +Query 1/1: Action query time = 4.413 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5610 +t=90: Selected seed 195 with value = 0.5610 +Query 1/1: Action query time = 4.587 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8413 +t=106: Selected seed 195 with value = 0.8413 +Query 1/1: Action query time = 5.074 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6071 +t=122: Selected seed 195 with value = 0.6071 +Query 1/1: Action query time = 5.081 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7519 +t=138: Selected seed 195 with value = 0.7519 +Query 1/1: Action query time = 4.892 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7376 +t=154: Selected seed 195 with value = 0.7376 +Query 1/1: Action query time = 5.026 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5784 +t=170: Selected seed 195 with value = 0.5784 +Query 1/1: Action query time = 5.430 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7571 +t=186: Selected seed 195 with value = 0.7571 +Query 1/1: Action query time = 5.279 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5822 +t=202: Selected seed 195 with value = 0.5822 +Query 1/1: Action query time = 4.734 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7429 +t=218: Selected seed 195 with value = 0.7429 +Query 1/1: Action query time = 3.850 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5954 +t=234: Selected seed 195 with value = 0.5954 +Query 1/1: Action query time = 3.933 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7528 +t=250: Selected seed 195 with value = 0.7528 +Query 1/1: Action query time = 3.824 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6036 +t=266: Selected seed 195 with value = 0.6036 +Query 1/1: Action query time = 3.659 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7601 +t=282: Selected seed 195 with value = 0.7601 +Query 1/1: Action query time = 3.529 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6026 +t=298: Selected seed 195 with value = 0.6026 +Saved rollout MP4 at path ./rollouts/demochan200_t3_s11/2026_08_01-13_03_00--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_47--demochan200_t6_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_47--demochan200_t6_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..cbab6b5b6439dd0048ee9103380064ab5849658f --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_47--demochan200_t6_s03.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t6_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 1.769 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4239 +t=10: Selected seed 195 with value = 0.4239 +Query 1/1: Action query time = 2.448 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5167 +t=26: Selected seed 195 with value = 0.5167 +Query 1/1: Action query time = 3.678 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6143 +t=42: Selected seed 195 with value = 0.6143 +Query 1/1: Action query time = 5.006 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7596 +t=58: Selected seed 195 with value = 0.7596 +Query 1/1: Action query time = 5.296 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8752 +t=74: Selected seed 195 with value = 0.8752 +Query 1/1: Action query time = 4.738 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9736 +t=90: Selected seed 195 with value = 0.9736 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s03/2026_08_01-13_09_47--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 5.813 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4495 +t=10: Selected seed 195 with value = 0.4495 +Query 1/1: Action query time = 4.990 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5108 +t=26: Selected seed 195 with value = 0.5108 +Query 1/1: Action query time = 4.388 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6150 +t=42: Selected seed 195 with value = 0.6150 +Query 1/1: Action query time = 5.087 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6745 +t=58: Selected seed 195 with value = 0.6745 +Query 1/1: Action query time = 5.049 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5666 +t=74: Selected seed 195 with value = 0.5666 +Query 1/1: Action query time = 5.199 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6504 +t=90: Selected seed 195 with value = 0.6504 +Query 1/1: Action query time = 5.286 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6578 +t=106: Selected seed 195 with value = 0.6578 +Query 1/1: Action query time = 5.334 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6863 +t=122: Selected seed 195 with value = 0.6863 +Query 1/1: Action query time = 5.088 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8095 +t=138: Selected seed 195 with value = 0.8095 +Query 1/1: Action query time = 5.194 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9266 +t=154: Selected seed 195 with value = 0.9266 +Query 1/1: Action query time = 4.987 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9301 +t=170: Selected seed 195 with value = 0.9301 +Query 1/1: Action query time = 5.088 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=186: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 4.637 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=202: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 4.363 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.553 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.269 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.039 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.770 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.599 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s03/2026_08_01-13_09_47--with_future_img--episode=2--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.366 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4219 +t=10: Selected seed 195 with value = 0.4219 +Query 1/1: Action query time = 5.270 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5021 +t=26: Selected seed 195 with value = 0.5021 +Query 1/1: Action query time = 3.562 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6227 +t=42: Selected seed 195 with value = 0.6227 +Query 1/1: Action query time = 2.667 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7642 +t=58: Selected seed 195 with value = 0.7642 +Query 1/1: Action query time = 4.505 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8685 +t=74: Selected seed 195 with value = 0.8685 +Query 1/1: Action query time = 4.700 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9589 +t=90: Selected seed 195 with value = 0.9589 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s03/2026_08_01-13_09_47--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_48--demochan200_t6_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_48--demochan200_t6_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..29c5ebb48f3e06ec035451a6be43e5944cf87fd6 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_48--demochan200_t6_s02.txt @@ -0,0 +1,209 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t6_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 5.178 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4428 +t=10: Selected seed 195 with value = 0.4428 +Query 1/1: Action query time = 5.259 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4886 +t=26: Selected seed 195 with value = 0.4886 +Query 1/1: Action query time = 5.610 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6182 +t=42: Selected seed 195 with value = 0.6182 +Query 1/1: Action query time = 5.008 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5419 +t=58: Selected seed 195 with value = 0.5419 +Query 1/1: Action query time = 4.594 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5581 +t=74: Selected seed 195 with value = 0.5581 +Query 1/1: Action query time = 5.325 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6325 +t=90: Selected seed 195 with value = 0.6325 +Query 1/1: Action query time = 5.628 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6077 +t=106: Selected seed 195 with value = 0.6077 +Query 1/1: Action query time = 5.232 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7798 +t=122: Selected seed 195 with value = 0.7798 +Query 1/1: Action query time = 5.628 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8514 +t=138: Selected seed 195 with value = 0.8514 +Query 1/1: Action query time = 5.107 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9693 +t=154: Selected seed 195 with value = 0.9693 +Query 1/1: Action query time = 4.998 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=170: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 4.801 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9967 +t=186: Selected seed 195 with value = 0.9967 +Query 1/1: Action query time = 5.132 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9949 +t=202: Selected seed 195 with value = 0.9949 +Query 1/1: Action query time = 4.750 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.921 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.068 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.093 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.284 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.261 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=298: Selected seed 195 with value = 0.9977 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s02/2026_08_01-13_09_48--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 5.589 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4559 +t=10: Selected seed 195 with value = 0.4559 +Query 1/1: Action query time = 4.797 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4955 +t=26: Selected seed 195 with value = 0.4955 +Query 1/1: Action query time = 4.566 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6205 +t=42: Selected seed 195 with value = 0.6205 +Query 1/1: Action query time = 4.460 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7631 +t=58: Selected seed 195 with value = 0.7631 +Query 1/1: Action query time = 4.625 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8805 +t=74: Selected seed 195 with value = 0.8805 +Query 1/1: Action query time = 4.216 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=90: Selected seed 195 with value = 0.9809 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s02/2026_08_01-13_09_48--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 3.708 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4583 +t=10: Selected seed 195 with value = 0.4583 +Query 1/1: Action query time = 5.519 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5102 +t=26: Selected seed 195 with value = 0.5102 +Query 1/1: Action query time = 4.908 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5986 +t=42: Selected seed 195 with value = 0.5986 +Query 1/1: Action query time = 3.866 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8460 +t=58: Selected seed 195 with value = 0.8460 +Query 1/1: Action query time = 4.265 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8953 +t=74: Selected seed 195 with value = 0.8953 +Query 1/1: Action query time = 4.386 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=90: Selected seed 195 with value = 0.9963 +Query 1/1: Action query time = 4.055 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9935 +t=106: Selected seed 195 with value = 0.9935 +Query 1/1: Action query time = 3.008 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9889 +t=122: Selected seed 195 with value = 0.9889 +Query 1/1: Action query time = 2.410 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9916 +t=138: Selected seed 195 with value = 0.9916 +Query 1/1: Action query time = 2.378 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9927 +t=154: Selected seed 195 with value = 0.9927 +Query 1/1: Action query time = 2.577 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=170: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 2.606 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9972 +t=186: Selected seed 195 with value = 0.9972 +Query 1/1: Action query time = 2.910 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=202: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 3.050 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=218: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 3.325 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=234: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 2.803 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=250: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 2.002 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=266: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 1.931 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=282: Selected seed 195 with value = 0.9982 +Query 1/1: Action query time = 1.219 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=298: Selected seed 195 with value = 0.9971 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s02/2026_08_01-13_09_48--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_48--demochan200_t6_s09.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_48--demochan200_t6_s09.txt new file mode 100644 index 0000000000000000000000000000000000000000..91c3a3eddcdc15820e30a642900caeb2bf76225a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_48--demochan200_t6_s09.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t6_s09', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='9,25,41', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 2.549 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4352 +t=10: Selected seed 195 with value = 0.4352 +Query 1/1: Action query time = 4.843 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5073 +t=26: Selected seed 195 with value = 0.5073 +Query 1/1: Action query time = 5.524 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5716 +t=42: Selected seed 195 with value = 0.5716 +Query 1/1: Action query time = 4.914 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6929 +t=58: Selected seed 195 with value = 0.6929 +Query 1/1: Action query time = 4.582 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8289 +t=74: Selected seed 195 with value = 0.8289 +Query 1/1: Action query time = 4.053 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9275 +t=90: Selected seed 195 with value = 0.9275 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s09/2026_08_01-13_09_48--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 3.884 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4502 +t=10: Selected seed 195 with value = 0.4502 +Query 1/1: Action query time = 4.199 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4886 +t=26: Selected seed 195 with value = 0.4886 +Query 1/1: Action query time = 4.463 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6351 +t=42: Selected seed 195 with value = 0.6351 +Query 1/1: Action query time = 4.969 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7337 +t=58: Selected seed 195 with value = 0.7337 +Query 1/1: Action query time = 4.617 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8358 +t=74: Selected seed 195 with value = 0.8358 +Query 1/1: Action query time = 4.868 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9352 +t=90: Selected seed 195 with value = 0.9352 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s09/2026_08_01-13_09_48--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.229 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4149 +t=10: Selected seed 195 with value = 0.4149 +Query 1/1: Action query time = 5.340 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4969 +t=26: Selected seed 195 with value = 0.4969 +Query 1/1: Action query time = 5.383 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5809 +t=42: Selected seed 195 with value = 0.5809 +Query 1/1: Action query time = 5.125 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6862 +t=58: Selected seed 195 with value = 0.6862 +Query 1/1: Action query time = 5.205 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8020 +t=74: Selected seed 195 with value = 0.8020 +Query 1/1: Action query time = 5.196 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9157 +t=90: Selected seed 195 with value = 0.9157 +Query 1/1: Action query time = 4.699 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9933 +t=106: Selected seed 195 with value = 0.9933 +Query 1/1: Action query time = 2.578 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=122: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 4.145 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=138: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 6.020 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.528 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.195 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.858 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.968 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.808 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.698 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.357 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.728 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=282: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 5.060 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9970 +t=298: Selected seed 195 with value = 0.9970 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s09/2026_08_01-13_09_48--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_49--demochan200_t6_s12.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_49--demochan200_t6_s12.txt new file mode 100644 index 0000000000000000000000000000000000000000..cab410804a5ecbfb3b5780734d15272f5999e9a0 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_09_49--demochan200_t6_s12.txt @@ -0,0 +1,209 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t6_s12', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='12,28,44', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 5.133 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4545 +t=10: Selected seed 195 with value = 0.4545 +Query 1/1: Action query time = 4.008 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5349 +t=26: Selected seed 195 with value = 0.5349 +Query 1/1: Action query time = 4.130 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6262 +t=42: Selected seed 195 with value = 0.6262 +Query 1/1: Action query time = 5.101 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6553 +t=58: Selected seed 195 with value = 0.6553 +Query 1/1: Action query time = 5.466 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5566 +t=74: Selected seed 195 with value = 0.5566 +Query 1/1: Action query time = 5.060 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6264 +t=90: Selected seed 195 with value = 0.6264 +Query 1/1: Action query time = 4.115 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8309 +t=106: Selected seed 195 with value = 0.8309 +Query 1/1: Action query time = 4.901 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8488 +t=122: Selected seed 195 with value = 0.8488 +Query 1/1: Action query time = 5.027 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8324 +t=138: Selected seed 195 with value = 0.8324 +Query 1/1: Action query time = 4.587 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9712 +t=154: Selected seed 195 with value = 0.9712 +Query 1/1: Action query time = 5.168 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8947 +t=170: Selected seed 195 with value = 0.8947 +Query 1/1: Action query time = 5.131 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8771 +t=186: Selected seed 195 with value = 0.8771 +Query 1/1: Action query time = 4.266 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9543 +t=202: Selected seed 195 with value = 0.9543 +Query 1/1: Action query time = 4.329 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.396 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.768 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.117 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.080 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.869 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s12/2026_08_01-13_09_49--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 4.174 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4483 +t=10: Selected seed 195 with value = 0.4483 +Query 1/1: Action query time = 5.855 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4800 +t=26: Selected seed 195 with value = 0.4800 +Query 1/1: Action query time = 4.912 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5890 +t=42: Selected seed 195 with value = 0.5890 +Query 1/1: Action query time = 4.805 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7031 +t=58: Selected seed 195 with value = 0.7031 +Query 1/1: Action query time = 4.916 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8232 +t=74: Selected seed 195 with value = 0.8232 +Query 1/1: Action query time = 4.303 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9432 +t=90: Selected seed 195 with value = 0.9432 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s12/2026_08_01-13_09_49--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 2.777 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4339 +t=10: Selected seed 195 with value = 0.4339 +Query 1/1: Action query time = 5.054 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5050 +t=26: Selected seed 195 with value = 0.5050 +Query 1/1: Action query time = 3.140 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5982 +t=42: Selected seed 195 with value = 0.5982 +Query 1/1: Action query time = 4.870 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5153 +t=58: Selected seed 195 with value = 0.5153 +Query 1/1: Action query time = 4.747 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5883 +t=74: Selected seed 195 with value = 0.5883 +Query 1/1: Action query time = 3.723 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5214 +t=90: Selected seed 195 with value = 0.5214 +Query 1/1: Action query time = 4.213 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8087 +t=106: Selected seed 195 with value = 0.8087 +Query 1/1: Action query time = 2.677 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9052 +t=122: Selected seed 195 with value = 0.9052 +Query 1/1: Action query time = 2.449 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9933 +t=138: Selected seed 195 with value = 0.9933 +Query 1/1: Action query time = 2.993 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9935 +t=154: Selected seed 195 with value = 0.9935 +Query 1/1: Action query time = 2.401 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9936 +t=170: Selected seed 195 with value = 0.9936 +Query 1/1: Action query time = 2.377 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=186: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 1.902 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.674 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.841 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8671 +t=234: Selected seed 195 with value = 0.8671 +Query 1/1: Action query time = 1.837 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8920 +t=250: Selected seed 195 with value = 0.8920 +Query 1/1: Action query time = 1.652 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.833 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.678 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9768 +t=298: Selected seed 195 with value = 0.9768 +Saved rollout MP4 at path ./rollouts/demochan200_t6_s12/2026_08_01-13_09_49--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_41_21--demochan200_t0_s10.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_41_21--demochan200_t0_s10.txt new file mode 100644 index 0000000000000000000000000000000000000000..76b48f4bcfc83bd7ddbc9cf63a25a33e758a1939 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_41_21--demochan200_t0_s10.txt @@ -0,0 +1,185 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t0_s10', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='10,26,42', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3405 +t=10: Selected seed 195 with value = 0.3405 +Query 1/1: Action query time = 4.934 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3963 +t=26: Selected seed 195 with value = 0.3963 +Query 1/1: Action query time = 4.930 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5050 +t=42: Selected seed 195 with value = 0.5050 +Query 1/1: Action query time = 4.628 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5892 +t=58: Selected seed 195 with value = 0.5892 +Query 1/1: Action query time = 5.778 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6877 +t=74: Selected seed 195 with value = 0.6877 +Query 1/1: Action query time = 5.177 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9452 +t=90: Selected seed 195 with value = 0.9452 +Query 1/1: Action query time = 4.168 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7960 +t=106: Selected seed 195 with value = 0.7960 +Query 1/1: Action query time = 4.895 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7393 +t=122: Selected seed 195 with value = 0.7393 +Query 1/1: Action query time = 4.698 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7657 +t=138: Selected seed 195 with value = 0.7657 +Query 1/1: Action query time = 4.147 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7743 +t=154: Selected seed 195 with value = 0.7743 +Query 1/1: Action query time = 3.698 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8701 +t=170: Selected seed 195 with value = 0.8701 +Query 1/1: Action query time = 4.081 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8544 +t=186: Selected seed 195 with value = 0.8544 +Query 1/1: Action query time = 5.091 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8494 +t=202: Selected seed 195 with value = 0.8494 +Query 1/1: Action query time = 5.350 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8447 +t=218: Selected seed 195 with value = 0.8447 +Query 1/1: Action query time = 5.618 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8376 +t=234: Selected seed 195 with value = 0.8376 +Query 1/1: Action query time = 5.082 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8315 +t=250: Selected seed 195 with value = 0.8315 +Query 1/1: Action query time = 4.911 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8207 +t=266: Selected seed 195 with value = 0.8207 +Query 1/1: Action query time = 4.377 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8135 +t=282: Selected seed 195 with value = 0.8135 +Query 1/1: Action query time = 4.130 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7843 +t=298: Selected seed 195 with value = 0.7843 +Saved rollout MP4 at path ./rollouts/demochan200_t0_s10/2026_08_01-13_41_21--with_future_img--episode=1--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.037 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3175 +t=10: Selected seed 195 with value = 0.3175 +Query 1/1: Action query time = 3.711 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3900 +t=26: Selected seed 195 with value = 0.3900 +Query 1/1: Action query time = 5.027 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4615 +t=42: Selected seed 195 with value = 0.4615 +Query 1/1: Action query time = 4.650 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5389 +t=58: Selected seed 195 with value = 0.5389 +Query 1/1: Action query time = 5.241 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6314 +t=74: Selected seed 195 with value = 0.6314 +Query 1/1: Action query time = 5.387 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7200 +t=90: Selected seed 195 with value = 0.7200 +Query 1/1: Action query time = 5.064 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9483 +t=106: Selected seed 195 with value = 0.9483 +Query 1/1: Action query time = 4.935 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9868 +t=122: Selected seed 195 with value = 0.9868 +Query 1/1: Action query time = 4.792 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9877 +t=138: Selected seed 195 with value = 0.9877 +Query 1/1: Action query time = 3.953 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t0_s10/2026_08_01-13_41_21--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.972 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3827 +t=10: Selected seed 195 with value = 0.3827 +Query 1/1: Action query time = 2.810 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4236 +t=26: Selected seed 195 with value = 0.4236 +Query 1/1: Action query time = 2.608 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4749 +t=42: Selected seed 195 with value = 0.4749 +Query 1/1: Action query time = 2.756 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5655 +t=58: Selected seed 195 with value = 0.5655 +Query 1/1: Action query time = 3.243 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6816 +t=74: Selected seed 195 with value = 0.6816 +Query 1/1: Action query time = 2.867 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9568 +t=90: Selected seed 195 with value = 0.9568 +Query 1/1: Action query time = 3.045 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=106: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 1.738 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9881 +t=122: Selected seed 195 with value = 0.9881 +Query 1/1: Action query time = 1.969 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t0_s10/2026_08_01-13_41_21--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_41_21--demochan200_t0_s13.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_41_21--demochan200_t0_s13.txt new file mode 100644 index 0000000000000000000000000000000000000000..eddd7c237d4b273dca14dd1b387f681bc3e2896c --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_41_21--demochan200_t0_s13.txt @@ -0,0 +1,153 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t0_s13', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='13,29,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.032 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3618 +t=10: Selected seed 195 with value = 0.3618 +Query 1/1: Action query time = 4.634 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4020 +t=26: Selected seed 195 with value = 0.4020 +Query 1/1: Action query time = 4.899 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4396 +t=42: Selected seed 195 with value = 0.4396 +Query 1/1: Action query time = 4.539 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5312 +t=58: Selected seed 195 with value = 0.5312 +Query 1/1: Action query time = 4.179 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6425 +t=74: Selected seed 195 with value = 0.6425 +Query 1/1: Action query time = 4.860 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7061 +t=90: Selected seed 195 with value = 0.7061 +Query 1/1: Action query time = 5.124 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9501 +t=106: Selected seed 195 with value = 0.9501 +Query 1/1: Action query time = 5.394 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9821 +t=122: Selected seed 195 with value = 0.9821 +Query 1/1: Action query time = 4.352 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.956 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t0_s13/2026_08_01-13_41_21--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.356 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3676 +t=10: Selected seed 195 with value = 0.3676 +Query 1/1: Action query time = 4.427 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3685 +t=26: Selected seed 195 with value = 0.3685 +Query 1/1: Action query time = 4.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4161 +t=42: Selected seed 195 with value = 0.4161 +Query 1/1: Action query time = 5.020 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5196 +t=58: Selected seed 195 with value = 0.5196 +Query 1/1: Action query time = 5.202 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6382 +t=74: Selected seed 195 with value = 0.6382 +Query 1/1: Action query time = 5.285 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8694 +t=90: Selected seed 195 with value = 0.8694 +Query 1/1: Action query time = 4.607 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9172 +t=106: Selected seed 195 with value = 0.9172 +Query 1/1: Action query time = 3.129 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9584 +t=122: Selected seed 195 with value = 0.9584 +Query 1/1: Action query time = 2.893 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9672 +t=138: Selected seed 195 with value = 0.9672 +Query 1/1: Action query time = 3.258 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=154: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 4.678 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t0_s13/2026_08_01-13_41_21--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.555 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3538 +t=10: Selected seed 195 with value = 0.3538 +Query 1/1: Action query time = 5.224 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3825 +t=26: Selected seed 195 with value = 0.3825 +Query 1/1: Action query time = 5.217 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4054 +t=42: Selected seed 195 with value = 0.4054 +Query 1/1: Action query time = 4.948 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4811 +t=58: Selected seed 195 with value = 0.4811 +Query 1/1: Action query time = 4.191 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6294 +t=74: Selected seed 195 with value = 0.6294 +Query 1/1: Action query time = 5.480 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7140 +t=90: Selected seed 195 with value = 0.7140 +Query 1/1: Action query time = 4.385 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9354 +t=106: Selected seed 195 with value = 0.9354 +Query 1/1: Action query time = 4.500 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=122: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 3.659 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t0_s13/2026_08_01-13_41_21--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_46_15--demochan200_t2_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_46_15--demochan200_t2_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..3c93ceb799c908a2bfe110d6b71c4cf6983e8200 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_46_15--demochan200_t2_s02.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t2_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.831 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4996 +t=10: Selected seed 195 with value = 0.4996 +Query 1/1: Action query time = 4.620 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5615 +t=26: Selected seed 195 with value = 0.5615 +Query 1/1: Action query time = 4.724 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7418 +t=42: Selected seed 195 with value = 0.7418 +Query 1/1: Action query time = 5.262 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7743 +t=58: Selected seed 195 with value = 0.7743 +Query 1/1: Action query time = 5.599 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9458 +t=74: Selected seed 195 with value = 0.9458 +Query 1/1: Action query time = 3.750 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.324 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.250 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.287 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.536 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.881 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.702 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.823 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.165 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.394 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.336 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.320 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9437 +t=266: Selected seed 195 with value = 0.9437 +Query 1/1: Action query time = 5.472 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8712 +t=282: Selected seed 195 with value = 0.8712 +Query 1/1: Action query time = 4.858 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8750 +t=298: Selected seed 195 with value = 0.8750 +Saved rollout MP4 at path ./rollouts/demochan200_t2_s02/2026_08_01-13_46_15--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.933 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5143 +t=10: Selected seed 195 with value = 0.5143 +Query 1/1: Action query time = 3.473 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5923 +t=26: Selected seed 195 with value = 0.5923 +Query 1/1: Action query time = 5.438 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8168 +t=42: Selected seed 195 with value = 0.8168 +Query 1/1: Action query time = 3.244 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8407 +t=58: Selected seed 195 with value = 0.8407 +Query 1/1: Action query time = 3.261 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9739 +t=74: Selected seed 195 with value = 0.9739 +Query 1/1: Action query time = 4.288 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.398 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.404 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.454 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.279 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.291 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.184 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.085 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.618 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.356 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.773 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.028 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.435 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9708 +t=282: Selected seed 195 with value = 0.9708 +Query 1/1: Action query time = 3.835 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8613 +t=298: Selected seed 195 with value = 0.8613 +Saved rollout MP4 at path ./rollouts/demochan200_t2_s02/2026_08_01-13_46_15--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.116 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4378 +t=10: Selected seed 195 with value = 0.4378 +Query 1/1: Action query time = 4.048 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5340 +t=26: Selected seed 195 with value = 0.5340 +Query 1/1: Action query time = 4.735 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7465 +t=42: Selected seed 195 with value = 0.7465 +Query 1/1: Action query time = 3.170 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7861 +t=58: Selected seed 195 with value = 0.7861 +Query 1/1: Action query time = 2.660 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8203 +t=74: Selected seed 195 with value = 0.8203 +Query 1/1: Action query time = 3.370 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9265 +t=90: Selected seed 195 with value = 0.9265 +Query 1/1: Action query time = 3.085 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9261 +t=106: Selected seed 195 with value = 0.9261 +Query 1/1: Action query time = 2.956 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9248 +t=122: Selected seed 195 with value = 0.9248 +Query 1/1: Action query time = 3.003 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8658 +t=138: Selected seed 195 with value = 0.8658 +Query 1/1: Action query time = 3.033 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9361 +t=154: Selected seed 195 with value = 0.9361 +Query 1/1: Action query time = 3.018 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8697 +t=170: Selected seed 195 with value = 0.8697 +Query 1/1: Action query time = 2.990 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9225 +t=186: Selected seed 195 with value = 0.9225 +Query 1/1: Action query time = 2.309 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8705 +t=202: Selected seed 195 with value = 0.8705 +Query 1/1: Action query time = 2.336 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9203 +t=218: Selected seed 195 with value = 0.9203 +Query 1/1: Action query time = 2.344 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8761 +t=234: Selected seed 195 with value = 0.8761 +Query 1/1: Action query time = 2.290 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8579 +t=250: Selected seed 195 with value = 0.8579 +Query 1/1: Action query time = 2.329 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8840 +t=266: Selected seed 195 with value = 0.8840 +Query 1/1: Action query time = 2.520 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8701 +t=282: Selected seed 195 with value = 0.8701 +Query 1/1: Action query time = 2.589 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8797 +t=298: Selected seed 195 with value = 0.8797 +Saved rollout MP4 at path ./rollouts/demochan200_t2_s02/2026_08_01-13_46_15--with_future_img--episode=3--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_46_16--demochan200_t2_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_46_16--demochan200_t2_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..03000c83065b974e548d2637292959cf7ac472e0 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_46_16--demochan200_t2_s08.txt @@ -0,0 +1,213 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t2_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.201 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5330 +t=10: Selected seed 195 with value = 0.5330 +Query 1/1: Action query time = 5.074 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5828 +t=26: Selected seed 195 with value = 0.5828 +Query 1/1: Action query time = 4.956 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8407 +t=42: Selected seed 195 with value = 0.8407 +Query 1/1: Action query time = 5.360 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7532 +t=58: Selected seed 195 with value = 0.7532 +Query 1/1: Action query time = 3.942 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8896 +t=74: Selected seed 195 with value = 0.8896 +Query 1/1: Action query time = 4.010 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9752 +t=90: Selected seed 195 with value = 0.9752 +Query 1/1: Action query time = 5.055 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.066 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.785 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.832 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.923 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.950 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.396 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.268 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.003 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.394 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.078 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9889 +t=266: Selected seed 195 with value = 0.9889 +Query 1/1: Action query time = 4.236 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8870 +t=282: Selected seed 195 with value = 0.8870 +Query 1/1: Action query time = 3.728 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9229 +t=298: Selected seed 195 with value = 0.9229 +Saved rollout MP4 at path ./rollouts/demochan200_t2_s08/2026_08_01-13_46_16--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 6.032 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4549 +t=10: Selected seed 195 with value = 0.4549 +Query 1/1: Action query time = 4.630 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5974 +t=26: Selected seed 195 with value = 0.5974 +Query 1/1: Action query time = 3.669 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6907 +t=42: Selected seed 195 with value = 0.6907 +Query 1/1: Action query time = 5.104 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7759 +t=58: Selected seed 195 with value = 0.7759 +Query 1/1: Action query time = 5.497 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8863 +t=74: Selected seed 195 with value = 0.8863 +Query 1/1: Action query time = 4.147 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9869 +t=90: Selected seed 195 with value = 0.9869 +Query 1/1: Action query time = 5.018 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t2_s08/2026_08_01-13_46_16--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.436 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5073 +t=10: Selected seed 195 with value = 0.5073 +Query 1/1: Action query time = 4.144 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5617 +t=26: Selected seed 195 with value = 0.5617 +Query 1/1: Action query time = 4.377 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7368 +t=42: Selected seed 195 with value = 0.7368 +Query 1/1: Action query time = 4.487 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7793 +t=58: Selected seed 195 with value = 0.7793 +Query 1/1: Action query time = 3.883 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8019 +t=74: Selected seed 195 with value = 0.8019 +Query 1/1: Action query time = 4.041 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=90: Selected seed 195 with value = 0.9989 +Query 1/1: Action query time = 3.665 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.733 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.471 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.748 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.636 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.521 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.353 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.186 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.013 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.604 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.639 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.217 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.490 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t2_s08/2026_08_01-13_46_16--with_future_img--episode=3--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_46_17--demochan200_t2_s15.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_46_17--demochan200_t2_s15.txt new file mode 100644 index 0000000000000000000000000000000000000000..bdc5e24ff4f9c4cd83357a35b169ba0ebe3fa72b --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_46_17--demochan200_t2_s15.txt @@ -0,0 +1,213 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t2_s15', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='15,31,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.224 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4931 +t=10: Selected seed 195 with value = 0.4931 +Query 1/1: Action query time = 4.765 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5656 +t=26: Selected seed 195 with value = 0.5656 +Query 1/1: Action query time = 5.222 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6751 +t=42: Selected seed 195 with value = 0.6751 +Query 1/1: Action query time = 4.167 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7629 +t=58: Selected seed 195 with value = 0.7629 +Query 1/1: Action query time = 3.693 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8986 +t=74: Selected seed 195 with value = 0.8986 +Query 1/1: Action query time = 5.877 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9884 +t=90: Selected seed 195 with value = 0.9884 +Query 1/1: Action query time = 5.151 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t2_s15/2026_08_01-13_46_17--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.688 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5106 +t=10: Selected seed 195 with value = 0.5106 +Query 1/1: Action query time = 4.786 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5802 +t=26: Selected seed 195 with value = 0.5802 +Query 1/1: Action query time = 4.962 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8238 +t=42: Selected seed 195 with value = 0.8238 +Query 1/1: Action query time = 5.022 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7471 +t=58: Selected seed 195 with value = 0.7471 +Query 1/1: Action query time = 4.862 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8471 +t=74: Selected seed 195 with value = 0.8471 +Query 1/1: Action query time = 4.718 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9892 +t=90: Selected seed 195 with value = 0.9892 +Query 1/1: Action query time = 4.654 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.484 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.068 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.899 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.481 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.462 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.809 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.113 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.562 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.892 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.339 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9794 +t=266: Selected seed 195 with value = 0.9794 +Query 1/1: Action query time = 5.928 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9223 +t=282: Selected seed 195 with value = 0.9223 +Query 1/1: Action query time = 5.763 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8226 +t=298: Selected seed 195 with value = 0.8226 +Saved rollout MP4 at path ./rollouts/demochan200_t2_s15/2026_08_01-13_46_17--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.844 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5222 +t=10: Selected seed 195 with value = 0.5222 +Query 1/1: Action query time = 5.088 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5858 +t=26: Selected seed 195 with value = 0.5858 +Query 1/1: Action query time = 4.352 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8273 +t=42: Selected seed 195 with value = 0.8273 +Query 1/1: Action query time = 4.565 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7741 +t=58: Selected seed 195 with value = 0.7741 +Query 1/1: Action query time = 4.144 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7700 +t=74: Selected seed 195 with value = 0.7700 +Query 1/1: Action query time = 4.111 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8300 +t=90: Selected seed 195 with value = 0.8300 +Query 1/1: Action query time = 4.013 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8378 +t=106: Selected seed 195 with value = 0.8378 +Query 1/1: Action query time = 4.062 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8039 +t=122: Selected seed 195 with value = 0.8039 +Query 1/1: Action query time = 4.269 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9324 +t=138: Selected seed 195 with value = 0.9324 +Query 1/1: Action query time = 3.023 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8468 +t=154: Selected seed 195 with value = 0.8468 +Query 1/1: Action query time = 4.068 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9306 +t=170: Selected seed 195 with value = 0.9306 +Query 1/1: Action query time = 3.551 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8846 +t=186: Selected seed 195 with value = 0.8846 +Query 1/1: Action query time = 2.735 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8917 +t=202: Selected seed 195 with value = 0.8917 +Query 1/1: Action query time = 2.025 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8169 +t=218: Selected seed 195 with value = 0.8169 +Query 1/1: Action query time = 3.720 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8705 +t=234: Selected seed 195 with value = 0.8705 +Query 1/1: Action query time = 4.659 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8335 +t=250: Selected seed 195 with value = 0.8335 +Query 1/1: Action query time = 3.887 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8324 +t=266: Selected seed 195 with value = 0.8324 +Query 1/1: Action query time = 3.332 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8389 +t=282: Selected seed 195 with value = 0.8389 +Query 1/1: Action query time = 2.206 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8517 +t=298: Selected seed 195 with value = 0.8517 +Saved rollout MP4 at path ./rollouts/demochan200_t2_s15/2026_08_01-13_46_17--with_future_img--episode=3--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_40--demochan200_t4_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_40--demochan200_t4_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..cf530c1fbd2aa88efc6811473004698e6d064c58 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_40--demochan200_t4_s03.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t4_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.015 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4861 +t=10: Selected seed 195 with value = 0.4861 +Query 1/1: Action query time = 4.854 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5884 +t=26: Selected seed 195 with value = 0.5884 +Query 1/1: Action query time = 4.914 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6492 +t=42: Selected seed 195 with value = 0.6492 +Query 1/1: Action query time = 4.491 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8103 +t=58: Selected seed 195 with value = 0.8103 +Query 1/1: Action query time = 5.184 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9557 +t=74: Selected seed 195 with value = 0.9557 +Query 1/1: Action query time = 3.834 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=90: Selected seed 195 with value = 0.9983 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s03/2026_08_01-13_51_40--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.789 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4963 +t=10: Selected seed 195 with value = 0.4963 +Query 1/1: Action query time = 4.545 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5857 +t=26: Selected seed 195 with value = 0.5857 +Query 1/1: Action query time = 4.899 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6455 +t=42: Selected seed 195 with value = 0.6455 +Query 1/1: Action query time = 5.016 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7629 +t=58: Selected seed 195 with value = 0.7629 +Query 1/1: Action query time = 5.262 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9201 +t=74: Selected seed 195 with value = 0.9201 +Query 1/1: Action query time = 4.265 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s03/2026_08_01-13_51_40--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.443 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5129 +t=10: Selected seed 195 with value = 0.5129 +Query 1/1: Action query time = 5.010 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5816 +t=26: Selected seed 195 with value = 0.5816 +Query 1/1: Action query time = 4.852 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6620 +t=42: Selected seed 195 with value = 0.6620 +Query 1/1: Action query time = 4.100 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7810 +t=58: Selected seed 195 with value = 0.7810 +Query 1/1: Action query time = 3.958 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9227 +t=74: Selected seed 195 with value = 0.9227 +Query 1/1: Action query time = 2.923 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s03/2026_08_01-13_51_40--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_40--demochan200_t4_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_40--demochan200_t4_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..49c0dc74b7d975d4f439ca9e8ec71cb77c2e60a5 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_40--demochan200_t4_s07.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t4_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.853 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5036 +t=10: Selected seed 195 with value = 0.5036 +Query 1/1: Action query time = 5.029 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5859 +t=26: Selected seed 195 with value = 0.5859 +Query 1/1: Action query time = 5.422 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6521 +t=42: Selected seed 195 with value = 0.6521 +Query 1/1: Action query time = 5.255 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7668 +t=58: Selected seed 195 with value = 0.7668 +Query 1/1: Action query time = 4.933 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9298 +t=74: Selected seed 195 with value = 0.9298 +Query 1/1: Action query time = 4.585 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9975 +t=90: Selected seed 195 with value = 0.9975 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s07/2026_08_01-13_51_40--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.335 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5030 +t=10: Selected seed 195 with value = 0.5030 +Query 1/1: Action query time = 5.237 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5906 +t=26: Selected seed 195 with value = 0.5906 +Query 1/1: Action query time = 5.698 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7030 +t=42: Selected seed 195 with value = 0.7030 +Query 1/1: Action query time = 5.669 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8324 +t=58: Selected seed 195 with value = 0.8324 +Query 1/1: Action query time = 5.154 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9655 +t=74: Selected seed 195 with value = 0.9655 +Query 1/1: Action query time = 3.838 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=90: Selected seed 195 with value = 0.9985 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s07/2026_08_01-13_51_40--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.013 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5018 +t=10: Selected seed 195 with value = 0.5018 +Query 1/1: Action query time = 5.354 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5863 +t=26: Selected seed 195 with value = 0.5863 +Query 1/1: Action query time = 5.335 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6415 +t=42: Selected seed 195 with value = 0.6415 +Query 1/1: Action query time = 4.370 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7884 +t=58: Selected seed 195 with value = 0.7884 +Query 1/1: Action query time = 5.099 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9330 +t=74: Selected seed 195 with value = 0.9330 +Query 1/1: Action query time = 2.808 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9860 +t=90: Selected seed 195 with value = 0.9860 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s07/2026_08_01-13_51_40--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_41--demochan200_t4_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_41--demochan200_t4_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..f79ad93de3a06439de12bbe9d00fed5cad18a65f --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_41--demochan200_t4_s06.txt @@ -0,0 +1,213 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t4_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.514 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5114 +t=10: Selected seed 195 with value = 0.5114 +Query 1/1: Action query time = 4.678 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5953 +t=26: Selected seed 195 with value = 0.5953 +Query 1/1: Action query time = 4.844 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6872 +t=42: Selected seed 195 with value = 0.6872 +Query 1/1: Action query time = 5.029 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8147 +t=58: Selected seed 195 with value = 0.8147 +Query 1/1: Action query time = 5.019 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9333 +t=74: Selected seed 195 with value = 0.9333 +Query 1/1: Action query time = 4.502 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=90: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 3.306 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=106: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 4.045 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.957 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9915 +t=138: Selected seed 195 with value = 0.9915 +Query 1/1: Action query time = 5.634 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9805 +t=154: Selected seed 195 with value = 0.9805 +Query 1/1: Action query time = 5.599 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9793 +t=170: Selected seed 195 with value = 0.9793 +Query 1/1: Action query time = 4.708 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9676 +t=186: Selected seed 195 with value = 0.9676 +Query 1/1: Action query time = 3.870 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=202: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 3.520 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=218: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 3.295 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=234: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 5.720 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=250: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 5.254 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9943 +t=266: Selected seed 195 with value = 0.9943 +Query 1/1: Action query time = 4.464 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9921 +t=282: Selected seed 195 with value = 0.9921 +Query 1/1: Action query time = 4.564 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9905 +t=298: Selected seed 195 with value = 0.9905 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s06/2026_08_01-13_51_41--with_future_img--episode=1--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 1.859 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5041 +t=10: Selected seed 195 with value = 0.5041 +Query 1/1: Action query time = 2.094 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6072 +t=26: Selected seed 195 with value = 0.6072 +Query 1/1: Action query time = 2.511 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6968 +t=42: Selected seed 195 with value = 0.6968 +Query 1/1: Action query time = 2.192 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8485 +t=58: Selected seed 195 with value = 0.8485 +Query 1/1: Action query time = 1.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8463 +t=74: Selected seed 195 with value = 0.8463 +Query 1/1: Action query time = 2.167 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9760 +t=90: Selected seed 195 with value = 0.9760 +Query 1/1: Action query time = 2.124 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9753 +t=106: Selected seed 195 with value = 0.9753 +Query 1/1: Action query time = 2.527 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9790 +t=122: Selected seed 195 with value = 0.9790 +Query 1/1: Action query time = 2.632 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9810 +t=138: Selected seed 195 with value = 0.9810 +Query 1/1: Action query time = 2.494 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9810 +t=154: Selected seed 195 with value = 0.9810 +Query 1/1: Action query time = 1.963 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9783 +t=170: Selected seed 195 with value = 0.9783 +Query 1/1: Action query time = 2.014 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=186: Selected seed 195 with value = 0.9809 +Query 1/1: Action query time = 2.353 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9799 +t=202: Selected seed 195 with value = 0.9799 +Query 1/1: Action query time = 1.992 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9753 +t=218: Selected seed 195 with value = 0.9753 +Query 1/1: Action query time = 1.872 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9726 +t=234: Selected seed 195 with value = 0.9726 +Query 1/1: Action query time = 1.892 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9689 +t=250: Selected seed 195 with value = 0.9689 +Query 1/1: Action query time = 1.841 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9699 +t=266: Selected seed 195 with value = 0.9699 +Query 1/1: Action query time = 1.663 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9730 +t=282: Selected seed 195 with value = 0.9730 +Query 1/1: Action query time = 1.714 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9714 +t=298: Selected seed 195 with value = 0.9714 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s06/2026_08_01-13_51_41--with_future_img--episode=2--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 0.981 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5043 +t=10: Selected seed 195 with value = 0.5043 +Query 1/1: Action query time = 0.961 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5391 +t=26: Selected seed 195 with value = 0.5391 +Query 1/1: Action query time = 0.967 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6384 +t=42: Selected seed 195 with value = 0.6384 +Query 1/1: Action query time = 0.967 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6302 +t=58: Selected seed 195 with value = 0.6302 +Query 1/1: Action query time = 0.997 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7488 +t=74: Selected seed 195 with value = 0.7488 +Query 1/1: Action query time = 0.972 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9276 +t=90: Selected seed 195 with value = 0.9276 +Query 1/1: Action query time = 0.973 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9837 +t=106: Selected seed 195 with value = 0.9837 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s06/2026_08_01-13_51_41--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_42--demochan200_t4_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_42--demochan200_t4_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..431bddebeb385fd3ddeb4b2d564a24a46eadd25a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_42--demochan200_t4_s14.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t4_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.221 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4959 +t=10: Selected seed 195 with value = 0.4959 +Query 1/1: Action query time = 5.377 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5822 +t=26: Selected seed 195 with value = 0.5822 +Query 1/1: Action query time = 4.893 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6426 +t=42: Selected seed 195 with value = 0.6426 +Query 1/1: Action query time = 4.921 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7837 +t=58: Selected seed 195 with value = 0.7837 +Query 1/1: Action query time = 4.989 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9407 +t=74: Selected seed 195 with value = 0.9407 +Query 1/1: Action query time = 2.772 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=90: Selected seed 195 with value = 0.9928 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s14/2026_08_01-13_51_42--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.622 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4926 +t=10: Selected seed 195 with value = 0.4926 +Query 1/1: Action query time = 4.407 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5923 +t=26: Selected seed 195 with value = 0.5923 +Query 1/1: Action query time = 4.851 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6581 +t=42: Selected seed 195 with value = 0.6581 +Query 1/1: Action query time = 4.862 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7726 +t=58: Selected seed 195 with value = 0.7726 +Query 1/1: Action query time = 4.743 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9329 +t=74: Selected seed 195 with value = 0.9329 +Query 1/1: Action query time = 3.647 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=90: Selected seed 195 with value = 0.9912 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s14/2026_08_01-13_51_42--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.209 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5001 +t=10: Selected seed 195 with value = 0.5001 +Query 1/1: Action query time = 5.152 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5852 +t=26: Selected seed 195 with value = 0.5852 +Query 1/1: Action query time = 5.119 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6733 +t=42: Selected seed 195 with value = 0.6733 +Query 1/1: Action query time = 4.872 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7970 +t=58: Selected seed 195 with value = 0.7970 +Query 1/1: Action query time = 4.287 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9425 +t=74: Selected seed 195 with value = 0.9425 +Query 1/1: Action query time = 2.596 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=90: Selected seed 195 with value = 0.9980 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s14/2026_08_01-13_51_42--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_42--demochan200_t4_s15.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_42--demochan200_t4_s15.txt new file mode 100644 index 0000000000000000000000000000000000000000..7fb341ec38d7c1a4b7c376e0947435ddebfc75ff --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_51_42--demochan200_t4_s15.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t4_s15', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='15,31,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.327 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5142 +t=10: Selected seed 195 with value = 0.5142 +Query 1/1: Action query time = 5.098 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6007 +t=26: Selected seed 195 with value = 0.6007 +Query 1/1: Action query time = 5.422 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6705 +t=42: Selected seed 195 with value = 0.6705 +Query 1/1: Action query time = 5.076 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7751 +t=58: Selected seed 195 with value = 0.7751 +Query 1/1: Action query time = 4.953 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9417 +t=74: Selected seed 195 with value = 0.9417 +Query 1/1: Action query time = 5.051 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=90: Selected seed 195 with value = 0.9924 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s15/2026_08_01-13_51_42--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.415 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4968 +t=10: Selected seed 195 with value = 0.4968 +Query 1/1: Action query time = 5.597 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5972 +t=26: Selected seed 195 with value = 0.5972 +Query 1/1: Action query time = 4.942 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6705 +t=42: Selected seed 195 with value = 0.6705 +Query 1/1: Action query time = 4.588 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8094 +t=58: Selected seed 195 with value = 0.8094 +Query 1/1: Action query time = 5.050 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9489 +t=74: Selected seed 195 with value = 0.9489 +Query 1/1: Action query time = 4.909 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9753 +t=90: Selected seed 195 with value = 0.9753 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s15/2026_08_01-13_51_42--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 2.035 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5164 +t=10: Selected seed 195 with value = 0.5164 +Query 1/1: Action query time = 5.096 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6017 +t=26: Selected seed 195 with value = 0.6017 +Query 1/1: Action query time = 4.857 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6642 +t=42: Selected seed 195 with value = 0.6642 +Query 1/1: Action query time = 5.006 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7749 +t=58: Selected seed 195 with value = 0.7749 +Query 1/1: Action query time = 5.005 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9329 +t=74: Selected seed 195 with value = 0.9329 +Query 1/1: Action query time = 4.362 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan200_t4_s15/2026_08_01-13_51_42--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_55_08--demochan200_t5_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_55_08--demochan200_t5_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..5f59a1e2d3a40830eca29d3796c670c4264aeb1b --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_01-13_55_08--demochan200_t5_s08.txt @@ -0,0 +1,141 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000200/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan200_t5_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 3.171 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3697 +t=10: Selected seed 195 with value = 0.3697 +Query 1/1: Action query time = 5.548 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3766 +t=26: Selected seed 195 with value = 0.3766 +Query 1/1: Action query time = 5.192 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4648 +t=42: Selected seed 195 with value = 0.4648 +Query 1/1: Action query time = 4.977 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5441 +t=58: Selected seed 195 with value = 0.5441 +Query 1/1: Action query time = 5.010 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6229 +t=74: Selected seed 195 with value = 0.6229 +Query 1/1: Action query time = 5.314 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7052 +t=90: Selected seed 195 with value = 0.7052 +Query 1/1: Action query time = 5.502 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8643 +t=106: Selected seed 195 with value = 0.8643 +Query 1/1: Action query time = 4.787 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9490 +t=122: Selected seed 195 with value = 0.9490 +Query 1/1: Action query time = 5.144 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9412 +t=138: Selected seed 195 with value = 0.9412 +Saved rollout MP4 at path ./rollouts/demochan200_t5_s08/2026_08_01-13_55_08--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 3.665 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3646 +t=10: Selected seed 195 with value = 0.3646 +Query 1/1: Action query time = 5.664 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3947 +t=26: Selected seed 195 with value = 0.3947 +Query 1/1: Action query time = 5.175 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4898 +t=42: Selected seed 195 with value = 0.4898 +Query 1/1: Action query time = 5.117 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5773 +t=58: Selected seed 195 with value = 0.5773 +Query 1/1: Action query time = 4.951 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6215 +t=74: Selected seed 195 with value = 0.6215 +Query 1/1: Action query time = 5.275 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6766 +t=90: Selected seed 195 with value = 0.6766 +Query 1/1: Action query time = 5.371 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7799 +t=106: Selected seed 195 with value = 0.7799 +Query 1/1: Action query time = 5.061 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9765 +t=122: Selected seed 195 with value = 0.9765 +Saved rollout MP4 at path ./rollouts/demochan200_t5_s08/2026_08_01-13_55_08--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 1.730 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3232 +t=10: Selected seed 195 with value = 0.3232 +Query 1/1: Action query time = 2.894 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3808 +t=26: Selected seed 195 with value = 0.3808 +Query 1/1: Action query time = 5.602 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4892 +t=42: Selected seed 195 with value = 0.4892 +Query 1/1: Action query time = 4.935 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5526 +t=58: Selected seed 195 with value = 0.5526 +Query 1/1: Action query time = 4.855 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5906 +t=74: Selected seed 195 with value = 0.5906 +Query 1/1: Action query time = 4.429 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6634 +t=90: Selected seed 195 with value = 0.6634 +Query 1/1: Action query time = 4.964 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7162 +t=106: Selected seed 195 with value = 0.7162 +Query 1/1: Action query time = 5.743 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8037 +t=122: Selected seed 195 with value = 0.8037 +Query 1/1: Action query time = 5.505 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9574 +t=138: Selected seed 195 with value = 0.9574 +Query 1/1: Action query time = 4.507 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9744 +t=154: Selected seed 195 with value = 0.9744 +Saved rollout MP4 at path ./rollouts/demochan200_t5_s08/2026_08_01-13_55_08--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_15_29--demochan400_t0_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_15_29--demochan400_t0_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..4afb28b5a52d164491d9e6d4034b5b9b2a547146 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_15_29--demochan400_t0_s07.txt @@ -0,0 +1,221 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t0_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.838 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3504 +t=10: Selected seed 195 with value = 0.3504 +Query 1/1: Action query time = 4.919 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4116 +t=26: Selected seed 195 with value = 0.4116 +Query 1/1: Action query time = 5.169 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4661 +t=42: Selected seed 195 with value = 0.4661 +Query 1/1: Action query time = 5.342 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5621 +t=58: Selected seed 195 with value = 0.5621 +Query 1/1: Action query time = 4.366 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6765 +t=74: Selected seed 195 with value = 0.6765 +Query 1/1: Action query time = 4.601 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8738 +t=90: Selected seed 195 with value = 0.8738 +Query 1/1: Action query time = 4.691 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9713 +t=106: Selected seed 195 with value = 0.9713 +Query 1/1: Action query time = 4.113 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9785 +t=122: Selected seed 195 with value = 0.9785 +Query 1/1: Action query time = 4.805 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t0_s07/2026_08_02-02_15_29--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.404 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3199 +t=10: Selected seed 195 with value = 0.3199 +Query 1/1: Action query time = 6.270 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4244 +t=26: Selected seed 195 with value = 0.4244 +Query 1/1: Action query time = 4.769 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4758 +t=42: Selected seed 195 with value = 0.4758 +Query 1/1: Action query time = 4.673 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5756 +t=58: Selected seed 195 with value = 0.5756 +Query 1/1: Action query time = 4.242 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6643 +t=74: Selected seed 195 with value = 0.6643 +Query 1/1: Action query time = 4.632 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7996 +t=90: Selected seed 195 with value = 0.7996 +Query 1/1: Action query time = 4.778 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7617 +t=106: Selected seed 195 with value = 0.7617 +Query 1/1: Action query time = 4.619 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7612 +t=122: Selected seed 195 with value = 0.7612 +Query 1/1: Action query time = 5.072 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7547 +t=138: Selected seed 195 with value = 0.7547 +Query 1/1: Action query time = 2.936 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8290 +t=154: Selected seed 195 with value = 0.8290 +Query 1/1: Action query time = 4.449 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8547 +t=170: Selected seed 195 with value = 0.8547 +Query 1/1: Action query time = 4.290 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8459 +t=186: Selected seed 195 with value = 0.8459 +Query 1/1: Action query time = 1.809 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8446 +t=202: Selected seed 195 with value = 0.8446 +Query 1/1: Action query time = 4.132 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8406 +t=218: Selected seed 195 with value = 0.8406 +Query 1/1: Action query time = 5.137 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8305 +t=234: Selected seed 195 with value = 0.8305 +Query 1/1: Action query time = 5.182 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8310 +t=250: Selected seed 195 with value = 0.8310 +Query 1/1: Action query time = 3.342 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8449 +t=266: Selected seed 195 with value = 0.8449 +Query 1/1: Action query time = 4.930 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8471 +t=282: Selected seed 195 with value = 0.8471 +Query 1/1: Action query time = 4.368 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8406 +t=298: Selected seed 195 with value = 0.8406 +Saved rollout MP4 at path ./rollouts/demochan400_t0_s07/2026_08_02-02_15_29--with_future_img--episode=2--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.509 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3275 +t=10: Selected seed 195 with value = 0.3275 +Query 1/1: Action query time = 4.352 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3937 +t=26: Selected seed 195 with value = 0.3937 +Query 1/1: Action query time = 4.236 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4483 +t=42: Selected seed 195 with value = 0.4483 +Query 1/1: Action query time = 4.042 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5452 +t=58: Selected seed 195 with value = 0.5452 +Query 1/1: Action query time = 5.424 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6588 +t=74: Selected seed 195 with value = 0.6588 +Query 1/1: Action query time = 5.059 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8262 +t=90: Selected seed 195 with value = 0.8262 +Query 1/1: Action query time = 4.072 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7277 +t=106: Selected seed 195 with value = 0.7277 +Query 1/1: Action query time = 3.714 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7129 +t=122: Selected seed 195 with value = 0.7129 +Query 1/1: Action query time = 5.350 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7130 +t=138: Selected seed 195 with value = 0.7130 +Query 1/1: Action query time = 4.958 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7047 +t=154: Selected seed 195 with value = 0.7047 +Query 1/1: Action query time = 5.427 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7357 +t=170: Selected seed 195 with value = 0.7357 +Query 1/1: Action query time = 3.733 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7346 +t=186: Selected seed 195 with value = 0.7346 +Query 1/1: Action query time = 3.789 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7484 +t=202: Selected seed 195 with value = 0.7484 +Query 1/1: Action query time = 1.800 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7504 +t=218: Selected seed 195 with value = 0.7504 +Query 1/1: Action query time = 2.920 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7429 +t=234: Selected seed 195 with value = 0.7429 +Query 1/1: Action query time = 3.652 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7442 +t=250: Selected seed 195 with value = 0.7442 +Query 1/1: Action query time = 3.673 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7522 +t=266: Selected seed 195 with value = 0.7522 +Query 1/1: Action query time = 3.975 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7516 +t=282: Selected seed 195 with value = 0.7516 +Query 1/1: Action query time = 3.801 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7501 +t=298: Selected seed 195 with value = 0.7501 +Saved rollout MP4 at path ./rollouts/demochan400_t0_s07/2026_08_02-02_15_29--with_future_img--episode=3--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_15_29--demochan400_t0_s15.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_15_29--demochan400_t0_s15.txt new file mode 100644 index 0000000000000000000000000000000000000000..525193357be63e7e24915518e2d9457e38e365f0 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_15_29--demochan400_t0_s15.txt @@ -0,0 +1,185 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t0_s15', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='15,31,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.698 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3539 +t=10: Selected seed 195 with value = 0.3539 +Query 1/1: Action query time = 5.382 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4113 +t=26: Selected seed 195 with value = 0.4113 +Query 1/1: Action query time = 5.855 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4584 +t=42: Selected seed 195 with value = 0.4584 +Query 1/1: Action query time = 4.260 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5254 +t=58: Selected seed 195 with value = 0.5254 +Query 1/1: Action query time = 5.267 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6477 +t=74: Selected seed 195 with value = 0.6477 +Query 1/1: Action query time = 4.757 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7486 +t=90: Selected seed 195 with value = 0.7486 +Query 1/1: Action query time = 4.227 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8998 +t=106: Selected seed 195 with value = 0.8998 +Query 1/1: Action query time = 5.150 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9803 +t=122: Selected seed 195 with value = 0.9803 +Query 1/1: Action query time = 5.178 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.725 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t0_s15/2026_08_02-02_15_29--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.070 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3458 +t=10: Selected seed 195 with value = 0.3458 +Query 1/1: Action query time = 5.133 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3905 +t=26: Selected seed 195 with value = 0.3905 +Query 1/1: Action query time = 5.925 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4408 +t=42: Selected seed 195 with value = 0.4408 +Query 1/1: Action query time = 5.658 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5361 +t=58: Selected seed 195 with value = 0.5361 +Query 1/1: Action query time = 5.215 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6461 +t=74: Selected seed 195 with value = 0.6461 +Query 1/1: Action query time = 4.914 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8336 +t=90: Selected seed 195 with value = 0.8336 +Query 1/1: Action query time = 4.533 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9625 +t=106: Selected seed 195 with value = 0.9625 +Query 1/1: Action query time = 3.909 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9856 +t=122: Selected seed 195 with value = 0.9856 +Query 1/1: Action query time = 3.485 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t0_s15/2026_08_02-02_15_29--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.787 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3794 +t=10: Selected seed 195 with value = 0.3794 +Query 1/1: Action query time = 4.270 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4233 +t=26: Selected seed 195 with value = 0.4233 +Query 1/1: Action query time = 5.135 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4855 +t=42: Selected seed 195 with value = 0.4855 +Query 1/1: Action query time = 4.509 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5946 +t=58: Selected seed 195 with value = 0.5946 +Query 1/1: Action query time = 4.723 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6931 +t=74: Selected seed 195 with value = 0.6931 +Query 1/1: Action query time = 5.497 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9096 +t=90: Selected seed 195 with value = 0.9096 +Query 1/1: Action query time = 3.688 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9795 +t=106: Selected seed 195 with value = 0.9795 +Query 1/1: Action query time = 3.852 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7185 +t=122: Selected seed 195 with value = 0.7185 +Query 1/1: Action query time = 3.914 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6991 +t=138: Selected seed 195 with value = 0.6991 +Query 1/1: Action query time = 3.934 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7153 +t=154: Selected seed 195 with value = 0.7153 +Query 1/1: Action query time = 4.089 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7201 +t=170: Selected seed 195 with value = 0.7201 +Query 1/1: Action query time = 2.347 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7301 +t=186: Selected seed 195 with value = 0.7301 +Query 1/1: Action query time = 2.317 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7388 +t=202: Selected seed 195 with value = 0.7388 +Query 1/1: Action query time = 3.501 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7459 +t=218: Selected seed 195 with value = 0.7459 +Query 1/1: Action query time = 5.139 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7534 +t=234: Selected seed 195 with value = 0.7534 +Query 1/1: Action query time = 4.911 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7626 +t=250: Selected seed 195 with value = 0.7626 +Query 1/1: Action query time = 4.379 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7751 +t=266: Selected seed 195 with value = 0.7751 +Query 1/1: Action query time = 3.824 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7949 +t=282: Selected seed 195 with value = 0.7949 +Query 1/1: Action query time = 4.889 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7866 +t=298: Selected seed 195 with value = 0.7866 +Saved rollout MP4 at path ./rollouts/demochan400_t0_s15/2026_08_02-02_15_29--with_future_img--episode=3--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_24--demochan400_t1_s04.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_24--demochan400_t1_s04.txt new file mode 100644 index 0000000000000000000000000000000000000000..a1e604f33fa63899cda59df8dee8f6ce1263b37d --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_24--demochan400_t1_s04.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t1_s04', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='4,20,36', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 3.613 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5057 +t=10: Selected seed 195 with value = 0.5057 +Query 1/1: Action query time = 5.708 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5693 +t=26: Selected seed 195 with value = 0.5693 +Query 1/1: Action query time = 5.172 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7129 +t=42: Selected seed 195 with value = 0.7129 +Query 1/1: Action query time = 4.394 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8276 +t=58: Selected seed 195 with value = 0.8276 +Query 1/1: Action query time = 5.529 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9252 +t=74: Selected seed 195 with value = 0.9252 +Query 1/1: Action query time = 4.942 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s04/2026_08_02-02_21_24--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 5.737 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4828 +t=10: Selected seed 195 with value = 0.4828 +Query 1/1: Action query time = 4.378 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5676 +t=26: Selected seed 195 with value = 0.5676 +Query 1/1: Action query time = 5.634 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6740 +t=42: Selected seed 195 with value = 0.6740 +Query 1/1: Action query time = 4.706 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7530 +t=58: Selected seed 195 with value = 0.7530 +Query 1/1: Action query time = 4.616 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8800 +t=74: Selected seed 195 with value = 0.8800 +Query 1/1: Action query time = 4.088 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9929 +t=90: Selected seed 195 with value = 0.9929 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s04/2026_08_02-02_21_24--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 4.436 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4941 +t=10: Selected seed 195 with value = 0.4941 +Query 1/1: Action query time = 4.052 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5872 +t=26: Selected seed 195 with value = 0.5872 +Query 1/1: Action query time = 4.929 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7197 +t=42: Selected seed 195 with value = 0.7197 +Query 1/1: Action query time = 5.300 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7961 +t=58: Selected seed 195 with value = 0.7961 +Query 1/1: Action query time = 4.843 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9013 +t=74: Selected seed 195 with value = 0.9013 +Query 1/1: Action query time = 3.781 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=90: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 3.895 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9950 +t=106: Selected seed 195 with value = 0.9950 +Query 1/1: Action query time = 2.931 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.319 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.571 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.296 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.499 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.350 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.336 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.400 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.327 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.558 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=266: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 1.245 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9625 +t=282: Selected seed 195 with value = 0.9625 +Query 1/1: Action query time = 0.970 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9290 +t=298: Selected seed 195 with value = 0.9290 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s04/2026_08_02-02_21_24--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_25--demochan400_t1_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_25--demochan400_t1_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..4ad0405fb0dc0cdb0a5b6bfed166f35675e1a1ec --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_25--demochan400_t1_s03.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t1_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.024 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5052 +t=10: Selected seed 195 with value = 0.5052 +Query 1/1: Action query time = 4.247 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6077 +t=26: Selected seed 195 with value = 0.6077 +Query 1/1: Action query time = 5.550 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7058 +t=42: Selected seed 195 with value = 0.7058 +Query 1/1: Action query time = 5.745 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7706 +t=58: Selected seed 195 with value = 0.7706 +Query 1/1: Action query time = 4.570 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8513 +t=74: Selected seed 195 with value = 0.8513 +Query 1/1: Action query time = 2.672 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s03/2026_08_02-02_21_25--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 3.580 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4896 +t=10: Selected seed 195 with value = 0.4896 +Query 1/1: Action query time = 5.316 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5606 +t=26: Selected seed 195 with value = 0.5606 +Query 1/1: Action query time = 5.066 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7029 +t=42: Selected seed 195 with value = 0.7029 +Query 1/1: Action query time = 4.929 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7806 +t=58: Selected seed 195 with value = 0.7806 +Query 1/1: Action query time = 4.427 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9099 +t=74: Selected seed 195 with value = 0.9099 +Query 1/1: Action query time = 4.462 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s03/2026_08_02-02_21_25--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.750 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4922 +t=10: Selected seed 195 with value = 0.4922 +Query 1/1: Action query time = 4.789 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5815 +t=26: Selected seed 195 with value = 0.5815 +Query 1/1: Action query time = 4.750 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7068 +t=42: Selected seed 195 with value = 0.7068 +Query 1/1: Action query time = 4.679 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8177 +t=58: Selected seed 195 with value = 0.8177 +Query 1/1: Action query time = 4.688 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9304 +t=74: Selected seed 195 with value = 0.9304 +Query 1/1: Action query time = 3.713 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=90: Selected seed 195 with value = 0.9957 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s03/2026_08_02-02_21_25--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_25--demochan400_t1_s12.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_25--demochan400_t1_s12.txt new file mode 100644 index 0000000000000000000000000000000000000000..57ff31fe08e64d0fbb04c75cea32aa6320d3c51f --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_25--demochan400_t1_s12.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t1_s12', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='12,28,44', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 1.805 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5169 +t=10: Selected seed 195 with value = 0.5169 +Query 1/1: Action query time = 1.836 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5977 +t=26: Selected seed 195 with value = 0.5977 +Query 1/1: Action query time = 5.062 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7039 +t=42: Selected seed 195 with value = 0.7039 +Query 1/1: Action query time = 3.005 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7580 +t=58: Selected seed 195 with value = 0.7580 +Query 1/1: Action query time = 5.077 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9223 +t=74: Selected seed 195 with value = 0.9223 +Query 1/1: Action query time = 5.822 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=90: Selected seed 195 with value = 0.9867 +Query 1/1: Action query time = 4.812 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=106: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 4.351 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.682 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.208 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.105 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.004 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.535 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.097 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.699 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.143 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9995 +t=250: Selected seed 195 with value = 0.9995 +Query 1/1: Action query time = 3.514 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9694 +t=266: Selected seed 195 with value = 0.9694 +Query 1/1: Action query time = 3.185 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8552 +t=282: Selected seed 195 with value = 0.8552 +Query 1/1: Action query time = 4.780 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9315 +t=298: Selected seed 195 with value = 0.9315 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s12/2026_08_02-02_21_25--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 3.973 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4901 +t=10: Selected seed 195 with value = 0.4901 +Query 1/1: Action query time = 5.028 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5756 +t=26: Selected seed 195 with value = 0.5756 +Query 1/1: Action query time = 4.347 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6889 +t=42: Selected seed 195 with value = 0.6889 +Query 1/1: Action query time = 3.011 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7485 +t=58: Selected seed 195 with value = 0.7485 +Query 1/1: Action query time = 3.753 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8786 +t=74: Selected seed 195 with value = 0.8786 +Query 1/1: Action query time = 1.979 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9951 +t=90: Selected seed 195 with value = 0.9951 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s12/2026_08_02-02_21_25--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 2.506 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5188 +t=10: Selected seed 195 with value = 0.5188 +Query 1/1: Action query time = 1.452 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5584 +t=26: Selected seed 195 with value = 0.5584 +Query 1/1: Action query time = 1.457 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6851 +t=42: Selected seed 195 with value = 0.6851 +Query 1/1: Action query time = 1.517 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7041 +t=58: Selected seed 195 with value = 0.7041 +Query 1/1: Action query time = 1.786 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8435 +t=74: Selected seed 195 with value = 0.8435 +Query 1/1: Action query time = 1.642 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=90: Selected seed 195 with value = 0.9992 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s12/2026_08_02-02_21_25--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_26--demochan400_t1_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_26--demochan400_t1_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..2a99ac12677c9605882ffc969493d3e614992e59 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_21_26--demochan400_t1_s14.txt @@ -0,0 +1,113 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t1_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 6.081 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5199 +t=10: Selected seed 195 with value = 0.5199 +Query 1/1: Action query time = 3.253 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5884 +t=26: Selected seed 195 with value = 0.5884 +Query 1/1: Action query time = 5.215 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7226 +t=42: Selected seed 195 with value = 0.7226 +Query 1/1: Action query time = 5.806 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8241 +t=58: Selected seed 195 with value = 0.8241 +Query 1/1: Action query time = 5.194 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9325 +t=74: Selected seed 195 with value = 0.9325 +Query 1/1: Action query time = 4.660 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=90: Selected seed 195 with value = 0.9978 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s14/2026_08_02-02_21_26--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 2.434 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5007 +t=10: Selected seed 195 with value = 0.5007 +Query 1/1: Action query time = 5.373 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5590 +t=26: Selected seed 195 with value = 0.5590 +Query 1/1: Action query time = 5.501 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6857 +t=42: Selected seed 195 with value = 0.6857 +Query 1/1: Action query time = 5.356 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7756 +t=58: Selected seed 195 with value = 0.7756 +Query 1/1: Action query time = 4.432 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8800 +t=74: Selected seed 195 with value = 0.8800 +Query 1/1: Action query time = 5.236 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9944 +t=90: Selected seed 195 with value = 0.9944 +Query 1/1: Action query time = 3.786 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=106: Selected seed 195 with value = 0.9960 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s14/2026_08_02-02_21_26--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 4.567 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5022 +t=10: Selected seed 195 with value = 0.5022 +Query 1/1: Action query time = 4.338 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5679 +t=26: Selected seed 195 with value = 0.5679 +Query 1/1: Action query time = 4.815 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6376 +t=42: Selected seed 195 with value = 0.6376 +Query 1/1: Action query time = 5.329 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7574 +t=58: Selected seed 195 with value = 0.7574 +Query 1/1: Action query time = 5.421 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8440 +t=74: Selected seed 195 with value = 0.8440 +Query 1/1: Action query time = 3.204 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9920 +t=90: Selected seed 195 with value = 0.9920 +Query 1/1: Action query time = 3.680 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9799 +t=106: Selected seed 195 with value = 0.9799 +Saved rollout MP4 at path ./rollouts/demochan400_t1_s14/2026_08_02-02_21_26--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_30--demochan400_t2_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_30--demochan400_t2_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..14423f802cf3c39a6d19b9a569d1635818bc72bd --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_30--demochan400_t2_s01.txt @@ -0,0 +1,344 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t2_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.801 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4765 +t=10: Selected seed 195 with value = 0.4765 +Query 1/1: Action query time = 4.373 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5438 +t=26: Selected seed 195 with value = 0.5438 +Query 1/1: Action query time = 5.421 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6950 +t=42: Selected seed 195 with value = 0.6950 +Query 1/1: Action query time = 5.709 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7409 +t=58: Selected seed 195 with value = 0.7409 +Query 1/1: Action query time = 3.954 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9003 +t=74: Selected seed 195 with value = 0.9003 +Query 1/1: Action query time = 3.841 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.723 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.801 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.133 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.166 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.118 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.832 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.136 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.122 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.120 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.073 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.045 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.152 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.884 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s01/2026_08_02-02_24_30--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.501 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5119 +t=10: Selected seed 195 with value = 0.5119 +Query 1/1: Action query time = 5.294 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5715 +t=26: Selected seed 195 with value = 0.5715 +Query 1/1: Action query time = 3.740 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7064 +t=42: Selected seed 195 with value = 0.7064 +Query 1/1: Action query time = 4.893 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7697 +t=58: Selected seed 195 with value = 0.7697 +Query 1/1: Action query time = 4.841 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8090 +t=74: Selected seed 195 with value = 0.8090 +Query 1/1: Action query time = 4.235 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9674 +t=90: Selected seed 195 with value = 0.9674 +Query 1/1: Action query time = 5.058 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.136 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.181 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.060 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.615 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.919 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.151 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.708 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.665 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.025 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.815 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.821 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.126 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s01/2026_08_02-02_24_30--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.131 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4966 +t=10: Selected seed 195 with value = 0.4966 +Query 1/1: Action query time = 4.077 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5630 +t=26: Selected seed 195 with value = 0.5630 +Query 1/1: Action query time = 3.663 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6946 +t=42: Selected seed 195 with value = 0.6946 +Query 1/1: Action query time = 5.225 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7233 +t=58: Selected seed 195 with value = 0.7233 +Query 1/1: Action query time = 4.457 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8851 +t=74: Selected seed 195 with value = 0.8851 +Query 1/1: Action query time = 4.707 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=90: Selected seed 195 with value = 0.9962 +Query 1/1: Action query time = 4.680 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.481 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.399 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.332 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.371 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.448 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.427 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.412 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.493 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.526 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.530 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.559 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.827 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s01/2026_08_02-02_24_30--with_future_img--episode=3--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 0.974 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5121 +t=10: Selected seed 195 with value = 0.5121 +Query 1/1: Action query time = 0.969 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5581 +t=26: Selected seed 195 with value = 0.5581 +Query 1/1: Action query time = 0.953 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6827 +t=42: Selected seed 195 with value = 0.6827 +Query 1/1: Action query time = 0.978 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7551 +t=58: Selected seed 195 with value = 0.7551 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8748 +t=74: Selected seed 195 with value = 0.8748 +Query 1/1: Action query time = 0.957 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9863 +t=90: Selected seed 195 with value = 0.9863 +Query 1/1: Action query time = 0.962 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.953 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.953 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.949 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.946 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.957 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.963 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.958 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.980 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9874 +t=282: Selected seed 195 with value = 0.9874 +Query 1/1: Action query time = 0.961 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9495 +t=298: Selected seed 195 with value = 0.9495 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s01/2026_08_02-02_24_30--with_future_img--episode=4--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 4 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_30--demochan400_t2_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_30--demochan400_t2_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..049bbf1c2226c74b691713ec9203c39bb78c5329 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_30--demochan400_t2_s06.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t2_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.641 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4860 +t=10: Selected seed 195 with value = 0.4860 +Query 1/1: Action query time = 4.541 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5750 +t=26: Selected seed 195 with value = 0.5750 +Query 1/1: Action query time = 4.878 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7137 +t=42: Selected seed 195 with value = 0.7137 +Query 1/1: Action query time = 4.875 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7174 +t=58: Selected seed 195 with value = 0.7174 +Query 1/1: Action query time = 5.462 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8861 +t=74: Selected seed 195 with value = 0.8861 +Query 1/1: Action query time = 4.494 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=90: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 4.424 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.152 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.763 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.043 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.079 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.188 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.026 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.678 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.237 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.485 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.101 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.855 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.923 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s06/2026_08_02-02_24_30--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.644 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4740 +t=10: Selected seed 195 with value = 0.4740 +Query 1/1: Action query time = 4.895 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5413 +t=26: Selected seed 195 with value = 0.5413 +Query 1/1: Action query time = 5.942 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8721 +t=42: Selected seed 195 with value = 0.8721 +Query 1/1: Action query time = 5.067 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7365 +t=58: Selected seed 195 with value = 0.7365 +Query 1/1: Action query time = 4.072 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8180 +t=74: Selected seed 195 with value = 0.8180 +Query 1/1: Action query time = 4.805 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9577 +t=90: Selected seed 195 with value = 0.9577 +Query 1/1: Action query time = 4.668 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9843 +t=106: Selected seed 195 with value = 0.9843 +Query 1/1: Action query time = 5.119 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.170 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.274 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.288 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.284 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.042 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.703 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.096 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.551 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.275 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.256 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.223 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s06/2026_08_02-02_24_30--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 2.864 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5097 +t=10: Selected seed 195 with value = 0.5097 +Query 1/1: Action query time = 4.946 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5848 +t=26: Selected seed 195 with value = 0.5848 +Query 1/1: Action query time = 4.577 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7282 +t=42: Selected seed 195 with value = 0.7282 +Query 1/1: Action query time = 4.702 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7441 +t=58: Selected seed 195 with value = 0.7441 +Query 1/1: Action query time = 4.779 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8295 +t=74: Selected seed 195 with value = 0.8295 +Query 1/1: Action query time = 4.848 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9795 +t=90: Selected seed 195 with value = 0.9795 +Query 1/1: Action query time = 4.620 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.499 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.467 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.489 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.546 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.558 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.552 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.523 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.546 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.519 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.518 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.558 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.786 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s06/2026_08_02-02_24_30--with_future_img--episode=3--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_30--demochan400_t2_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_30--demochan400_t2_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..f756b31294953280b92c60bf740f51dfd4f002e0 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_30--demochan400_t2_s07.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t2_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 7.097 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4984 +t=10: Selected seed 195 with value = 0.4984 +Query 1/1: Action query time = 4.910 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5676 +t=26: Selected seed 195 with value = 0.5676 +Query 1/1: Action query time = 4.678 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7325 +t=42: Selected seed 195 with value = 0.7325 +Query 1/1: Action query time = 5.002 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7795 +t=58: Selected seed 195 with value = 0.7795 +Query 1/1: Action query time = 5.320 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7864 +t=74: Selected seed 195 with value = 0.7864 +Query 1/1: Action query time = 4.086 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9096 +t=90: Selected seed 195 with value = 0.9096 +Query 1/1: Action query time = 5.216 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=106: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 5.398 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.628 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.603 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.975 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.097 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.632 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.936 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.073 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.385 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.144 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.851 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9757 +t=282: Selected seed 195 with value = 0.9757 +Query 1/1: Action query time = 3.269 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9600 +t=298: Selected seed 195 with value = 0.9600 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s07/2026_08_02-02_24_30--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.397 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5200 +t=10: Selected seed 195 with value = 0.5200 +Query 1/1: Action query time = 4.960 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5884 +t=26: Selected seed 195 with value = 0.5884 +Query 1/1: Action query time = 5.133 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7263 +t=42: Selected seed 195 with value = 0.7263 +Query 1/1: Action query time = 3.499 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7916 +t=58: Selected seed 195 with value = 0.7916 +Query 1/1: Action query time = 3.950 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9397 +t=74: Selected seed 195 with value = 0.9397 +Query 1/1: Action query time = 5.285 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=90: Selected seed 195 with value = 0.9982 +Query 1/1: Action query time = 5.387 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.924 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.027 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.118 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.330 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.119 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.912 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.147 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.126 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.111 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.703 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9910 +t=266: Selected seed 195 with value = 0.9910 +Query 1/1: Action query time = 4.388 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9755 +t=282: Selected seed 195 with value = 0.9755 +Query 1/1: Action query time = 2.942 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9451 +t=298: Selected seed 195 with value = 0.9451 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s07/2026_08_02-02_24_30--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.558 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4659 +t=10: Selected seed 195 with value = 0.4659 +Query 1/1: Action query time = 4.083 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5590 +t=26: Selected seed 195 with value = 0.5590 +Query 1/1: Action query time = 5.385 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6970 +t=42: Selected seed 195 with value = 0.6970 +Query 1/1: Action query time = 4.516 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7884 +t=58: Selected seed 195 with value = 0.7884 +Query 1/1: Action query time = 4.766 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7796 +t=74: Selected seed 195 with value = 0.7796 +Query 1/1: Action query time = 4.858 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9069 +t=90: Selected seed 195 with value = 0.9069 +Query 1/1: Action query time = 4.764 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=106: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 4.702 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.621 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.587 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.617 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.647 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.586 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.543 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.568 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.534 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.200 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9866 +t=266: Selected seed 195 with value = 0.9866 +Query 1/1: Action query time = 3.568 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9582 +t=282: Selected seed 195 with value = 0.9582 +Query 1/1: Action query time = 1.989 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9380 +t=298: Selected seed 195 with value = 0.9380 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s07/2026_08_02-02_24_30--with_future_img--episode=3--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_32--demochan400_t2_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_32--demochan400_t2_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..64f0480bb64fcd9d03bb170322a7241865debdb8 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_24_32--demochan400_t2_s14.txt @@ -0,0 +1,209 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t2_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.193 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5056 +t=10: Selected seed 195 with value = 0.5056 +Query 1/1: Action query time = 4.791 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5709 +t=26: Selected seed 195 with value = 0.5709 +Query 1/1: Action query time = 3.410 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7185 +t=42: Selected seed 195 with value = 0.7185 +Query 1/1: Action query time = 4.506 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7557 +t=58: Selected seed 195 with value = 0.7557 +Query 1/1: Action query time = 5.273 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9129 +t=74: Selected seed 195 with value = 0.9129 +Query 1/1: Action query time = 5.055 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=90: Selected seed 195 with value = 0.9946 +Query 1/1: Action query time = 4.947 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.129 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.622 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.622 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.878 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.994 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.774 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.211 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.614 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.310 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.059 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.533 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.515 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s14/2026_08_02-02_24_32--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.019 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5198 +t=10: Selected seed 195 with value = 0.5198 +Query 1/1: Action query time = 4.762 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5831 +t=26: Selected seed 195 with value = 0.5831 +Query 1/1: Action query time = 5.631 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6993 +t=42: Selected seed 195 with value = 0.6993 +Query 1/1: Action query time = 4.748 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7487 +t=58: Selected seed 195 with value = 0.7487 +Query 1/1: Action query time = 4.671 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8933 +t=74: Selected seed 195 with value = 0.8933 +Query 1/1: Action query time = 3.826 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=90: Selected seed 195 with value = 0.9983 +Query 1/1: Action query time = 4.696 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.230 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.288 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.496 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.084 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.682 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.418 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.054 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.826 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.795 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.126 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.195 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.104 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9953 +t=298: Selected seed 195 with value = 0.9953 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s14/2026_08_02-02_24_32--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 1.159 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5009 +t=10: Selected seed 195 with value = 0.5009 +Query 1/1: Action query time = 2.325 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5882 +t=26: Selected seed 195 with value = 0.5882 +Query 1/1: Action query time = 2.365 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6636 +t=42: Selected seed 195 with value = 0.6636 +Query 1/1: Action query time = 1.680 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7963 +t=58: Selected seed 195 with value = 0.7963 +Query 1/1: Action query time = 2.153 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9304 +t=74: Selected seed 195 with value = 0.9304 +Query 1/1: Action query time = 2.081 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9947 +t=90: Selected seed 195 with value = 0.9947 +Saved rollout MP4 at path ./rollouts/demochan400_t2_s14/2026_08_02-02_24_32--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_31_01--demochan400_t3_s13.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_31_01--demochan400_t3_s13.txt new file mode 100644 index 0000000000000000000000000000000000000000..78382791550c13a9d5fd520da8541d45a27a2941 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_31_01--demochan400_t3_s13.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t3_s13', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='13,29,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 5.798 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3864 +t=10: Selected seed 195 with value = 0.3864 +Query 1/1: Action query time = 5.066 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5106 +t=26: Selected seed 195 with value = 0.5106 +Query 1/1: Action query time = 4.628 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6289 +t=42: Selected seed 195 with value = 0.6289 +Query 1/1: Action query time = 5.330 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6627 +t=58: Selected seed 195 with value = 0.6627 +Query 1/1: Action query time = 5.194 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7377 +t=74: Selected seed 195 with value = 0.7377 +Query 1/1: Action query time = 5.110 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7824 +t=90: Selected seed 195 with value = 0.7824 +Query 1/1: Action query time = 5.524 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8828 +t=106: Selected seed 195 with value = 0.8828 +Query 1/1: Action query time = 4.961 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8006 +t=122: Selected seed 195 with value = 0.8006 +Query 1/1: Action query time = 4.398 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8319 +t=138: Selected seed 195 with value = 0.8319 +Query 1/1: Action query time = 5.316 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8847 +t=154: Selected seed 195 with value = 0.8847 +Query 1/1: Action query time = 5.169 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8876 +t=170: Selected seed 195 with value = 0.8876 +Query 1/1: Action query time = 4.578 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6683 +t=186: Selected seed 195 with value = 0.6683 +Query 1/1: Action query time = 5.144 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5640 +t=202: Selected seed 195 with value = 0.5640 +Query 1/1: Action query time = 5.257 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6234 +t=218: Selected seed 195 with value = 0.6234 +Query 1/1: Action query time = 5.123 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7625 +t=234: Selected seed 195 with value = 0.7625 +Query 1/1: Action query time = 5.698 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8449 +t=250: Selected seed 195 with value = 0.8449 +Query 1/1: Action query time = 5.443 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8422 +t=266: Selected seed 195 with value = 0.8422 +Query 1/1: Action query time = 4.487 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7311 +t=282: Selected seed 195 with value = 0.7311 +Query 1/1: Action query time = 2.339 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8654 +t=298: Selected seed 195 with value = 0.8654 +Saved rollout MP4 at path ./rollouts/demochan400_t3_s13/2026_08_02-02_31_01--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 5.670 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3927 +t=10: Selected seed 195 with value = 0.3927 +Query 1/1: Action query time = 4.595 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5254 +t=26: Selected seed 195 with value = 0.5254 +Query 1/1: Action query time = 4.395 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6575 +t=42: Selected seed 195 with value = 0.6575 +Query 1/1: Action query time = 4.857 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6454 +t=58: Selected seed 195 with value = 0.6454 +Query 1/1: Action query time = 4.833 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6063 +t=74: Selected seed 195 with value = 0.6063 +Query 1/1: Action query time = 5.738 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6077 +t=90: Selected seed 195 with value = 0.6077 +Query 1/1: Action query time = 5.101 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7406 +t=106: Selected seed 195 with value = 0.7406 +Query 1/1: Action query time = 5.322 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7051 +t=122: Selected seed 195 with value = 0.7051 +Query 1/1: Action query time = 4.857 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8849 +t=138: Selected seed 195 with value = 0.8849 +Query 1/1: Action query time = 5.086 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7978 +t=154: Selected seed 195 with value = 0.7978 +Query 1/1: Action query time = 5.493 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8317 +t=170: Selected seed 195 with value = 0.8317 +Query 1/1: Action query time = 5.426 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9415 +t=186: Selected seed 195 with value = 0.9415 +Query 1/1: Action query time = 5.054 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9366 +t=202: Selected seed 195 with value = 0.9366 +Query 1/1: Action query time = 4.796 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9692 +t=218: Selected seed 195 with value = 0.9692 +Query 1/1: Action query time = 4.897 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9631 +t=234: Selected seed 195 with value = 0.9631 +Query 1/1: Action query time = 4.744 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9598 +t=250: Selected seed 195 with value = 0.9598 +Query 1/1: Action query time = 4.846 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9781 +t=266: Selected seed 195 with value = 0.9781 +Query 1/1: Action query time = 4.803 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=282: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 2.709 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t3_s13/2026_08_02-02_31_01--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 5.352 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3745 +t=10: Selected seed 195 with value = 0.3745 +Query 1/1: Action query time = 4.740 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4862 +t=26: Selected seed 195 with value = 0.4862 +Query 1/1: Action query time = 4.762 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6626 +t=42: Selected seed 195 with value = 0.6626 +Query 1/1: Action query time = 5.184 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7165 +t=58: Selected seed 195 with value = 0.7165 +Query 1/1: Action query time = 4.454 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7803 +t=74: Selected seed 195 with value = 0.7803 +Query 1/1: Action query time = 4.768 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6881 +t=90: Selected seed 195 with value = 0.6881 +Query 1/1: Action query time = 4.257 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7662 +t=106: Selected seed 195 with value = 0.7662 +Query 1/1: Action query time = 4.916 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8732 +t=122: Selected seed 195 with value = 0.8732 +Query 1/1: Action query time = 5.214 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8984 +t=138: Selected seed 195 with value = 0.8984 +Query 1/1: Action query time = 5.008 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9245 +t=154: Selected seed 195 with value = 0.9245 +Query 1/1: Action query time = 4.594 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8073 +t=170: Selected seed 195 with value = 0.8073 +Query 1/1: Action query time = 5.260 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9669 +t=186: Selected seed 195 with value = 0.9669 +Query 1/1: Action query time = 5.401 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9555 +t=202: Selected seed 195 with value = 0.9555 +Query 1/1: Action query time = 4.594 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9349 +t=218: Selected seed 195 with value = 0.9349 +Query 1/1: Action query time = 4.252 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9611 +t=234: Selected seed 195 with value = 0.9611 +Query 1/1: Action query time = 4.166 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9629 +t=250: Selected seed 195 with value = 0.9629 +Query 1/1: Action query time = 3.698 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9754 +t=266: Selected seed 195 with value = 0.9754 +Query 1/1: Action query time = 3.332 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9872 +t=282: Selected seed 195 with value = 0.9872 +Query 1/1: Action query time = 1.746 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9889 +t=298: Selected seed 195 with value = 0.9889 +Saved rollout MP4 at path ./rollouts/demochan400_t3_s13/2026_08_02-02_31_01--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_31_01--demochan400_t3_s15.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_31_01--demochan400_t3_s15.txt new file mode 100644 index 0000000000000000000000000000000000000000..dacad08d544bf3a184f00b247c4fb904bcb9c7d7 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_31_01--demochan400_t3_s15.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t3_s15', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='15,31,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 5.128 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4050 +t=10: Selected seed 195 with value = 0.4050 +Query 1/1: Action query time = 4.467 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4518 +t=26: Selected seed 195 with value = 0.4518 +Query 1/1: Action query time = 5.105 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6372 +t=42: Selected seed 195 with value = 0.6372 +Query 1/1: Action query time = 5.342 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6148 +t=58: Selected seed 195 with value = 0.6148 +Query 1/1: Action query time = 4.515 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5904 +t=74: Selected seed 195 with value = 0.5904 +Query 1/1: Action query time = 4.710 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7878 +t=90: Selected seed 195 with value = 0.7878 +Query 1/1: Action query time = 4.467 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9215 +t=106: Selected seed 195 with value = 0.9215 +Query 1/1: Action query time = 4.870 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9314 +t=122: Selected seed 195 with value = 0.9314 +Query 1/1: Action query time = 5.288 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7751 +t=138: Selected seed 195 with value = 0.7751 +Query 1/1: Action query time = 4.577 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9503 +t=154: Selected seed 195 with value = 0.9503 +Query 1/1: Action query time = 4.542 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8760 +t=170: Selected seed 195 with value = 0.8760 +Query 1/1: Action query time = 5.518 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9420 +t=186: Selected seed 195 with value = 0.9420 +Query 1/1: Action query time = 5.182 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9454 +t=202: Selected seed 195 with value = 0.9454 +Query 1/1: Action query time = 4.975 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9499 +t=218: Selected seed 195 with value = 0.9499 +Query 1/1: Action query time = 4.939 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9652 +t=234: Selected seed 195 with value = 0.9652 +Query 1/1: Action query time = 5.160 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9786 +t=250: Selected seed 195 with value = 0.9786 +Query 1/1: Action query time = 4.594 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9802 +t=266: Selected seed 195 with value = 0.9802 +Query 1/1: Action query time = 4.904 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9848 +t=282: Selected seed 195 with value = 0.9848 +Query 1/1: Action query time = 4.507 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t3_s15/2026_08_02-02_31_01--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.951 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3975 +t=10: Selected seed 195 with value = 0.3975 +Query 1/1: Action query time = 5.355 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5193 +t=26: Selected seed 195 with value = 0.5193 +Query 1/1: Action query time = 4.795 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6476 +t=42: Selected seed 195 with value = 0.6476 +Query 1/1: Action query time = 4.663 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6186 +t=58: Selected seed 195 with value = 0.6186 +Query 1/1: Action query time = 5.265 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5764 +t=74: Selected seed 195 with value = 0.5764 +Query 1/1: Action query time = 4.333 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7709 +t=90: Selected seed 195 with value = 0.7709 +Query 1/1: Action query time = 4.789 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8645 +t=106: Selected seed 195 with value = 0.8645 +Query 1/1: Action query time = 4.745 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8434 +t=122: Selected seed 195 with value = 0.8434 +Query 1/1: Action query time = 5.113 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6925 +t=138: Selected seed 195 with value = 0.6925 +Query 1/1: Action query time = 5.154 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7640 +t=154: Selected seed 195 with value = 0.7640 +Query 1/1: Action query time = 5.065 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7795 +t=170: Selected seed 195 with value = 0.7795 +Query 1/1: Action query time = 4.651 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9021 +t=186: Selected seed 195 with value = 0.9021 +Query 1/1: Action query time = 4.989 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9726 +t=202: Selected seed 195 with value = 0.9726 +Query 1/1: Action query time = 4.871 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9746 +t=218: Selected seed 195 with value = 0.9746 +Query 1/1: Action query time = 4.988 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9729 +t=234: Selected seed 195 with value = 0.9729 +Query 1/1: Action query time = 4.961 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9888 +t=250: Selected seed 195 with value = 0.9888 +Query 1/1: Action query time = 3.503 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9907 +t=266: Selected seed 195 with value = 0.9907 +Query 1/1: Action query time = 3.955 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9894 +t=282: Selected seed 195 with value = 0.9894 +Query 1/1: Action query time = 4.607 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9845 +t=298: Selected seed 195 with value = 0.9845 +Saved rollout MP4 at path ./rollouts/demochan400_t3_s15/2026_08_02-02_31_01--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 3.289 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3879 +t=10: Selected seed 195 with value = 0.3879 +Query 1/1: Action query time = 5.293 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4840 +t=26: Selected seed 195 with value = 0.4840 +Query 1/1: Action query time = 4.771 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6036 +t=42: Selected seed 195 with value = 0.6036 +Query 1/1: Action query time = 5.111 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5954 +t=58: Selected seed 195 with value = 0.5954 +Query 1/1: Action query time = 5.027 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6436 +t=74: Selected seed 195 with value = 0.6436 +Query 1/1: Action query time = 4.743 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7661 +t=90: Selected seed 195 with value = 0.7661 +Query 1/1: Action query time = 4.783 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7769 +t=106: Selected seed 195 with value = 0.7769 +Query 1/1: Action query time = 4.886 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6706 +t=122: Selected seed 195 with value = 0.6706 +Query 1/1: Action query time = 4.553 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7695 +t=138: Selected seed 195 with value = 0.7695 +Query 1/1: Action query time = 4.902 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9707 +t=154: Selected seed 195 with value = 0.9707 +Query 1/1: Action query time = 5.429 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9758 +t=170: Selected seed 195 with value = 0.9758 +Query 1/1: Action query time = 4.844 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9796 +t=186: Selected seed 195 with value = 0.9796 +Query 1/1: Action query time = 4.891 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9807 +t=202: Selected seed 195 with value = 0.9807 +Query 1/1: Action query time = 4.809 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9777 +t=218: Selected seed 195 with value = 0.9777 +Query 1/1: Action query time = 5.327 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9692 +t=234: Selected seed 195 with value = 0.9692 +Query 1/1: Action query time = 3.182 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9624 +t=250: Selected seed 195 with value = 0.9624 +Query 1/1: Action query time = 4.068 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9539 +t=266: Selected seed 195 with value = 0.9539 +Query 1/1: Action query time = 3.602 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7368 +t=282: Selected seed 195 with value = 0.7368 +Query 1/1: Action query time = 3.339 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7515 +t=298: Selected seed 195 with value = 0.7515 +Saved rollout MP4 at path ./rollouts/demochan400_t3_s15/2026_08_02-02_31_01--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_37_50--demochan400_t4_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_37_50--demochan400_t4_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..88f02b5e5dbea3623c6ff96c085aefc9da9210e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_37_50--demochan400_t4_s01.txt @@ -0,0 +1,136 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t4_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.934 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5199 +t=10: Selected seed 195 with value = 0.5199 +Query 1/1: Action query time = 3.738 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5614 +t=26: Selected seed 195 with value = 0.5614 +Query 1/1: Action query time = 4.653 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6546 +t=42: Selected seed 195 with value = 0.6546 +Query 1/1: Action query time = 4.922 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7824 +t=58: Selected seed 195 with value = 0.7824 +Query 1/1: Action query time = 4.676 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9287 +t=74: Selected seed 195 with value = 0.9287 +Query 1/1: Action query time = 4.448 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9856 +t=90: Selected seed 195 with value = 0.9856 +Saved rollout MP4 at path ./rollouts/demochan400_t4_s01/2026_08_02-02_37_50--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.264 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4901 +t=10: Selected seed 195 with value = 0.4901 +Query 1/1: Action query time = 4.278 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5851 +t=26: Selected seed 195 with value = 0.5851 +Query 1/1: Action query time = 4.960 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6777 +t=42: Selected seed 195 with value = 0.6777 +Query 1/1: Action query time = 4.557 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7843 +t=58: Selected seed 195 with value = 0.7843 +Query 1/1: Action query time = 5.046 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9493 +t=74: Selected seed 195 with value = 0.9493 +Query 1/1: Action query time = 3.672 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=90: Selected seed 195 with value = 0.9954 +Saved rollout MP4 at path ./rollouts/demochan400_t4_s01/2026_08_02-02_37_50--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.187 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5146 +t=10: Selected seed 195 with value = 0.5146 +Query 1/1: Action query time = 4.788 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5857 +t=26: Selected seed 195 with value = 0.5857 +Query 1/1: Action query time = 4.491 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6761 +t=42: Selected seed 195 with value = 0.6761 +Query 1/1: Action query time = 5.459 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8120 +t=58: Selected seed 195 with value = 0.8120 +Query 1/1: Action query time = 4.359 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9567 +t=74: Selected seed 195 with value = 0.9567 +Query 1/1: Action query time = 4.132 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=90: Selected seed 195 with value = 0.9963 +Saved rollout MP4 at path ./rollouts/demochan400_t4_s01/2026_08_02-02_37_50--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 3.645 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5170 +t=10: Selected seed 195 with value = 0.5170 +Query 1/1: Action query time = 2.819 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5760 +t=26: Selected seed 195 with value = 0.5760 +Query 1/1: Action query time = 1.789 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6613 +t=42: Selected seed 195 with value = 0.6613 +Query 1/1: Action query time = 1.871 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7738 +t=58: Selected seed 195 with value = 0.7738 +Query 1/1: Action query time = 1.605 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9173 +t=74: Selected seed 195 with value = 0.9173 +Query 1/1: Action query time = 1.316 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9975 +t=90: Selected seed 195 with value = 0.9975 +Saved rollout MP4 at path ./rollouts/demochan400_t4_s01/2026_08_02-02_37_50--with_future_img--episode=4--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 4 +Total successes: 4 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_37_51--demochan400_t4_s10.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_37_51--demochan400_t4_s10.txt new file mode 100644 index 0000000000000000000000000000000000000000..25b78dfe381fcb3ddcde5cf4a7ed40b7c6b17a2f --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_37_51--demochan400_t4_s10.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t4_s10', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='10,26,42', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.811 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4863 +t=10: Selected seed 195 with value = 0.4863 +Query 1/1: Action query time = 4.142 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5734 +t=26: Selected seed 195 with value = 0.5734 +Query 1/1: Action query time = 4.353 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6620 +t=42: Selected seed 195 with value = 0.6620 +Query 1/1: Action query time = 5.131 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7679 +t=58: Selected seed 195 with value = 0.7679 +Query 1/1: Action query time = 5.415 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9139 +t=74: Selected seed 195 with value = 0.9139 +Query 1/1: Action query time = 4.328 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t4_s10/2026_08_02-02_37_51--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.941 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4822 +t=10: Selected seed 195 with value = 0.4822 +Query 1/1: Action query time = 2.944 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5625 +t=26: Selected seed 195 with value = 0.5625 +Query 1/1: Action query time = 4.424 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6551 +t=42: Selected seed 195 with value = 0.6551 +Query 1/1: Action query time = 4.354 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7701 +t=58: Selected seed 195 with value = 0.7701 +Query 1/1: Action query time = 5.549 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9324 +t=74: Selected seed 195 with value = 0.9324 +Query 1/1: Action query time = 5.068 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=90: Selected seed 195 with value = 0.9993 +Saved rollout MP4 at path ./rollouts/demochan400_t4_s10/2026_08_02-02_37_51--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.471 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4913 +t=10: Selected seed 195 with value = 0.4913 +Query 1/1: Action query time = 2.572 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5718 +t=26: Selected seed 195 with value = 0.5718 +Query 1/1: Action query time = 2.567 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6719 +t=42: Selected seed 195 with value = 0.6719 +Query 1/1: Action query time = 5.677 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7697 +t=58: Selected seed 195 with value = 0.7697 +Query 1/1: Action query time = 4.862 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9162 +t=74: Selected seed 195 with value = 0.9162 +Query 1/1: Action query time = 4.180 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=90: Selected seed 195 with value = 0.9809 +Saved rollout MP4 at path ./rollouts/demochan400_t4_s10/2026_08_02-02_37_51--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_40_38--demochan400_t5_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_40_38--demochan400_t5_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..48a2cec224b7dd1ae62519d733beed425f986355 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_40_38--demochan400_t5_s02.txt @@ -0,0 +1,181 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t5_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 1.854 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3334 +t=10: Selected seed 195 with value = 0.3334 +Query 1/1: Action query time = 2.367 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3808 +t=26: Selected seed 195 with value = 0.3808 +Query 1/1: Action query time = 4.397 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4371 +t=42: Selected seed 195 with value = 0.4371 +Query 1/1: Action query time = 4.193 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5167 +t=58: Selected seed 195 with value = 0.5167 +Query 1/1: Action query time = 4.890 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5970 +t=74: Selected seed 195 with value = 0.5970 +Query 1/1: Action query time = 5.220 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6540 +t=90: Selected seed 195 with value = 0.6540 +Query 1/1: Action query time = 4.972 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7668 +t=106: Selected seed 195 with value = 0.7668 +Query 1/1: Action query time = 4.697 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9176 +t=122: Selected seed 195 with value = 0.9176 +Query 1/1: Action query time = 5.172 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t5_s02/2026_08_02-02_40_38--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.911 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3357 +t=10: Selected seed 195 with value = 0.3357 +Query 1/1: Action query time = 3.834 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3994 +t=26: Selected seed 195 with value = 0.3994 +Query 1/1: Action query time = 4.089 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4830 +t=42: Selected seed 195 with value = 0.4830 +Query 1/1: Action query time = 5.402 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5368 +t=58: Selected seed 195 with value = 0.5368 +Query 1/1: Action query time = 5.016 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6030 +t=74: Selected seed 195 with value = 0.6030 +Query 1/1: Action query time = 4.886 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6961 +t=90: Selected seed 195 with value = 0.6961 +Query 1/1: Action query time = 5.005 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8412 +t=106: Selected seed 195 with value = 0.8412 +Query 1/1: Action query time = 5.595 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9391 +t=122: Selected seed 195 with value = 0.9391 +Query 1/1: Action query time = 4.971 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/demochan400_t5_s02/2026_08_02-02_40_38--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 2.816 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3317 +t=10: Selected seed 195 with value = 0.3317 +Query 1/1: Action query time = 2.837 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3105 +t=26: Selected seed 195 with value = 0.3105 +Query 1/1: Action query time = 3.115 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4936 +t=42: Selected seed 195 with value = 0.4936 +Query 1/1: Action query time = 4.456 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4653 +t=58: Selected seed 195 with value = 0.4653 +Query 1/1: Action query time = 5.551 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6113 +t=74: Selected seed 195 with value = 0.6113 +Query 1/1: Action query time = 5.839 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7214 +t=90: Selected seed 195 with value = 0.7214 +Query 1/1: Action query time = 5.588 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7930 +t=106: Selected seed 195 with value = 0.7930 +Query 1/1: Action query time = 5.108 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8544 +t=122: Selected seed 195 with value = 0.8544 +Query 1/1: Action query time = 4.348 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9363 +t=138: Selected seed 195 with value = 0.9363 +Query 1/1: Action query time = 5.045 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9332 +t=154: Selected seed 195 with value = 0.9332 +Query 1/1: Action query time = 4.918 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9246 +t=170: Selected seed 195 with value = 0.9246 +Query 1/1: Action query time = 1.880 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8407 +t=186: Selected seed 195 with value = 0.8407 +Query 1/1: Action query time = 1.697 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8405 +t=202: Selected seed 195 with value = 0.8405 +Query 1/1: Action query time = 2.289 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9918 +t=218: Selected seed 195 with value = 0.9918 +Query 1/1: Action query time = 1.349 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=234: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 3.446 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9948 +t=250: Selected seed 195 with value = 0.9948 +Query 1/1: Action query time = 3.237 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9887 +t=266: Selected seed 195 with value = 0.9887 +Query 1/1: Action query time = 3.293 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9780 +t=282: Selected seed 195 with value = 0.9780 +Query 1/1: Action query time = 3.048 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9735 +t=298: Selected seed 195 with value = 0.9735 +Saved rollout MP4 at path ./rollouts/demochan400_t5_s02/2026_08_02-02_40_38--with_future_img--episode=3--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s05.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s05.txt new file mode 100644 index 0000000000000000000000000000000000000000..4155b67ee7375fa6abb935575851f859bbacfed5 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s05.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t6_s05', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,21,37', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 4.133 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4509 +t=10: Selected seed 195 with value = 0.4509 +Query 1/1: Action query time = 5.303 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4874 +t=26: Selected seed 195 with value = 0.4874 +Query 1/1: Action query time = 5.410 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5728 +t=42: Selected seed 195 with value = 0.5728 +Query 1/1: Action query time = 5.302 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6094 +t=58: Selected seed 195 with value = 0.6094 +Query 1/1: Action query time = 5.045 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5927 +t=74: Selected seed 195 with value = 0.5927 +Query 1/1: Action query time = 3.998 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6966 +t=90: Selected seed 195 with value = 0.6966 +Query 1/1: Action query time = 4.367 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7582 +t=106: Selected seed 195 with value = 0.7582 +Query 1/1: Action query time = 4.569 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6670 +t=122: Selected seed 195 with value = 0.6670 +Query 1/1: Action query time = 3.707 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8281 +t=138: Selected seed 195 with value = 0.8281 +Query 1/1: Action query time = 5.214 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6746 +t=154: Selected seed 195 with value = 0.6746 +Query 1/1: Action query time = 5.442 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8407 +t=170: Selected seed 195 with value = 0.8407 +Query 1/1: Action query time = 5.781 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6809 +t=186: Selected seed 195 with value = 0.6809 +Query 1/1: Action query time = 3.569 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8014 +t=202: Selected seed 195 with value = 0.8014 +Query 1/1: Action query time = 4.443 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8310 +t=218: Selected seed 195 with value = 0.8310 +Query 1/1: Action query time = 3.302 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7746 +t=234: Selected seed 195 with value = 0.7746 +Query 1/1: Action query time = 4.530 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8707 +t=250: Selected seed 195 with value = 0.8707 +Query 1/1: Action query time = 4.342 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9359 +t=266: Selected seed 195 with value = 0.9359 +Query 1/1: Action query time = 4.583 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9591 +t=282: Selected seed 195 with value = 0.9591 +Query 1/1: Action query time = 4.569 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9863 +t=298: Selected seed 195 with value = 0.9863 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s05/2026_08_02-02_45_26--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 3.725 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4523 +t=10: Selected seed 195 with value = 0.4523 +Query 1/1: Action query time = 3.176 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4961 +t=26: Selected seed 195 with value = 0.4961 +Query 1/1: Action query time = 2.839 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6281 +t=42: Selected seed 195 with value = 0.6281 +Query 1/1: Action query time = 2.581 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7304 +t=58: Selected seed 195 with value = 0.7304 +Query 1/1: Action query time = 1.783 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8819 +t=74: Selected seed 195 with value = 0.8819 +Query 1/1: Action query time = 2.025 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=90: Selected seed 195 with value = 0.9977 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s05/2026_08_02-02_45_26--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 1.482 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4314 +t=10: Selected seed 195 with value = 0.4314 +Query 1/1: Action query time = 1.503 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5076 +t=26: Selected seed 195 with value = 0.5076 +Query 1/1: Action query time = 1.906 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5871 +t=42: Selected seed 195 with value = 0.5871 +Query 1/1: Action query time = 1.994 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6931 +t=58: Selected seed 195 with value = 0.6931 +Query 1/1: Action query time = 1.169 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7923 +t=74: Selected seed 195 with value = 0.7923 +Query 1/1: Action query time = 1.191 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9166 +t=90: Selected seed 195 with value = 0.9166 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s05/2026_08_02-02_45_26--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..29700bdcb082f0d85001562c0934396dedb9a84a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s06.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t6_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 6.800 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4836 +t=10: Selected seed 195 with value = 0.4836 +Query 1/1: Action query time = 4.892 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5254 +t=26: Selected seed 195 with value = 0.5254 +Query 1/1: Action query time = 4.831 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6031 +t=42: Selected seed 195 with value = 0.6031 +Query 1/1: Action query time = 4.000 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7222 +t=58: Selected seed 195 with value = 0.7222 +Query 1/1: Action query time = 4.772 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8410 +t=74: Selected seed 195 with value = 0.8410 +Query 1/1: Action query time = 2.628 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9915 +t=90: Selected seed 195 with value = 0.9915 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s06/2026_08_02-02_45_26--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 5.923 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4826 +t=10: Selected seed 195 with value = 0.4826 +Query 1/1: Action query time = 6.147 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4728 +t=26: Selected seed 195 with value = 0.4728 +Query 1/1: Action query time = 5.184 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5770 +t=42: Selected seed 195 with value = 0.5770 +Query 1/1: Action query time = 3.338 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6987 +t=58: Selected seed 195 with value = 0.6987 +Query 1/1: Action query time = 4.292 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8331 +t=74: Selected seed 195 with value = 0.8331 +Query 1/1: Action query time = 5.133 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9518 +t=90: Selected seed 195 with value = 0.9518 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s06/2026_08_02-02_45_26--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.835 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4293 +t=10: Selected seed 195 with value = 0.4293 +Query 1/1: Action query time = 4.679 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5397 +t=26: Selected seed 195 with value = 0.5397 +Query 1/1: Action query time = 4.483 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6486 +t=42: Selected seed 195 with value = 0.6486 +Query 1/1: Action query time = 4.943 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7190 +t=58: Selected seed 195 with value = 0.7190 +Query 1/1: Action query time = 4.724 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8545 +t=74: Selected seed 195 with value = 0.8545 +Query 1/1: Action query time = 2.625 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9766 +t=90: Selected seed 195 with value = 0.9766 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s06/2026_08_02-02_45_26--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..665f1f28bb037c4c3f8ac1a38842b3084065a75e --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s08.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t6_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 4.183 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4658 +t=10: Selected seed 195 with value = 0.4658 +Query 1/1: Action query time = 4.510 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5464 +t=26: Selected seed 195 with value = 0.5464 +Query 1/1: Action query time = 5.121 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6697 +t=42: Selected seed 195 with value = 0.6697 +Query 1/1: Action query time = 5.255 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7392 +t=58: Selected seed 195 with value = 0.7392 +Query 1/1: Action query time = 4.812 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8624 +t=74: Selected seed 195 with value = 0.8624 +Query 1/1: Action query time = 4.529 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9883 +t=90: Selected seed 195 with value = 0.9883 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s08/2026_08_02-02_45_26--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 5.719 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4363 +t=10: Selected seed 195 with value = 0.4363 +Query 1/1: Action query time = 4.767 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5568 +t=26: Selected seed 195 with value = 0.5568 +Query 1/1: Action query time = 4.441 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6432 +t=42: Selected seed 195 with value = 0.6432 +Query 1/1: Action query time = 5.141 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7257 +t=58: Selected seed 195 with value = 0.7257 +Query 1/1: Action query time = 4.496 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8681 +t=74: Selected seed 195 with value = 0.8681 +Query 1/1: Action query time = 4.896 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9830 +t=90: Selected seed 195 with value = 0.9830 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s08/2026_08_02-02_45_26--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 3.777 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4514 +t=10: Selected seed 195 with value = 0.4514 +Query 1/1: Action query time = 5.046 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5183 +t=26: Selected seed 195 with value = 0.5183 +Query 1/1: Action query time = 5.320 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6197 +t=42: Selected seed 195 with value = 0.6197 +Query 1/1: Action query time = 4.739 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6753 +t=58: Selected seed 195 with value = 0.6753 +Query 1/1: Action query time = 3.998 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8586 +t=74: Selected seed 195 with value = 0.8586 +Query 1/1: Action query time = 4.728 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8571 +t=90: Selected seed 195 with value = 0.8571 +Query 1/1: Action query time = 3.027 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9306 +t=106: Selected seed 195 with value = 0.9306 +Query 1/1: Action query time = 1.839 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9863 +t=122: Selected seed 195 with value = 0.9863 +Query 1/1: Action query time = 2.064 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.666 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9786 +t=154: Selected seed 195 with value = 0.9786 +Query 1/1: Action query time = 2.140 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9229 +t=170: Selected seed 195 with value = 0.9229 +Query 1/1: Action query time = 1.920 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9289 +t=186: Selected seed 195 with value = 0.9289 +Query 1/1: Action query time = 2.134 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9699 +t=202: Selected seed 195 with value = 0.9699 +Query 1/1: Action query time = 1.938 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9447 +t=218: Selected seed 195 with value = 0.9447 +Query 1/1: Action query time = 1.796 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9522 +t=234: Selected seed 195 with value = 0.9522 +Query 1/1: Action query time = 1.669 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7193 +t=250: Selected seed 195 with value = 0.7193 +Query 1/1: Action query time = 1.183 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8947 +t=266: Selected seed 195 with value = 0.8947 +Query 1/1: Action query time = 1.164 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8991 +t=282: Selected seed 195 with value = 0.8991 +Query 1/1: Action query time = 0.976 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8111 +t=298: Selected seed 195 with value = 0.8111 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s08/2026_08_02-02_45_26--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s12.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s12.txt new file mode 100644 index 0000000000000000000000000000000000000000..83be92fc2f1ef4639411ea3cafa2495b047682b3 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-02_45_26--demochan400_t6_s12.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_demochan_from40k_2gpu/checkpoints/iter_000000400/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='demochan400_t6_s12', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='12,28,44', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 3.024 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4640 +t=10: Selected seed 195 with value = 0.4640 +Query 1/1: Action query time = 4.201 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5241 +t=26: Selected seed 195 with value = 0.5241 +Query 1/1: Action query time = 4.977 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6343 +t=42: Selected seed 195 with value = 0.6343 +Query 1/1: Action query time = 4.711 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6978 +t=58: Selected seed 195 with value = 0.6978 +Query 1/1: Action query time = 5.151 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8067 +t=74: Selected seed 195 with value = 0.8067 +Query 1/1: Action query time = 4.849 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9645 +t=90: Selected seed 195 with value = 0.9645 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s12/2026_08_02-02_45_26--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 3.276 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4470 +t=10: Selected seed 195 with value = 0.4470 +Query 1/1: Action query time = 4.995 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4777 +t=26: Selected seed 195 with value = 0.4777 +Query 1/1: Action query time = 3.598 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6245 +t=42: Selected seed 195 with value = 0.6245 +Query 1/1: Action query time = 4.635 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7120 +t=58: Selected seed 195 with value = 0.7120 +Query 1/1: Action query time = 5.726 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8325 +t=74: Selected seed 195 with value = 0.8325 +Query 1/1: Action query time = 4.068 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=90: Selected seed 195 with value = 0.9864 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s12/2026_08_02-02_45_26--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.787 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4363 +t=10: Selected seed 195 with value = 0.4363 +Query 1/1: Action query time = 3.955 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5268 +t=26: Selected seed 195 with value = 0.5268 +Query 1/1: Action query time = 5.358 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6468 +t=42: Selected seed 195 with value = 0.6468 +Query 1/1: Action query time = 5.410 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6745 +t=58: Selected seed 195 with value = 0.6745 +Query 1/1: Action query time = 4.788 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6891 +t=74: Selected seed 195 with value = 0.6891 +Query 1/1: Action query time = 3.775 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8105 +t=90: Selected seed 195 with value = 0.8105 +Query 1/1: Action query time = 4.452 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8463 +t=106: Selected seed 195 with value = 0.8463 +Query 1/1: Action query time = 3.261 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9080 +t=122: Selected seed 195 with value = 0.9080 +Query 1/1: Action query time = 2.114 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9875 +t=138: Selected seed 195 with value = 0.9875 +Query 1/1: Action query time = 1.843 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=154: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 1.749 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=170: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 1.997 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=186: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 1.747 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9810 +t=202: Selected seed 195 with value = 0.9810 +Query 1/1: Action query time = 1.905 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8941 +t=218: Selected seed 195 with value = 0.8941 +Query 1/1: Action query time = 1.757 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9995 +t=234: Selected seed 195 with value = 0.9995 +Query 1/1: Action query time = 1.927 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=250: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 1.657 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8605 +t=266: Selected seed 195 with value = 0.8605 +Query 1/1: Action query time = 1.768 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8658 +t=282: Selected seed 195 with value = 0.8658 +Query 1/1: Action query time = 1.282 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9744 +t=298: Selected seed 195 with value = 0.9744 +Saved rollout MP4 at path ./rollouts/demochan400_t6_s12/2026_08_02-02_45_26--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_34_47--realcl800_t0_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_34_47--realcl800_t0_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..889f364dfc581a0f21553dd53b29d7b233637e4c --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_34_47--realcl800_t0_s03.txt @@ -0,0 +1,133 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t0_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 1.926 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2274 +t=10: Selected seed 195 with value = 0.2274 +Query 1/1: Action query time = 1.980 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2383 +t=26: Selected seed 195 with value = 0.2383 +Query 1/1: Action query time = 4.063 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4149 +t=42: Selected seed 195 with value = 0.4149 +Query 1/1: Action query time = 4.758 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4378 +t=58: Selected seed 195 with value = 0.4378 +Query 1/1: Action query time = 4.857 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5245 +t=74: Selected seed 195 with value = 0.5245 +Query 1/1: Action query time = 5.038 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6535 +t=90: Selected seed 195 with value = 0.6535 +Query 1/1: Action query time = 4.588 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6884 +t=106: Selected seed 195 with value = 0.6884 +Query 1/1: Action query time = 4.783 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8518 +t=122: Selected seed 195 with value = 0.8518 +Query 1/1: Action query time = 5.524 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9581 +t=138: Selected seed 195 with value = 0.9581 +Saved rollout MP4 at path ./rollouts/realcl800_t0_s03/2026_08_02-03_34_47--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 1.727 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3690 +t=10: Selected seed 195 with value = 0.3690 +Query 1/1: Action query time = 2.175 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4179 +t=26: Selected seed 195 with value = 0.4179 +Query 1/1: Action query time = 4.919 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4785 +t=42: Selected seed 195 with value = 0.4785 +Query 1/1: Action query time = 4.904 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5528 +t=58: Selected seed 195 with value = 0.5528 +Query 1/1: Action query time = 4.987 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6454 +t=74: Selected seed 195 with value = 0.6454 +Query 1/1: Action query time = 4.899 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7783 +t=90: Selected seed 195 with value = 0.7783 +Query 1/1: Action query time = 4.917 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9055 +t=106: Selected seed 195 with value = 0.9055 +Saved rollout MP4 at path ./rollouts/realcl800_t0_s03/2026_08_02-03_34_47--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.774 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3060 +t=10: Selected seed 195 with value = 0.3060 +Query 1/1: Action query time = 5.055 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3080 +t=26: Selected seed 195 with value = 0.3080 +Query 1/1: Action query time = 3.320 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3528 +t=42: Selected seed 195 with value = 0.3528 +Query 1/1: Action query time = 2.929 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4376 +t=58: Selected seed 195 with value = 0.4376 +Query 1/1: Action query time = 5.056 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4913 +t=74: Selected seed 195 with value = 0.4913 +Query 1/1: Action query time = 4.270 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5763 +t=90: Selected seed 195 with value = 0.5763 +Query 1/1: Action query time = 4.585 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7519 +t=106: Selected seed 195 with value = 0.7519 +Query 1/1: Action query time = 4.928 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8806 +t=122: Selected seed 195 with value = 0.8806 +Query 1/1: Action query time = 5.871 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=138: Selected seed 195 with value = 0.9926 +Saved rollout MP4 at path ./rollouts/realcl800_t0_s03/2026_08_02-03_34_47--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_34_47--realcl800_t0_s05.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_34_47--realcl800_t0_s05.txt new file mode 100644 index 0000000000000000000000000000000000000000..58fe5e0897474ff9383645fd2843eaaad2605f0f --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_34_47--realcl800_t0_s05.txt @@ -0,0 +1,137 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t0_s05', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,21,37', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.550 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3028 +t=10: Selected seed 195 with value = 0.3028 +Query 1/1: Action query time = 2.542 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3449 +t=26: Selected seed 195 with value = 0.3449 +Query 1/1: Action query time = 5.105 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4166 +t=42: Selected seed 195 with value = 0.4166 +Query 1/1: Action query time = 4.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4743 +t=58: Selected seed 195 with value = 0.4743 +Query 1/1: Action query time = 4.953 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5534 +t=74: Selected seed 195 with value = 0.5534 +Query 1/1: Action query time = 5.253 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6001 +t=90: Selected seed 195 with value = 0.6001 +Query 1/1: Action query time = 5.301 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6816 +t=106: Selected seed 195 with value = 0.6816 +Query 1/1: Action query time = 5.464 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8262 +t=122: Selected seed 195 with value = 0.8262 +Query 1/1: Action query time = 3.907 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9813 +t=138: Selected seed 195 with value = 0.9813 +Saved rollout MP4 at path ./rollouts/realcl800_t0_s05/2026_08_02-03_34_47--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.152 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3012 +t=10: Selected seed 195 with value = 0.3012 +Query 1/1: Action query time = 5.724 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2654 +t=26: Selected seed 195 with value = 0.2654 +Query 1/1: Action query time = 5.499 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3305 +t=42: Selected seed 195 with value = 0.3305 +Query 1/1: Action query time = 5.518 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4312 +t=58: Selected seed 195 with value = 0.4312 +Query 1/1: Action query time = 5.610 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5758 +t=74: Selected seed 195 with value = 0.5758 +Query 1/1: Action query time = 4.778 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6449 +t=90: Selected seed 195 with value = 0.6449 +Query 1/1: Action query time = 5.019 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7646 +t=106: Selected seed 195 with value = 0.7646 +Query 1/1: Action query time = 5.237 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8997 +t=122: Selected seed 195 with value = 0.8997 +Query 1/1: Action query time = 3.092 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9942 +t=138: Selected seed 195 with value = 0.9942 +Saved rollout MP4 at path ./rollouts/realcl800_t0_s05/2026_08_02-03_34_47--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.237 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2928 +t=10: Selected seed 195 with value = 0.2928 +Query 1/1: Action query time = 4.233 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3202 +t=26: Selected seed 195 with value = 0.3202 +Query 1/1: Action query time = 4.426 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3908 +t=42: Selected seed 195 with value = 0.3908 +Query 1/1: Action query time = 4.482 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4852 +t=58: Selected seed 195 with value = 0.4852 +Query 1/1: Action query time = 5.407 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5837 +t=74: Selected seed 195 with value = 0.5837 +Query 1/1: Action query time = 4.747 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6626 +t=90: Selected seed 195 with value = 0.6626 +Query 1/1: Action query time = 4.898 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8012 +t=106: Selected seed 195 with value = 0.8012 +Query 1/1: Action query time = 3.167 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9399 +t=122: Selected seed 195 with value = 0.9399 +Saved rollout MP4 at path ./rollouts/realcl800_t0_s05/2026_08_02-03_34_47--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_34_48--realcl800_t0_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_34_48--realcl800_t0_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..1520d04ee0b6614c2a7cef8da92b5072557c3830 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_34_48--realcl800_t0_s06.txt @@ -0,0 +1,145 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t0_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.891 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2267 +t=10: Selected seed 195 with value = 0.2267 +Query 1/1: Action query time = 4.895 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2366 +t=26: Selected seed 195 with value = 0.2366 +Query 1/1: Action query time = 5.132 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2798 +t=42: Selected seed 195 with value = 0.2798 +Query 1/1: Action query time = 4.604 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3129 +t=58: Selected seed 195 with value = 0.3129 +Query 1/1: Action query time = 4.936 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3747 +t=74: Selected seed 195 with value = 0.3747 +Query 1/1: Action query time = 5.313 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4211 +t=90: Selected seed 195 with value = 0.4211 +Query 1/1: Action query time = 5.256 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5235 +t=106: Selected seed 195 with value = 0.5235 +Query 1/1: Action query time = 3.766 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9198 +t=122: Selected seed 195 with value = 0.9198 +Saved rollout MP4 at path ./rollouts/realcl800_t0_s06/2026_08_02-03_34_48--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.084 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3492 +t=10: Selected seed 195 with value = 0.3492 +Query 1/1: Action query time = 4.922 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3926 +t=26: Selected seed 195 with value = 0.3926 +Query 1/1: Action query time = 4.965 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4393 +t=42: Selected seed 195 with value = 0.4393 +Query 1/1: Action query time = 4.678 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5275 +t=58: Selected seed 195 with value = 0.5275 +Query 1/1: Action query time = 4.948 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6204 +t=74: Selected seed 195 with value = 0.6204 +Query 1/1: Action query time = 4.297 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7581 +t=90: Selected seed 195 with value = 0.7581 +Query 1/1: Action query time = 4.875 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8655 +t=106: Selected seed 195 with value = 0.8655 +Query 1/1: Action query time = 5.168 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9925 +t=122: Selected seed 195 with value = 0.9925 +Saved rollout MP4 at path ./rollouts/realcl800_t0_s06/2026_08_02-03_34_48--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.812 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3338 +t=10: Selected seed 195 with value = 0.3338 +Query 1/1: Action query time = 4.095 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3828 +t=26: Selected seed 195 with value = 0.3828 +Query 1/1: Action query time = 5.453 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4394 +t=42: Selected seed 195 with value = 0.4394 +Query 1/1: Action query time = 4.835 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5063 +t=58: Selected seed 195 with value = 0.5063 +Query 1/1: Action query time = 5.682 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6192 +t=74: Selected seed 195 with value = 0.6192 +Query 1/1: Action query time = 4.663 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6926 +t=90: Selected seed 195 with value = 0.6926 +Query 1/1: Action query time = 4.265 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6902 +t=106: Selected seed 195 with value = 0.6902 +Query 1/1: Action query time = 4.158 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8511 +t=122: Selected seed 195 with value = 0.8511 +Query 1/1: Action query time = 4.868 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6944 +t=138: Selected seed 195 with value = 0.6944 +Query 1/1: Action query time = 3.177 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7500 +t=154: Selected seed 195 with value = 0.7500 +Query 1/1: Action query time = 1.816 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8116 +t=170: Selected seed 195 with value = 0.8116 +Query 1/1: Action query time = 2.708 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9640 +t=186: Selected seed 195 with value = 0.9640 +Saved rollout MP4 at path ./rollouts/realcl800_t0_s06/2026_08_02-03_34_48--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_03--realcl800_t1_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_03--realcl800_t1_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..150e6aaf144007c76310292af35f5f9dfb6b12f5 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_03--realcl800_t1_s01.txt @@ -0,0 +1,132 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t1_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 3.791 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4500 +t=10: Selected seed 195 with value = 0.4500 +Query 1/1: Action query time = 5.174 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5027 +t=26: Selected seed 195 with value = 0.5027 +Query 1/1: Action query time = 4.764 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6016 +t=42: Selected seed 195 with value = 0.6016 +Query 1/1: Action query time = 4.940 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6934 +t=58: Selected seed 195 with value = 0.6934 +Query 1/1: Action query time = 5.415 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8286 +t=74: Selected seed 195 with value = 0.8286 +Query 1/1: Action query time = 4.002 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=90: Selected seed 195 with value = 0.9922 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s01/2026_08_02-03_38_03--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 3.436 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4259 +t=10: Selected seed 195 with value = 0.4259 +Query 1/1: Action query time = 4.857 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5111 +t=26: Selected seed 195 with value = 0.5111 +Query 1/1: Action query time = 4.588 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6162 +t=42: Selected seed 195 with value = 0.6162 +Query 1/1: Action query time = 5.407 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7338 +t=58: Selected seed 195 with value = 0.7338 +Query 1/1: Action query time = 5.431 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8761 +t=74: Selected seed 195 with value = 0.8761 +Query 1/1: Action query time = 3.665 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9897 +t=90: Selected seed 195 with value = 0.9897 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s01/2026_08_02-03_38_03--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 2.940 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4143 +t=10: Selected seed 195 with value = 0.4143 +Query 1/1: Action query time = 5.066 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5105 +t=26: Selected seed 195 with value = 0.5105 +Query 1/1: Action query time = 5.257 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6022 +t=42: Selected seed 195 with value = 0.6022 +Query 1/1: Action query time = 5.412 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6903 +t=58: Selected seed 195 with value = 0.6903 +Query 1/1: Action query time = 5.039 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7924 +t=74: Selected seed 195 with value = 0.7924 +Query 1/1: Action query time = 3.384 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9650 +t=90: Selected seed 195 with value = 0.9650 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s01/2026_08_02-03_38_03--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 2.052 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4740 +t=10: Selected seed 195 with value = 0.4740 +Query 1/1: Action query time = 1.149 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5604 +t=26: Selected seed 195 with value = 0.5604 +Query 1/1: Action query time = 1.143 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6203 +t=42: Selected seed 195 with value = 0.6203 +Query 1/1: Action query time = 1.279 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7557 +t=58: Selected seed 195 with value = 0.7557 +Query 1/1: Action query time = 1.161 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9065 +t=74: Selected seed 195 with value = 0.9065 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s01/2026_08_02-03_38_03--with_future_img--episode=4--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 4 +Total successes: 4 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_04--realcl800_t1_s04.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_04--realcl800_t1_s04.txt new file mode 100644 index 0000000000000000000000000000000000000000..bbf14bc2e9312e76d653232467267e8afd28480a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_04--realcl800_t1_s04.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t1_s04', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='4,20,36', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.991 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4337 +t=10: Selected seed 195 with value = 0.4337 +Query 1/1: Action query time = 4.969 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5044 +t=26: Selected seed 195 with value = 0.5044 +Query 1/1: Action query time = 5.199 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5808 +t=42: Selected seed 195 with value = 0.5808 +Query 1/1: Action query time = 5.224 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6794 +t=58: Selected seed 195 with value = 0.6794 +Query 1/1: Action query time = 4.077 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7781 +t=74: Selected seed 195 with value = 0.7781 +Query 1/1: Action query time = 2.844 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9308 +t=90: Selected seed 195 with value = 0.9308 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s04/2026_08_02-03_38_04--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 3.826 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4320 +t=10: Selected seed 195 with value = 0.4320 +Query 1/1: Action query time = 4.761 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5178 +t=26: Selected seed 195 with value = 0.5178 +Query 1/1: Action query time = 5.053 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6167 +t=42: Selected seed 195 with value = 0.6167 +Query 1/1: Action query time = 4.756 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6957 +t=58: Selected seed 195 with value = 0.6957 +Query 1/1: Action query time = 3.702 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8340 +t=74: Selected seed 195 with value = 0.8340 +Query 1/1: Action query time = 4.051 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9869 +t=90: Selected seed 195 with value = 0.9869 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s04/2026_08_02-03_38_04--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.089 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4585 +t=10: Selected seed 195 with value = 0.4585 +Query 1/1: Action query time = 5.616 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5449 +t=26: Selected seed 195 with value = 0.5449 +Query 1/1: Action query time = 5.312 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6374 +t=42: Selected seed 195 with value = 0.6374 +Query 1/1: Action query time = 4.793 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7445 +t=58: Selected seed 195 with value = 0.7445 +Query 1/1: Action query time = 3.516 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8633 +t=74: Selected seed 195 with value = 0.8633 +Query 1/1: Action query time = 2.182 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9913 +t=90: Selected seed 195 with value = 0.9913 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s04/2026_08_02-03_38_04--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_04--realcl800_t1_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_04--realcl800_t1_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..9951fa8318b7f67401794ed29ea860014769c581 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_04--realcl800_t1_s06.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t1_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 3.360 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3639 +t=10: Selected seed 195 with value = 0.3639 +Query 1/1: Action query time = 5.378 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4712 +t=26: Selected seed 195 with value = 0.4712 +Query 1/1: Action query time = 4.760 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5926 +t=42: Selected seed 195 with value = 0.5926 +Query 1/1: Action query time = 4.483 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6914 +t=58: Selected seed 195 with value = 0.6914 +Query 1/1: Action query time = 4.299 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7975 +t=74: Selected seed 195 with value = 0.7975 +Query 1/1: Action query time = 4.232 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9204 +t=90: Selected seed 195 with value = 0.9204 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s06/2026_08_02-03_38_04--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 2.614 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3860 +t=10: Selected seed 195 with value = 0.3860 +Query 1/1: Action query time = 3.743 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5075 +t=26: Selected seed 195 with value = 0.5075 +Query 1/1: Action query time = 5.714 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6062 +t=42: Selected seed 195 with value = 0.6062 +Query 1/1: Action query time = 5.402 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6955 +t=58: Selected seed 195 with value = 0.6955 +Query 1/1: Action query time = 5.047 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8173 +t=74: Selected seed 195 with value = 0.8173 +Query 1/1: Action query time = 3.306 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9618 +t=90: Selected seed 195 with value = 0.9618 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s06/2026_08_02-03_38_04--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 2.747 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4279 +t=10: Selected seed 195 with value = 0.4279 +Query 1/1: Action query time = 4.668 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4841 +t=26: Selected seed 195 with value = 0.4841 +Query 1/1: Action query time = 4.916 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6029 +t=42: Selected seed 195 with value = 0.6029 +Query 1/1: Action query time = 4.469 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6874 +t=58: Selected seed 195 with value = 0.6874 +Query 1/1: Action query time = 5.009 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8209 +t=74: Selected seed 195 with value = 0.8209 +Query 1/1: Action query time = 4.371 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9715 +t=90: Selected seed 195 with value = 0.9715 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s06/2026_08_02-03_38_04--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_04--realcl800_t1_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_04--realcl800_t1_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..f84c97459afd0675b185158399d7f216b5a68e62 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_04--realcl800_t1_s11.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t1_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.080 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4492 +t=10: Selected seed 195 with value = 0.4492 +Query 1/1: Action query time = 4.916 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5201 +t=26: Selected seed 195 with value = 0.5201 +Query 1/1: Action query time = 5.254 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6114 +t=42: Selected seed 195 with value = 0.6114 +Query 1/1: Action query time = 5.290 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7021 +t=58: Selected seed 195 with value = 0.7021 +Query 1/1: Action query time = 5.456 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8123 +t=74: Selected seed 195 with value = 0.8123 +Query 1/1: Action query time = 4.732 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9645 +t=90: Selected seed 195 with value = 0.9645 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s11/2026_08_02-03_38_04--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 4.293 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4029 +t=10: Selected seed 195 with value = 0.4029 +Query 1/1: Action query time = 4.496 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5197 +t=26: Selected seed 195 with value = 0.5197 +Query 1/1: Action query time = 5.100 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6028 +t=42: Selected seed 195 with value = 0.6028 +Query 1/1: Action query time = 5.421 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7059 +t=58: Selected seed 195 with value = 0.7059 +Query 1/1: Action query time = 3.793 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8209 +t=74: Selected seed 195 with value = 0.8209 +Query 1/1: Action query time = 4.811 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9674 +t=90: Selected seed 195 with value = 0.9674 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s11/2026_08_02-03_38_04--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.251 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4581 +t=10: Selected seed 195 with value = 0.4581 +Query 1/1: Action query time = 4.689 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5334 +t=26: Selected seed 195 with value = 0.5334 +Query 1/1: Action query time = 4.875 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6237 +t=42: Selected seed 195 with value = 0.6237 +Query 1/1: Action query time = 4.528 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6895 +t=58: Selected seed 195 with value = 0.6895 +Query 1/1: Action query time = 4.440 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8223 +t=74: Selected seed 195 with value = 0.8223 +Query 1/1: Action query time = 3.929 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9748 +t=90: Selected seed 195 with value = 0.9748 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s11/2026_08_02-03_38_04--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_05--realcl800_t1_s13.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_05--realcl800_t1_s13.txt new file mode 100644 index 0000000000000000000000000000000000000000..4dd01f8393babb3e1d395c98d48b6330b530347d --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_38_05--realcl800_t1_s13.txt @@ -0,0 +1,109 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t1_s13', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='13,29,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.478 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4530 +t=10: Selected seed 195 with value = 0.4530 +Query 1/1: Action query time = 4.766 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5255 +t=26: Selected seed 195 with value = 0.5255 +Query 1/1: Action query time = 5.252 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6097 +t=42: Selected seed 195 with value = 0.6097 +Query 1/1: Action query time = 5.180 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6992 +t=58: Selected seed 195 with value = 0.6992 +Query 1/1: Action query time = 4.885 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8198 +t=74: Selected seed 195 with value = 0.8198 +Query 1/1: Action query time = 4.649 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9738 +t=90: Selected seed 195 with value = 0.9738 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s13/2026_08_02-03_38_05--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 2.672 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4913 +t=10: Selected seed 195 with value = 0.4913 +Query 1/1: Action query time = 5.771 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5370 +t=26: Selected seed 195 with value = 0.5370 +Query 1/1: Action query time = 5.192 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6250 +t=42: Selected seed 195 with value = 0.6250 +Query 1/1: Action query time = 4.568 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7412 +t=58: Selected seed 195 with value = 0.7412 +Query 1/1: Action query time = 4.490 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8387 +t=74: Selected seed 195 with value = 0.8387 +Query 1/1: Action query time = 5.322 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=90: Selected seed 195 with value = 0.9914 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s13/2026_08_02-03_38_05--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 2.226 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3828 +t=10: Selected seed 195 with value = 0.3828 +Query 1/1: Action query time = 3.859 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4945 +t=26: Selected seed 195 with value = 0.4945 +Query 1/1: Action query time = 5.305 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5518 +t=42: Selected seed 195 with value = 0.5518 +Query 1/1: Action query time = 4.962 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6197 +t=58: Selected seed 195 with value = 0.6197 +Query 1/1: Action query time = 5.002 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7267 +t=74: Selected seed 195 with value = 0.7267 +Query 1/1: Action query time = 4.754 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8714 +t=90: Selected seed 195 with value = 0.8714 +Query 1/1: Action query time = 3.552 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=106: Selected seed 195 with value = 0.9914 +Saved rollout MP4 at path ./rollouts/realcl800_t1_s13/2026_08_02-03_38_05--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_40_30--realcl800_t2_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_40_30--realcl800_t2_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..f879230c090080060e93dba6e4714e2642d746ac --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_40_30--realcl800_t2_s03.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t2_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.192 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4112 +t=10: Selected seed 195 with value = 0.4112 +Query 1/1: Action query time = 2.182 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4854 +t=26: Selected seed 195 with value = 0.4854 +Query 1/1: Action query time = 4.322 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6069 +t=42: Selected seed 195 with value = 0.6069 +Query 1/1: Action query time = 4.603 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7701 +t=58: Selected seed 195 with value = 0.7701 +Query 1/1: Action query time = 5.010 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9070 +t=74: Selected seed 195 with value = 0.9070 +Query 1/1: Action query time = 5.033 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9939 +t=90: Selected seed 195 with value = 0.9939 +Saved rollout MP4 at path ./rollouts/realcl800_t2_s03/2026_08_02-03_40_30--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.165 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4250 +t=10: Selected seed 195 with value = 0.4250 +Query 1/1: Action query time = 2.031 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5144 +t=26: Selected seed 195 with value = 0.5144 +Query 1/1: Action query time = 2.424 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6275 +t=42: Selected seed 195 with value = 0.6275 +Query 1/1: Action query time = 4.967 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7394 +t=58: Selected seed 195 with value = 0.7394 +Query 1/1: Action query time = 4.943 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8547 +t=74: Selected seed 195 with value = 0.8547 +Query 1/1: Action query time = 4.892 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=90: Selected seed 195 with value = 0.9981 +Saved rollout MP4 at path ./rollouts/realcl800_t2_s03/2026_08_02-03_40_30--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.762 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4223 +t=10: Selected seed 195 with value = 0.4223 +Query 1/1: Action query time = 2.176 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4831 +t=26: Selected seed 195 with value = 0.4831 +Query 1/1: Action query time = 4.511 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6019 +t=42: Selected seed 195 with value = 0.6019 +Query 1/1: Action query time = 3.483 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6927 +t=58: Selected seed 195 with value = 0.6927 +Query 1/1: Action query time = 5.044 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8448 +t=74: Selected seed 195 with value = 0.8448 +Query 1/1: Action query time = 5.340 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9950 +t=90: Selected seed 195 with value = 0.9950 +Saved rollout MP4 at path ./rollouts/realcl800_t2_s03/2026_08_02-03_40_30--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_40_30--realcl800_t2_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_40_30--realcl800_t2_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..4fc7af9fb4e11b88d8280683af00a6c8a83f6037 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_40_30--realcl800_t2_s06.txt @@ -0,0 +1,109 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t2_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.012 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3111 +t=10: Selected seed 195 with value = 0.3111 +Query 1/1: Action query time = 3.339 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3726 +t=26: Selected seed 195 with value = 0.3726 +Query 1/1: Action query time = 5.224 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4829 +t=42: Selected seed 195 with value = 0.4829 +Query 1/1: Action query time = 5.300 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5939 +t=58: Selected seed 195 with value = 0.5939 +Query 1/1: Action query time = 5.216 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7151 +t=74: Selected seed 195 with value = 0.7151 +Query 1/1: Action query time = 3.737 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8413 +t=90: Selected seed 195 with value = 0.8413 +Query 1/1: Action query time = 3.785 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=106: Selected seed 195 with value = 0.9886 +Saved rollout MP4 at path ./rollouts/realcl800_t2_s06/2026_08_02-03_40_30--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.310 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4298 +t=10: Selected seed 195 with value = 0.4298 +Query 1/1: Action query time = 3.708 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5150 +t=26: Selected seed 195 with value = 0.5150 +Query 1/1: Action query time = 5.364 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5778 +t=42: Selected seed 195 with value = 0.5778 +Query 1/1: Action query time = 5.072 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7322 +t=58: Selected seed 195 with value = 0.7322 +Query 1/1: Action query time = 3.705 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8631 +t=74: Selected seed 195 with value = 0.8631 +Query 1/1: Action query time = 4.671 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=90: Selected seed 195 with value = 0.9961 +Saved rollout MP4 at path ./rollouts/realcl800_t2_s06/2026_08_02-03_40_30--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.334 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4359 +t=10: Selected seed 195 with value = 0.4359 +Query 1/1: Action query time = 4.037 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4859 +t=26: Selected seed 195 with value = 0.4859 +Query 1/1: Action query time = 4.746 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6173 +t=42: Selected seed 195 with value = 0.6173 +Query 1/1: Action query time = 5.188 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7210 +t=58: Selected seed 195 with value = 0.7210 +Query 1/1: Action query time = 3.818 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8560 +t=74: Selected seed 195 with value = 0.8560 +Query 1/1: Action query time = 3.300 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=90: Selected seed 195 with value = 0.9809 +Saved rollout MP4 at path ./rollouts/realcl800_t2_s06/2026_08_02-03_40_30--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_25--realcl800_t3_s12.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_25--realcl800_t3_s12.txt new file mode 100644 index 0000000000000000000000000000000000000000..d174f1c855fb76f3e576b647c9adce7895b6d9aa --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_25--realcl800_t3_s12.txt @@ -0,0 +1,185 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t3_s12', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='12,28,44', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 3.031 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2076 +t=10: Selected seed 195 with value = 0.2076 +Query 1/1: Action query time = 4.753 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2214 +t=26: Selected seed 195 with value = 0.2214 +Query 1/1: Action query time = 5.580 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2214 +t=42: Selected seed 195 with value = 0.2214 +Query 1/1: Action query time = 5.019 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2283 +t=58: Selected seed 195 with value = 0.2283 +Query 1/1: Action query time = 4.953 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2345 +t=74: Selected seed 195 with value = 0.2345 +Query 1/1: Action query time = 5.169 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2612 +t=90: Selected seed 195 with value = 0.2612 +Query 1/1: Action query time = 4.771 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2787 +t=106: Selected seed 195 with value = 0.2787 +Query 1/1: Action query time = 4.827 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4825 +t=122: Selected seed 195 with value = 0.4825 +Query 1/1: Action query time = 5.137 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5777 +t=138: Selected seed 195 with value = 0.5777 +Query 1/1: Action query time = 5.140 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7096 +t=154: Selected seed 195 with value = 0.7096 +Query 1/1: Action query time = 4.515 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8349 +t=170: Selected seed 195 with value = 0.8349 +Query 1/1: Action query time = 4.672 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9802 +t=186: Selected seed 195 with value = 0.9802 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s12/2026_08_02-03_43_25--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 3.584 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1773 +t=10: Selected seed 195 with value = 0.1773 +Query 1/1: Action query time = 5.041 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1959 +t=26: Selected seed 195 with value = 0.1959 +Query 1/1: Action query time = 5.305 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2333 +t=42: Selected seed 195 with value = 0.2333 +Query 1/1: Action query time = 4.693 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2617 +t=58: Selected seed 195 with value = 0.2617 +Query 1/1: Action query time = 4.666 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2976 +t=74: Selected seed 195 with value = 0.2976 +Query 1/1: Action query time = 4.869 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2010 +t=90: Selected seed 195 with value = 0.2010 +Query 1/1: Action query time = 3.971 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2105 +t=106: Selected seed 195 with value = 0.2105 +Query 1/1: Action query time = 4.018 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5329 +t=122: Selected seed 195 with value = 0.5329 +Query 1/1: Action query time = 5.204 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5724 +t=138: Selected seed 195 with value = 0.5724 +Query 1/1: Action query time = 5.262 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4947 +t=154: Selected seed 195 with value = 0.4947 +Query 1/1: Action query time = 5.036 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5779 +t=170: Selected seed 195 with value = 0.5779 +Query 1/1: Action query time = 4.821 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6854 +t=186: Selected seed 195 with value = 0.6854 +Query 1/1: Action query time = 3.153 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7730 +t=202: Selected seed 195 with value = 0.7730 +Query 1/1: Action query time = 3.499 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8989 +t=218: Selected seed 195 with value = 0.8989 +Query 1/1: Action query time = 5.377 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9923 +t=234: Selected seed 195 with value = 0.9923 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s12/2026_08_02-03_43_25--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.795 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1933 +t=10: Selected seed 195 with value = 0.1933 +Query 1/1: Action query time = 3.988 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2212 +t=26: Selected seed 195 with value = 0.2212 +Query 1/1: Action query time = 3.322 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2445 +t=42: Selected seed 195 with value = 0.2445 +Query 1/1: Action query time = 4.918 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2880 +t=58: Selected seed 195 with value = 0.2880 +Query 1/1: Action query time = 4.082 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3497 +t=74: Selected seed 195 with value = 0.3497 +Query 1/1: Action query time = 5.194 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4125 +t=90: Selected seed 195 with value = 0.4125 +Query 1/1: Action query time = 5.332 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4878 +t=106: Selected seed 195 with value = 0.4878 +Query 1/1: Action query time = 5.223 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5484 +t=122: Selected seed 195 with value = 0.5484 +Query 1/1: Action query time = 3.527 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6650 +t=138: Selected seed 195 with value = 0.6650 +Query 1/1: Action query time = 3.773 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7725 +t=154: Selected seed 195 with value = 0.7725 +Query 1/1: Action query time = 3.702 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9234 +t=170: Selected seed 195 with value = 0.9234 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s12/2026_08_02-03_43_25--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_26--realcl800_t3_s09.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_26--realcl800_t3_s09.txt new file mode 100644 index 0000000000000000000000000000000000000000..1f0f5fc0bb8aa4e64df94dec07ff0e1fe8f43499 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_26--realcl800_t3_s09.txt @@ -0,0 +1,169 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t3_s09', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='9,25,41', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 5.623 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1413 +t=10: Selected seed 195 with value = 0.1413 +Query 1/1: Action query time = 5.678 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2253 +t=26: Selected seed 195 with value = 0.2253 +Query 1/1: Action query time = 5.213 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2423 +t=42: Selected seed 195 with value = 0.2423 +Query 1/1: Action query time = 4.863 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2992 +t=58: Selected seed 195 with value = 0.2992 +Query 1/1: Action query time = 5.092 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1817 +t=74: Selected seed 195 with value = 0.1817 +Query 1/1: Action query time = 4.779 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4659 +t=90: Selected seed 195 with value = 0.4659 +Query 1/1: Action query time = 5.154 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5401 +t=106: Selected seed 195 with value = 0.5401 +Query 1/1: Action query time = 5.137 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5740 +t=122: Selected seed 195 with value = 0.5740 +Query 1/1: Action query time = 5.243 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6551 +t=138: Selected seed 195 with value = 0.6551 +Query 1/1: Action query time = 4.392 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7677 +t=154: Selected seed 195 with value = 0.7677 +Query 1/1: Action query time = 4.541 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8492 +t=170: Selected seed 195 with value = 0.8492 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s09/2026_08_02-03_43_26--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.655 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1684 +t=10: Selected seed 195 with value = 0.1684 +Query 1/1: Action query time = 4.836 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2138 +t=26: Selected seed 195 with value = 0.2138 +Query 1/1: Action query time = 5.310 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2110 +t=42: Selected seed 195 with value = 0.2110 +Query 1/1: Action query time = 5.299 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2382 +t=58: Selected seed 195 with value = 0.2382 +Query 1/1: Action query time = 5.161 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2630 +t=74: Selected seed 195 with value = 0.2630 +Query 1/1: Action query time = 3.839 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3907 +t=90: Selected seed 195 with value = 0.3907 +Query 1/1: Action query time = 4.053 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4892 +t=106: Selected seed 195 with value = 0.4892 +Query 1/1: Action query time = 4.814 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5520 +t=122: Selected seed 195 with value = 0.5520 +Query 1/1: Action query time = 4.933 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6292 +t=138: Selected seed 195 with value = 0.6292 +Query 1/1: Action query time = 4.359 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7459 +t=154: Selected seed 195 with value = 0.7459 +Query 1/1: Action query time = 4.594 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8624 +t=170: Selected seed 195 with value = 0.8624 +Query 1/1: Action query time = 4.132 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9844 +t=186: Selected seed 195 with value = 0.9844 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s09/2026_08_02-03_43_26--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 3.169 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1930 +t=10: Selected seed 195 with value = 0.1930 +Query 1/1: Action query time = 5.156 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2318 +t=26: Selected seed 195 with value = 0.2318 +Query 1/1: Action query time = 4.981 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2392 +t=42: Selected seed 195 with value = 0.2392 +Query 1/1: Action query time = 4.968 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2264 +t=58: Selected seed 195 with value = 0.2264 +Query 1/1: Action query time = 3.472 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2488 +t=74: Selected seed 195 with value = 0.2488 +Query 1/1: Action query time = 3.429 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2759 +t=90: Selected seed 195 with value = 0.2759 +Query 1/1: Action query time = 5.073 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3557 +t=106: Selected seed 195 with value = 0.3557 +Query 1/1: Action query time = 3.867 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4010 +t=122: Selected seed 195 with value = 0.4010 +Query 1/1: Action query time = 5.179 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6152 +t=138: Selected seed 195 with value = 0.6152 +Query 1/1: Action query time = 5.483 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7349 +t=154: Selected seed 195 with value = 0.7349 +Query 1/1: Action query time = 6.097 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8918 +t=170: Selected seed 195 with value = 0.8918 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s09/2026_08_02-03_43_26--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_26--realcl800_t3_s10.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_26--realcl800_t3_s10.txt new file mode 100644 index 0000000000000000000000000000000000000000..0b8289935ad5485fb4d1c72a70a994f908b1e064 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_26--realcl800_t3_s10.txt @@ -0,0 +1,197 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t3_s10', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='10,26,42', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 5.358 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2177 +t=10: Selected seed 195 with value = 0.2177 +Query 1/1: Action query time = 4.518 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2464 +t=26: Selected seed 195 with value = 0.2464 +Query 1/1: Action query time = 4.555 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2768 +t=42: Selected seed 195 with value = 0.2768 +Query 1/1: Action query time = 5.370 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3184 +t=58: Selected seed 195 with value = 0.3184 +Query 1/1: Action query time = 5.097 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3717 +t=74: Selected seed 195 with value = 0.3717 +Query 1/1: Action query time = 4.915 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4294 +t=90: Selected seed 195 with value = 0.4294 +Query 1/1: Action query time = 5.325 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4987 +t=106: Selected seed 195 with value = 0.4987 +Query 1/1: Action query time = 4.939 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5484 +t=122: Selected seed 195 with value = 0.5484 +Query 1/1: Action query time = 4.396 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7153 +t=138: Selected seed 195 with value = 0.7153 +Query 1/1: Action query time = 4.940 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7969 +t=154: Selected seed 195 with value = 0.7969 +Query 1/1: Action query time = 5.558 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9253 +t=170: Selected seed 195 with value = 0.9253 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s10/2026_08_02-03_43_26--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 1.821 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2047 +t=10: Selected seed 195 with value = 0.2047 +Query 1/1: Action query time = 2.751 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2197 +t=26: Selected seed 195 with value = 0.2197 +Query 1/1: Action query time = 5.804 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2381 +t=42: Selected seed 195 with value = 0.2381 +Query 1/1: Action query time = 4.919 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2839 +t=58: Selected seed 195 with value = 0.2839 +Query 1/1: Action query time = 4.680 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3357 +t=74: Selected seed 195 with value = 0.3357 +Query 1/1: Action query time = 4.906 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4047 +t=90: Selected seed 195 with value = 0.4047 +Query 1/1: Action query time = 5.291 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4906 +t=106: Selected seed 195 with value = 0.4906 +Query 1/1: Action query time = 3.847 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5740 +t=122: Selected seed 195 with value = 0.5740 +Query 1/1: Action query time = 3.879 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6882 +t=138: Selected seed 195 with value = 0.6882 +Query 1/1: Action query time = 5.084 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8146 +t=154: Selected seed 195 with value = 0.8146 +Query 1/1: Action query time = 4.803 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9181 +t=170: Selected seed 195 with value = 0.9181 +Query 1/1: Action query time = 5.248 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2614 +t=186: Selected seed 195 with value = 0.2614 +Query 1/1: Action query time = 4.855 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2598 +t=202: Selected seed 195 with value = 0.2598 +Query 1/1: Action query time = 2.562 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2806 +t=218: Selected seed 195 with value = 0.2806 +Query 1/1: Action query time = 3.766 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2500 +t=234: Selected seed 195 with value = 0.2500 +Query 1/1: Action query time = 3.185 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2936 +t=250: Selected seed 195 with value = 0.2936 +Query 1/1: Action query time = 5.220 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2410 +t=266: Selected seed 195 with value = 0.2410 +Query 1/1: Action query time = 5.111 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2444 +t=282: Selected seed 195 with value = 0.2444 +Query 1/1: Action query time = 5.220 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2473 +t=298: Selected seed 195 with value = 0.2473 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s10/2026_08_02-03_43_26--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.907 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1876 +t=10: Selected seed 195 with value = 0.1876 +Query 1/1: Action query time = 3.904 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2198 +t=26: Selected seed 195 with value = 0.2198 +Query 1/1: Action query time = 5.102 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2358 +t=42: Selected seed 195 with value = 0.2358 +Query 1/1: Action query time = 5.466 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2984 +t=58: Selected seed 195 with value = 0.2984 +Query 1/1: Action query time = 6.056 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3750 +t=74: Selected seed 195 with value = 0.3750 +Query 1/1: Action query time = 4.507 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4415 +t=90: Selected seed 195 with value = 0.4415 +Query 1/1: Action query time = 3.628 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4945 +t=106: Selected seed 195 with value = 0.4945 +Query 1/1: Action query time = 2.879 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4960 +t=122: Selected seed 195 with value = 0.4960 +Query 1/1: Action query time = 2.643 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6772 +t=138: Selected seed 195 with value = 0.6772 +Query 1/1: Action query time = 2.480 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7895 +t=154: Selected seed 195 with value = 0.7895 +Query 1/1: Action query time = 2.043 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9690 +t=170: Selected seed 195 with value = 0.9690 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s10/2026_08_02-03_43_26--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_26--realcl800_t3_s15.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_26--realcl800_t3_s15.txt new file mode 100644 index 0000000000000000000000000000000000000000..2990f470521620abca92fddbf65140276b885ef4 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_43_26--realcl800_t3_s15.txt @@ -0,0 +1,205 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t3_s15', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='15,31,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 5.820 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1676 +t=10: Selected seed 195 with value = 0.1676 +Query 1/1: Action query time = 4.825 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1895 +t=26: Selected seed 195 with value = 0.1895 +Query 1/1: Action query time = 5.107 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2282 +t=42: Selected seed 195 with value = 0.2282 +Query 1/1: Action query time = 5.147 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2498 +t=58: Selected seed 195 with value = 0.2498 +Query 1/1: Action query time = 5.348 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2935 +t=74: Selected seed 195 with value = 0.2935 +Query 1/1: Action query time = 4.916 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3853 +t=90: Selected seed 195 with value = 0.3853 +Query 1/1: Action query time = 4.711 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4710 +t=106: Selected seed 195 with value = 0.4710 +Query 1/1: Action query time = 4.975 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5839 +t=122: Selected seed 195 with value = 0.5839 +Query 1/1: Action query time = 5.413 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4840 +t=138: Selected seed 195 with value = 0.4840 +Query 1/1: Action query time = 5.358 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7199 +t=154: Selected seed 195 with value = 0.7199 +Query 1/1: Action query time = 4.033 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8617 +t=170: Selected seed 195 with value = 0.8617 +Query 1/1: Action query time = 1.618 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8523 +t=186: Selected seed 195 with value = 0.8523 +Query 1/1: Action query time = 2.572 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8782 +t=202: Selected seed 195 with value = 0.8782 +Query 1/1: Action query time = 3.120 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8549 +t=218: Selected seed 195 with value = 0.8549 +Query 1/1: Action query time = 5.047 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3271 +t=234: Selected seed 195 with value = 0.3271 +Query 1/1: Action query time = 4.989 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3281 +t=250: Selected seed 195 with value = 0.3281 +Query 1/1: Action query time = 4.872 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3231 +t=266: Selected seed 195 with value = 0.3231 +Query 1/1: Action query time = 4.833 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8870 +t=282: Selected seed 195 with value = 0.8870 +Query 1/1: Action query time = 5.031 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9545 +t=298: Selected seed 195 with value = 0.9545 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s15/2026_08_02-03_43_26--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 5.230 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1823 +t=10: Selected seed 195 with value = 0.1823 +Query 1/1: Action query time = 4.854 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2197 +t=26: Selected seed 195 with value = 0.2197 +Query 1/1: Action query time = 5.246 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2119 +t=42: Selected seed 195 with value = 0.2119 +Query 1/1: Action query time = 4.793 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2429 +t=58: Selected seed 195 with value = 0.2429 +Query 1/1: Action query time = 3.008 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2079 +t=74: Selected seed 195 with value = 0.2079 +Query 1/1: Action query time = 4.312 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3446 +t=90: Selected seed 195 with value = 0.3446 +Query 1/1: Action query time = 5.102 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4776 +t=106: Selected seed 195 with value = 0.4776 +Query 1/1: Action query time = 3.215 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4951 +t=122: Selected seed 195 with value = 0.4951 +Query 1/1: Action query time = 4.700 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5930 +t=138: Selected seed 195 with value = 0.5930 +Query 1/1: Action query time = 5.527 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7135 +t=154: Selected seed 195 with value = 0.7135 +Query 1/1: Action query time = 4.219 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8822 +t=170: Selected seed 195 with value = 0.8822 +Query 1/1: Action query time = 4.726 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9471 +t=186: Selected seed 195 with value = 0.9471 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s15/2026_08_02-03_43_26--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 5.659 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1740 +t=10: Selected seed 195 with value = 0.1740 +Query 1/1: Action query time = 4.993 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2018 +t=26: Selected seed 195 with value = 0.2018 +Query 1/1: Action query time = 4.275 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2174 +t=42: Selected seed 195 with value = 0.2174 +Query 1/1: Action query time = 4.360 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2541 +t=58: Selected seed 195 with value = 0.2541 +Query 1/1: Action query time = 4.054 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2993 +t=74: Selected seed 195 with value = 0.2993 +Query 1/1: Action query time = 3.902 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3654 +t=90: Selected seed 195 with value = 0.3654 +Query 1/1: Action query time = 1.809 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4454 +t=106: Selected seed 195 with value = 0.4454 +Query 1/1: Action query time = 2.454 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5308 +t=122: Selected seed 195 with value = 0.5308 +Query 1/1: Action query time = 2.348 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6328 +t=138: Selected seed 195 with value = 0.6328 +Query 1/1: Action query time = 2.131 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7321 +t=154: Selected seed 195 with value = 0.7321 +Query 1/1: Action query time = 0.965 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8237 +t=170: Selected seed 195 with value = 0.8237 +Query 1/1: Action query time = 1.192 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9515 +t=186: Selected seed 195 with value = 0.9515 +Saved rollout MP4 at path ./rollouts/realcl800_t3_s15/2026_08_02-03_43_26--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_19--realcl800_t4_s00.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_19--realcl800_t4_s00.txt new file mode 100644 index 0000000000000000000000000000000000000000..71f9ece564e8a1befb2545ba1a4e84a6c55b3516 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_19--realcl800_t4_s00.txt @@ -0,0 +1,184 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t4_s00', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,16,32,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.638 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4265 +t=10: Selected seed 195 with value = 0.4265 +Query 1/1: Action query time = 4.803 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5170 +t=26: Selected seed 195 with value = 0.5170 +Query 1/1: Action query time = 5.069 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5849 +t=42: Selected seed 195 with value = 0.5849 +Query 1/1: Action query time = 4.848 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6899 +t=58: Selected seed 195 with value = 0.6899 +Query 1/1: Action query time = 4.641 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8043 +t=74: Selected seed 195 with value = 0.8043 +Query 1/1: Action query time = 4.486 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9690 +t=90: Selected seed 195 with value = 0.9690 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s00/2026_08_02-03_48_19--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.583 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4766 +t=10: Selected seed 195 with value = 0.4766 +Query 1/1: Action query time = 5.092 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5609 +t=26: Selected seed 195 with value = 0.5609 +Query 1/1: Action query time = 4.975 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6646 +t=42: Selected seed 195 with value = 0.6646 +Query 1/1: Action query time = 5.089 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7329 +t=58: Selected seed 195 with value = 0.7329 +Query 1/1: Action query time = 4.840 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8795 +t=74: Selected seed 195 with value = 0.8795 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s00/2026_08_02-03_48_19--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.983 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4723 +t=10: Selected seed 195 with value = 0.4723 +Query 1/1: Action query time = 3.265 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5559 +t=26: Selected seed 195 with value = 0.5559 +Query 1/1: Action query time = 5.021 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6513 +t=42: Selected seed 195 with value = 0.6513 +Query 1/1: Action query time = 5.140 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7291 +t=58: Selected seed 195 with value = 0.7291 +Query 1/1: Action query time = 4.989 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8643 +t=74: Selected seed 195 with value = 0.8643 +Query 1/1: Action query time = 4.379 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9533 +t=90: Selected seed 195 with value = 0.9533 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s00/2026_08_02-03_48_19--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 3.265 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4409 +t=10: Selected seed 195 with value = 0.4409 +Query 1/1: Action query time = 2.267 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5533 +t=26: Selected seed 195 with value = 0.5533 +Query 1/1: Action query time = 2.047 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6066 +t=42: Selected seed 195 with value = 0.6066 +Query 1/1: Action query time = 1.868 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6843 +t=58: Selected seed 195 with value = 0.6843 +Query 1/1: Action query time = 1.840 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7927 +t=74: Selected seed 195 with value = 0.7927 +Query 1/1: Action query time = 0.956 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9217 +t=90: Selected seed 195 with value = 0.9217 +Query 1/1: Action query time = 0.956 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9085 +t=106: Selected seed 195 with value = 0.9085 +Query 1/1: Action query time = 0.968 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9732 +t=122: Selected seed 195 with value = 0.9732 +Query 1/1: Action query time = 0.961 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9656 +t=138: Selected seed 195 with value = 0.9656 +Query 1/1: Action query time = 0.957 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9579 +t=154: Selected seed 195 with value = 0.9579 +Query 1/1: Action query time = 0.950 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8882 +t=170: Selected seed 195 with value = 0.8882 +Query 1/1: Action query time = 0.954 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8201 +t=186: Selected seed 195 with value = 0.8201 +Query 1/1: Action query time = 0.986 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7390 +t=202: Selected seed 195 with value = 0.7390 +Query 1/1: Action query time = 1.103 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8202 +t=218: Selected seed 195 with value = 0.8202 +Query 1/1: Action query time = 1.143 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8732 +t=234: Selected seed 195 with value = 0.8732 +Query 1/1: Action query time = 0.967 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9251 +t=250: Selected seed 195 with value = 0.9251 +Query 1/1: Action query time = 0.983 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9648 +t=266: Selected seed 195 with value = 0.9648 +Query 1/1: Action query time = 0.944 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9821 +t=282: Selected seed 195 with value = 0.9821 +Query 1/1: Action query time = 0.958 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9793 +t=298: Selected seed 195 with value = 0.9793 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s00/2026_08_02-03_48_19--with_future_img--episode=4--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 4 +# successes: 3 (75.0%) +Current task success rate: 0.75 +Current total success rate: 0.75 +Final results: +Total episodes: 4 +Total successes: 3 +Overall success rate: 0.7500 (75.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_19--realcl800_t4_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_19--realcl800_t4_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..053d03353cbcdae4da24111926a534cf9adc211a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_19--realcl800_t4_s03.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t4_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.286 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4407 +t=10: Selected seed 195 with value = 0.4407 +Query 1/1: Action query time = 4.812 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5333 +t=26: Selected seed 195 with value = 0.5333 +Query 1/1: Action query time = 4.961 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6192 +t=42: Selected seed 195 with value = 0.6192 +Query 1/1: Action query time = 4.953 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7211 +t=58: Selected seed 195 with value = 0.7211 +Query 1/1: Action query time = 5.337 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8764 +t=74: Selected seed 195 with value = 0.8764 +Query 1/1: Action query time = 3.846 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9923 +t=90: Selected seed 195 with value = 0.9923 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s03/2026_08_02-03_48_19--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.148 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4279 +t=10: Selected seed 195 with value = 0.4279 +Query 1/1: Action query time = 5.348 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5220 +t=26: Selected seed 195 with value = 0.5220 +Query 1/1: Action query time = 5.086 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5977 +t=42: Selected seed 195 with value = 0.5977 +Query 1/1: Action query time = 5.203 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7373 +t=58: Selected seed 195 with value = 0.7373 +Query 1/1: Action query time = 4.309 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8597 +t=74: Selected seed 195 with value = 0.8597 +Query 1/1: Action query time = 3.424 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=90: Selected seed 195 with value = 0.9978 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s03/2026_08_02-03_48_19--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.034 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4601 +t=10: Selected seed 195 with value = 0.4601 +Query 1/1: Action query time = 4.730 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5256 +t=26: Selected seed 195 with value = 0.5256 +Query 1/1: Action query time = 5.282 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6191 +t=42: Selected seed 195 with value = 0.6191 +Query 1/1: Action query time = 4.922 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7014 +t=58: Selected seed 195 with value = 0.7014 +Query 1/1: Action query time = 4.052 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8425 +t=74: Selected seed 195 with value = 0.8425 +Query 1/1: Action query time = 3.979 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9758 +t=90: Selected seed 195 with value = 0.9758 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s03/2026_08_02-03_48_19--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_20--realcl800_t4_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_20--realcl800_t4_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..1d42a3a97f82093fbba5e110616fe676a932979c --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_20--realcl800_t4_s08.txt @@ -0,0 +1,97 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t4_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.748 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4444 +t=10: Selected seed 195 with value = 0.4444 +Query 1/1: Action query time = 4.816 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5292 +t=26: Selected seed 195 with value = 0.5292 +Query 1/1: Action query time = 5.076 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6014 +t=42: Selected seed 195 with value = 0.6014 +Query 1/1: Action query time = 5.339 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7452 +t=58: Selected seed 195 with value = 0.7452 +Query 1/1: Action query time = 4.454 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9079 +t=74: Selected seed 195 with value = 0.9079 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s08/2026_08_02-03_48_20--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.250 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4330 +t=10: Selected seed 195 with value = 0.4330 +Query 1/1: Action query time = 3.940 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5269 +t=26: Selected seed 195 with value = 0.5269 +Query 1/1: Action query time = 4.939 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6068 +t=42: Selected seed 195 with value = 0.6068 +Query 1/1: Action query time = 5.096 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7802 +t=58: Selected seed 195 with value = 0.7802 +Query 1/1: Action query time = 4.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9323 +t=74: Selected seed 195 with value = 0.9323 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s08/2026_08_02-03_48_20--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.526 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4468 +t=10: Selected seed 195 with value = 0.4468 +Query 1/1: Action query time = 4.074 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5437 +t=26: Selected seed 195 with value = 0.5437 +Query 1/1: Action query time = 4.941 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6406 +t=42: Selected seed 195 with value = 0.6406 +Query 1/1: Action query time = 4.956 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7267 +t=58: Selected seed 195 with value = 0.7267 +Query 1/1: Action query time = 4.044 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8844 +t=74: Selected seed 195 with value = 0.8844 +Query 1/1: Action query time = 5.067 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9750 +t=90: Selected seed 195 with value = 0.9750 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s08/2026_08_02-03_48_20--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_21--realcl800_t4_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_21--realcl800_t4_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..803b632a7e59b354415bd4667d49866aefd4a8f5 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_48_21--realcl800_t4_s11.txt @@ -0,0 +1,153 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t4_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.568 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4448 +t=10: Selected seed 195 with value = 0.4448 +Query 1/1: Action query time = 4.823 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5545 +t=26: Selected seed 195 with value = 0.5545 +Query 1/1: Action query time = 4.843 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6128 +t=42: Selected seed 195 with value = 0.6128 +Query 1/1: Action query time = 4.986 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6947 +t=58: Selected seed 195 with value = 0.6947 +Query 1/1: Action query time = 4.281 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8189 +t=74: Selected seed 195 with value = 0.8189 +Query 1/1: Action query time = 4.112 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9874 +t=90: Selected seed 195 with value = 0.9874 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s11/2026_08_02-03_48_21--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.503 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4684 +t=10: Selected seed 195 with value = 0.4684 +Query 1/1: Action query time = 5.067 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5454 +t=26: Selected seed 195 with value = 0.5454 +Query 1/1: Action query time = 4.903 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6288 +t=42: Selected seed 195 with value = 0.6288 +Query 1/1: Action query time = 4.822 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7256 +t=58: Selected seed 195 with value = 0.7256 +Query 1/1: Action query time = 2.133 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8228 +t=74: Selected seed 195 with value = 0.8228 +Query 1/1: Action query time = 3.154 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7733 +t=90: Selected seed 195 with value = 0.7733 +Query 1/1: Action query time = 4.484 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9848 +t=106: Selected seed 195 with value = 0.9848 +Query 1/1: Action query time = 4.727 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=122: Selected seed 195 with value = 0.9870 +Query 1/1: Action query time = 4.729 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9856 +t=138: Selected seed 195 with value = 0.9856 +Query 1/1: Action query time = 4.272 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9828 +t=154: Selected seed 195 with value = 0.9828 +Query 1/1: Action query time = 5.076 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=170: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 4.213 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9687 +t=186: Selected seed 195 with value = 0.9687 +Query 1/1: Action query time = 1.915 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9514 +t=202: Selected seed 195 with value = 0.9514 +Query 1/1: Action query time = 0.984 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9146 +t=218: Selected seed 195 with value = 0.9146 +Query 1/1: Action query time = 1.296 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8706 +t=234: Selected seed 195 with value = 0.8706 +Query 1/1: Action query time = 0.984 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8835 +t=250: Selected seed 195 with value = 0.8835 +Query 1/1: Action query time = 1.081 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8941 +t=266: Selected seed 195 with value = 0.8941 +Query 1/1: Action query time = 0.976 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8872 +t=282: Selected seed 195 with value = 0.8872 +Query 1/1: Action query time = 1.294 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8737 +t=298: Selected seed 195 with value = 0.8737 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s11/2026_08_02-03_48_21--with_future_img--episode=2--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4662 +t=10: Selected seed 195 with value = 0.4662 +Query 1/1: Action query time = 0.981 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5318 +t=26: Selected seed 195 with value = 0.5318 +Query 1/1: Action query time = 0.989 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6144 +t=42: Selected seed 195 with value = 0.6144 +Query 1/1: Action query time = 1.118 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6766 +t=58: Selected seed 195 with value = 0.6766 +Query 1/1: Action query time = 1.173 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8654 +t=74: Selected seed 195 with value = 0.8654 +Saved rollout MP4 at path ./rollouts/realcl800_t4_s11/2026_08_02-03_48_21--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_07--realcl800_t5_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_07--realcl800_t5_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..5fb8de8a9d6a59854605e7d9d45fd2f8a96e669a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_07--realcl800_t5_s07.txt @@ -0,0 +1,125 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t5_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 4.390 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3443 +t=10: Selected seed 195 with value = 0.3443 +Query 1/1: Action query time = 5.022 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3924 +t=26: Selected seed 195 with value = 0.3924 +Query 1/1: Action query time = 3.566 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4063 +t=42: Selected seed 195 with value = 0.4063 +Query 1/1: Action query time = 4.768 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5888 +t=58: Selected seed 195 with value = 0.5888 +Query 1/1: Action query time = 5.761 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6899 +t=74: Selected seed 195 with value = 0.6899 +Query 1/1: Action query time = 5.111 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8182 +t=90: Selected seed 195 with value = 0.8182 +Query 1/1: Action query time = 4.260 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9666 +t=106: Selected seed 195 with value = 0.9666 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s07/2026_08_02-03_51_07--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.171 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1984 +t=10: Selected seed 195 with value = 0.1984 +Query 1/1: Action query time = 3.191 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3635 +t=26: Selected seed 195 with value = 0.3635 +Query 1/1: Action query time = 3.795 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4529 +t=42: Selected seed 195 with value = 0.4529 +Query 1/1: Action query time = 5.172 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3461 +t=58: Selected seed 195 with value = 0.3461 +Query 1/1: Action query time = 5.894 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2495 +t=74: Selected seed 195 with value = 0.2495 +Query 1/1: Action query time = 4.606 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6885 +t=90: Selected seed 195 with value = 0.6885 +Query 1/1: Action query time = 4.824 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7678 +t=106: Selected seed 195 with value = 0.7678 +Query 1/1: Action query time = 4.830 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8704 +t=122: Selected seed 195 with value = 0.8704 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s07/2026_08_02-03_51_07--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 3.849 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3126 +t=10: Selected seed 195 with value = 0.3126 +Query 1/1: Action query time = 4.335 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3458 +t=26: Selected seed 195 with value = 0.3458 +Query 1/1: Action query time = 3.418 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4347 +t=42: Selected seed 195 with value = 0.4347 +Query 1/1: Action query time = 4.089 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4777 +t=58: Selected seed 195 with value = 0.4777 +Query 1/1: Action query time = 4.498 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5583 +t=74: Selected seed 195 with value = 0.5583 +Query 1/1: Action query time = 5.361 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6723 +t=90: Selected seed 195 with value = 0.6723 +Query 1/1: Action query time = 4.437 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7730 +t=106: Selected seed 195 with value = 0.7730 +Query 1/1: Action query time = 5.357 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8686 +t=122: Selected seed 195 with value = 0.8686 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s07/2026_08_02-03_51_07--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_07--realcl800_t5_s12.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_07--realcl800_t5_s12.txt new file mode 100644 index 0000000000000000000000000000000000000000..b2eee2c81e82ffad05c59fb9891162f4033e96f7 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_07--realcl800_t5_s12.txt @@ -0,0 +1,133 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t5_s12', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='12,28,44', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 2.329 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3118 +t=10: Selected seed 195 with value = 0.3118 +Query 1/1: Action query time = 3.404 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2742 +t=26: Selected seed 195 with value = 0.2742 +Query 1/1: Action query time = 4.884 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3723 +t=42: Selected seed 195 with value = 0.3723 +Query 1/1: Action query time = 5.138 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5304 +t=58: Selected seed 195 with value = 0.5304 +Query 1/1: Action query time = 3.147 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6229 +t=74: Selected seed 195 with value = 0.6229 +Query 1/1: Action query time = 5.209 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7311 +t=90: Selected seed 195 with value = 0.7311 +Query 1/1: Action query time = 5.382 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8293 +t=106: Selected seed 195 with value = 0.8293 +Query 1/1: Action query time = 5.230 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9763 +t=122: Selected seed 195 with value = 0.9763 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s12/2026_08_02-03_51_07--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.080 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2272 +t=10: Selected seed 195 with value = 0.2272 +Query 1/1: Action query time = 2.904 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2727 +t=26: Selected seed 195 with value = 0.2727 +Query 1/1: Action query time = 4.782 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3115 +t=42: Selected seed 195 with value = 0.3115 +Query 1/1: Action query time = 4.160 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4661 +t=58: Selected seed 195 with value = 0.4661 +Query 1/1: Action query time = 5.511 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5028 +t=74: Selected seed 195 with value = 0.5028 +Query 1/1: Action query time = 5.356 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6366 +t=90: Selected seed 195 with value = 0.6366 +Query 1/1: Action query time = 5.190 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7550 +t=106: Selected seed 195 with value = 0.7550 +Query 1/1: Action query time = 4.922 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9000 +t=122: Selected seed 195 with value = 0.9000 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s12/2026_08_02-03_51_07--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 4.404 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3051 +t=10: Selected seed 195 with value = 0.3051 +Query 1/1: Action query time = 4.417 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1652 +t=26: Selected seed 195 with value = 0.1652 +Query 1/1: Action query time = 3.781 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2380 +t=42: Selected seed 195 with value = 0.2380 +Query 1/1: Action query time = 2.676 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4433 +t=58: Selected seed 195 with value = 0.4433 +Query 1/1: Action query time = 3.411 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5344 +t=74: Selected seed 195 with value = 0.5344 +Query 1/1: Action query time = 5.692 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2988 +t=90: Selected seed 195 with value = 0.2988 +Query 1/1: Action query time = 5.297 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3790 +t=106: Selected seed 195 with value = 0.3790 +Query 1/1: Action query time = 5.664 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7465 +t=122: Selected seed 195 with value = 0.7465 +Query 1/1: Action query time = 5.055 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8604 +t=138: Selected seed 195 with value = 0.8604 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s12/2026_08_02-03_51_07--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_08--realcl800_t5_s13.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_08--realcl800_t5_s13.txt new file mode 100644 index 0000000000000000000000000000000000000000..4d57829c0146a28f535cc8b8cea45dc882a43557 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_08--realcl800_t5_s13.txt @@ -0,0 +1,137 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t5_s13', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='13,29,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 4.780 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1790 +t=10: Selected seed 195 with value = 0.1790 +Query 1/1: Action query time = 4.675 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2248 +t=26: Selected seed 195 with value = 0.2248 +Query 1/1: Action query time = 5.900 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4713 +t=42: Selected seed 195 with value = 0.4713 +Query 1/1: Action query time = 5.865 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3378 +t=58: Selected seed 195 with value = 0.3378 +Query 1/1: Action query time = 4.406 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3669 +t=74: Selected seed 195 with value = 0.3669 +Query 1/1: Action query time = 3.896 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4227 +t=90: Selected seed 195 with value = 0.4227 +Query 1/1: Action query time = 4.723 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5048 +t=106: Selected seed 195 with value = 0.5048 +Query 1/1: Action query time = 4.833 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6029 +t=122: Selected seed 195 with value = 0.6029 +Query 1/1: Action query time = 3.303 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6957 +t=138: Selected seed 195 with value = 0.6957 +Query 1/1: Action query time = 2.838 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8171 +t=154: Selected seed 195 with value = 0.8171 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s13/2026_08_02-03_51_08--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 5.469 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3265 +t=10: Selected seed 195 with value = 0.3265 +Query 1/1: Action query time = 4.831 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4013 +t=26: Selected seed 195 with value = 0.4013 +Query 1/1: Action query time = 5.076 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5227 +t=42: Selected seed 195 with value = 0.5227 +Query 1/1: Action query time = 4.873 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5741 +t=58: Selected seed 195 with value = 0.5741 +Query 1/1: Action query time = 4.919 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5842 +t=74: Selected seed 195 with value = 0.5842 +Query 1/1: Action query time = 4.849 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6879 +t=90: Selected seed 195 with value = 0.6879 +Query 1/1: Action query time = 4.587 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7803 +t=106: Selected seed 195 with value = 0.7803 +Query 1/1: Action query time = 3.435 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8881 +t=122: Selected seed 195 with value = 0.8881 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s13/2026_08_02-03_51_08--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 4.231 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2358 +t=10: Selected seed 195 with value = 0.2358 +Query 1/1: Action query time = 5.764 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3017 +t=26: Selected seed 195 with value = 0.3017 +Query 1/1: Action query time = 5.121 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3370 +t=42: Selected seed 195 with value = 0.3370 +Query 1/1: Action query time = 4.860 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3560 +t=58: Selected seed 195 with value = 0.3560 +Query 1/1: Action query time = 4.704 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6016 +t=74: Selected seed 195 with value = 0.6016 +Query 1/1: Action query time = 3.521 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6357 +t=90: Selected seed 195 with value = 0.6357 +Query 1/1: Action query time = 4.413 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7900 +t=106: Selected seed 195 with value = 0.7900 +Query 1/1: Action query time = 4.007 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9012 +t=122: Selected seed 195 with value = 0.9012 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s13/2026_08_02-03_51_08--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_08--realcl800_t5_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_08--realcl800_t5_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..25cd2368abc18be4fa1961ae0d124339b556dfa2 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_51_08--realcl800_t5_s14.txt @@ -0,0 +1,133 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t5_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 5.484 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2399 +t=10: Selected seed 195 with value = 0.2399 +Query 1/1: Action query time = 5.133 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2510 +t=26: Selected seed 195 with value = 0.2510 +Query 1/1: Action query time = 5.884 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3392 +t=42: Selected seed 195 with value = 0.3392 +Query 1/1: Action query time = 5.358 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5159 +t=58: Selected seed 195 with value = 0.5159 +Query 1/1: Action query time = 4.289 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2853 +t=74: Selected seed 195 with value = 0.2853 +Query 1/1: Action query time = 4.323 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3566 +t=90: Selected seed 195 with value = 0.3566 +Query 1/1: Action query time = 4.803 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7560 +t=106: Selected seed 195 with value = 0.7560 +Query 1/1: Action query time = 4.675 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8303 +t=122: Selected seed 195 with value = 0.8303 +Query 1/1: Action query time = 3.317 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9632 +t=138: Selected seed 195 with value = 0.9632 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s14/2026_08_02-03_51_08--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 3.189 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1636 +t=10: Selected seed 195 with value = 0.1636 +Query 1/1: Action query time = 5.215 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2239 +t=26: Selected seed 195 with value = 0.2239 +Query 1/1: Action query time = 4.625 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2797 +t=42: Selected seed 195 with value = 0.2797 +Query 1/1: Action query time = 4.863 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3237 +t=58: Selected seed 195 with value = 0.3237 +Query 1/1: Action query time = 4.912 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5764 +t=74: Selected seed 195 with value = 0.5764 +Query 1/1: Action query time = 4.975 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7333 +t=90: Selected seed 195 with value = 0.7333 +Query 1/1: Action query time = 4.459 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5987 +t=106: Selected seed 195 with value = 0.5987 +Query 1/1: Action query time = 4.521 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6624 +t=122: Selected seed 195 with value = 0.6624 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s14/2026_08_02-03_51_08--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 3.541 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1831 +t=10: Selected seed 195 with value = 0.1831 +Query 1/1: Action query time = 3.809 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2000 +t=26: Selected seed 195 with value = 0.2000 +Query 1/1: Action query time = 2.878 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4165 +t=42: Selected seed 195 with value = 0.4165 +Query 1/1: Action query time = 4.958 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4360 +t=58: Selected seed 195 with value = 0.4360 +Query 1/1: Action query time = 5.058 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5262 +t=74: Selected seed 195 with value = 0.5262 +Query 1/1: Action query time = 5.399 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6198 +t=90: Selected seed 195 with value = 0.6198 +Query 1/1: Action query time = 5.208 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7362 +t=106: Selected seed 195 with value = 0.7362 +Query 1/1: Action query time = 3.537 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8745 +t=122: Selected seed 195 with value = 0.8745 +Saved rollout MP4 at path ./rollouts/realcl800_t5_s14/2026_08_02-03_51_08--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_49--realcl800_t6_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_49--realcl800_t6_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..47256c4bcf2aee7a2c61dad32563f711c5e7aca2 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_49--realcl800_t6_s03.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t6_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 3.973 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4163 +t=10: Selected seed 195 with value = 0.4163 +Query 1/1: Action query time = 4.879 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4912 +t=26: Selected seed 195 with value = 0.4912 +Query 1/1: Action query time = 4.508 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5842 +t=42: Selected seed 195 with value = 0.5842 +Query 1/1: Action query time = 4.526 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6933 +t=58: Selected seed 195 with value = 0.6933 +Query 1/1: Action query time = 4.455 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8291 +t=74: Selected seed 195 with value = 0.8291 +Query 1/1: Action query time = 4.799 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9732 +t=90: Selected seed 195 with value = 0.9732 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s03/2026_08_02-03_54_49--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 4.635 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4304 +t=10: Selected seed 195 with value = 0.4304 +Query 1/1: Action query time = 4.501 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5053 +t=26: Selected seed 195 with value = 0.5053 +Query 1/1: Action query time = 4.648 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6031 +t=42: Selected seed 195 with value = 0.6031 +Query 1/1: Action query time = 5.326 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6959 +t=58: Selected seed 195 with value = 0.6959 +Query 1/1: Action query time = 5.360 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8058 +t=74: Selected seed 195 with value = 0.8058 +Query 1/1: Action query time = 4.863 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9465 +t=90: Selected seed 195 with value = 0.9465 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s03/2026_08_02-03_54_49--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 5.344 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4231 +t=10: Selected seed 195 with value = 0.4231 +Query 1/1: Action query time = 5.212 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5161 +t=26: Selected seed 195 with value = 0.5161 +Query 1/1: Action query time = 4.719 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6290 +t=42: Selected seed 195 with value = 0.6290 +Query 1/1: Action query time = 4.861 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7351 +t=58: Selected seed 195 with value = 0.7351 +Query 1/1: Action query time = 4.426 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8653 +t=74: Selected seed 195 with value = 0.8653 +Query 1/1: Action query time = 3.116 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9877 +t=90: Selected seed 195 with value = 0.9877 +Query 1/1: Action query time = 2.831 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=106: Selected seed 195 with value = 0.9909 +Query 1/1: Action query time = 2.503 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9833 +t=122: Selected seed 195 with value = 0.9833 +Query 1/1: Action query time = 4.034 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9829 +t=138: Selected seed 195 with value = 0.9829 +Query 1/1: Action query time = 5.425 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9791 +t=154: Selected seed 195 with value = 0.9791 +Query 1/1: Action query time = 5.462 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9679 +t=170: Selected seed 195 with value = 0.9679 +Query 1/1: Action query time = 4.601 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9495 +t=186: Selected seed 195 with value = 0.9495 +Query 1/1: Action query time = 3.614 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9390 +t=202: Selected seed 195 with value = 0.9390 +Query 1/1: Action query time = 3.472 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9040 +t=218: Selected seed 195 with value = 0.9040 +Query 1/1: Action query time = 3.115 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9585 +t=234: Selected seed 195 with value = 0.9585 +Query 1/1: Action query time = 5.048 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9756 +t=250: Selected seed 195 with value = 0.9756 +Query 1/1: Action query time = 3.933 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9653 +t=266: Selected seed 195 with value = 0.9653 +Query 1/1: Action query time = 4.481 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9685 +t=282: Selected seed 195 with value = 0.9685 +Query 1/1: Action query time = 4.312 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9684 +t=298: Selected seed 195 with value = 0.9684 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s03/2026_08_02-03_54_49--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_50--realcl800_t6_s10.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_50--realcl800_t6_s10.txt new file mode 100644 index 0000000000000000000000000000000000000000..25490ed772a1d7b940cf22a505c5713be6ac0da7 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_50--realcl800_t6_s10.txt @@ -0,0 +1,101 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t6_s10', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='10,26,42', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 6.144 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4311 +t=10: Selected seed 195 with value = 0.4311 +Query 1/1: Action query time = 5.105 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4996 +t=26: Selected seed 195 with value = 0.4996 +Query 1/1: Action query time = 5.228 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6372 +t=42: Selected seed 195 with value = 0.6372 +Query 1/1: Action query time = 4.246 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7302 +t=58: Selected seed 195 with value = 0.7302 +Query 1/1: Action query time = 3.835 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8593 +t=74: Selected seed 195 with value = 0.8593 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s10/2026_08_02-03_54_50--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 3.159 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4509 +t=10: Selected seed 195 with value = 0.4509 +Query 1/1: Action query time = 3.798 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5323 +t=26: Selected seed 195 with value = 0.5323 +Query 1/1: Action query time = 5.446 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6271 +t=42: Selected seed 195 with value = 0.6271 +Query 1/1: Action query time = 4.464 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6894 +t=58: Selected seed 195 with value = 0.6894 +Query 1/1: Action query time = 4.946 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8185 +t=74: Selected seed 195 with value = 0.8185 +Query 1/1: Action query time = 4.940 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=90: Selected seed 195 with value = 0.9870 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s10/2026_08_02-03_54_50--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.212 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4227 +t=10: Selected seed 195 with value = 0.4227 +Query 1/1: Action query time = 4.410 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4941 +t=26: Selected seed 195 with value = 0.4941 +Query 1/1: Action query time = 5.025 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6198 +t=42: Selected seed 195 with value = 0.6198 +Query 1/1: Action query time = 4.929 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6970 +t=58: Selected seed 195 with value = 0.6970 +Query 1/1: Action query time = 4.784 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8302 +t=74: Selected seed 195 with value = 0.8302 +Query 1/1: Action query time = 4.771 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9847 +t=90: Selected seed 195 with value = 0.9847 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s10/2026_08_02-03_54_50--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_50--realcl800_t6_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_50--realcl800_t6_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..1202f96410f95dca34b43a407c4322c38063bf02 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_50--realcl800_t6_s11.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t6_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 2.825 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4690 +t=10: Selected seed 195 with value = 0.4690 +Query 1/1: Action query time = 4.750 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5138 +t=26: Selected seed 195 with value = 0.5138 +Query 1/1: Action query time = 4.688 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6336 +t=42: Selected seed 195 with value = 0.6336 +Query 1/1: Action query time = 4.569 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7310 +t=58: Selected seed 195 with value = 0.7310 +Query 1/1: Action query time = 4.922 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8400 +t=74: Selected seed 195 with value = 0.8400 +Query 1/1: Action query time = 4.782 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9402 +t=90: Selected seed 195 with value = 0.9402 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s11/2026_08_02-03_54_50--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 3.005 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4340 +t=10: Selected seed 195 with value = 0.4340 +Query 1/1: Action query time = 3.187 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5227 +t=26: Selected seed 195 with value = 0.5227 +Query 1/1: Action query time = 5.082 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5829 +t=42: Selected seed 195 with value = 0.5829 +Query 1/1: Action query time = 4.736 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4990 +t=58: Selected seed 195 with value = 0.4990 +Query 1/1: Action query time = 4.252 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7730 +t=74: Selected seed 195 with value = 0.7730 +Query 1/1: Action query time = 4.935 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9172 +t=90: Selected seed 195 with value = 0.9172 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s11/2026_08_02-03_54_50--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.011 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4355 +t=10: Selected seed 195 with value = 0.4355 +Query 1/1: Action query time = 3.808 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3877 +t=26: Selected seed 195 with value = 0.3877 +Query 1/1: Action query time = 4.755 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5783 +t=42: Selected seed 195 with value = 0.5783 +Query 1/1: Action query time = 5.114 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6944 +t=58: Selected seed 195 with value = 0.6944 +Query 1/1: Action query time = 5.170 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8261 +t=74: Selected seed 195 with value = 0.8261 +Query 1/1: Action query time = 5.122 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9584 +t=90: Selected seed 195 with value = 0.9584 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s11/2026_08_02-03_54_50--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_51--realcl800_t6_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_51--realcl800_t6_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..de51bee14a82dd340a6d8cfa80677f145972179e --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-03_54_51--realcl800_t6_s14.txt @@ -0,0 +1,153 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_real_cl_from40k_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='realcl800_t6_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 5.975 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4457 +t=10: Selected seed 195 with value = 0.4457 +Query 1/1: Action query time = 5.000 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5026 +t=26: Selected seed 195 with value = 0.5026 +Query 1/1: Action query time = 5.182 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6365 +t=42: Selected seed 195 with value = 0.6365 +Query 1/1: Action query time = 4.633 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6645 +t=58: Selected seed 195 with value = 0.6645 +Query 1/1: Action query time = 3.501 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5940 +t=74: Selected seed 195 with value = 0.5940 +Query 1/1: Action query time = 3.815 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6439 +t=90: Selected seed 195 with value = 0.6439 +Query 1/1: Action query time = 2.542 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7641 +t=106: Selected seed 195 with value = 0.7641 +Query 1/1: Action query time = 4.323 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8392 +t=122: Selected seed 195 with value = 0.8392 +Query 1/1: Action query time = 4.852 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9221 +t=138: Selected seed 195 with value = 0.9221 +Query 1/1: Action query time = 4.893 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9875 +t=154: Selected seed 195 with value = 0.9875 +Query 1/1: Action query time = 5.031 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9698 +t=170: Selected seed 195 with value = 0.9698 +Query 1/1: Action query time = 4.517 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9692 +t=186: Selected seed 195 with value = 0.9692 +Query 1/1: Action query time = 3.542 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9740 +t=202: Selected seed 195 with value = 0.9740 +Query 1/1: Action query time = 4.060 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9744 +t=218: Selected seed 195 with value = 0.9744 +Query 1/1: Action query time = 5.158 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9750 +t=234: Selected seed 195 with value = 0.9750 +Query 1/1: Action query time = 4.969 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9769 +t=250: Selected seed 195 with value = 0.9769 +Query 1/1: Action query time = 4.705 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9765 +t=266: Selected seed 195 with value = 0.9765 +Query 1/1: Action query time = 4.915 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9760 +t=282: Selected seed 195 with value = 0.9760 +Query 1/1: Action query time = 3.981 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9771 +t=298: Selected seed 195 with value = 0.9771 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s14/2026_08_02-03_54_51--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 1.222 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4251 +t=10: Selected seed 195 with value = 0.4251 +Query 1/1: Action query time = 1.269 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5140 +t=26: Selected seed 195 with value = 0.5140 +Query 1/1: Action query time = 1.304 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6249 +t=42: Selected seed 195 with value = 0.6249 +Query 1/1: Action query time = 1.516 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7053 +t=58: Selected seed 195 with value = 0.7053 +Query 1/1: Action query time = 1.384 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8356 +t=74: Selected seed 195 with value = 0.8356 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s14/2026_08_02-03_54_51--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 0.976 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4455 +t=10: Selected seed 195 with value = 0.4455 +Query 1/1: Action query time = 0.987 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5045 +t=26: Selected seed 195 with value = 0.5045 +Query 1/1: Action query time = 0.981 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6258 +t=42: Selected seed 195 with value = 0.6258 +Query 1/1: Action query time = 0.986 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7136 +t=58: Selected seed 195 with value = 0.7136 +Query 1/1: Action query time = 1.225 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8492 +t=74: Selected seed 195 with value = 0.8492 +Query 1/1: Action query time = 0.992 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=90: Selected seed 195 with value = 0.9886 +Saved rollout MP4 at path ./rollouts/realcl800_t6_s14/2026_08_02-03_54_51--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_44_19--fcv2_175_t0_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_44_19--fcv2_175_t0_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..eaa9981dfc2e179d8a663559caad09ee502ef96a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_44_19--fcv2_175_t0_s01.txt @@ -0,0 +1,168 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t0_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.363 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3693 +t=10: Selected seed 195 with value = 0.3693 +Query 1/1: Action query time = 3.178 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4172 +t=26: Selected seed 195 with value = 0.4172 +Query 1/1: Action query time = 4.364 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4861 +t=42: Selected seed 195 with value = 0.4861 +Query 1/1: Action query time = 5.136 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5676 +t=58: Selected seed 195 with value = 0.5676 +Query 1/1: Action query time = 4.679 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6740 +t=74: Selected seed 195 with value = 0.6740 +Query 1/1: Action query time = 5.182 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7641 +t=90: Selected seed 195 with value = 0.7641 +Query 1/1: Action query time = 4.964 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9152 +t=106: Selected seed 195 with value = 0.9152 +Query 1/1: Action query time = 4.683 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t0_s01/2026_08_02-04_44_19--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.547 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3633 +t=10: Selected seed 195 with value = 0.3633 +Query 1/1: Action query time = 3.605 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4253 +t=26: Selected seed 195 with value = 0.4253 +Query 1/1: Action query time = 6.153 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5169 +t=42: Selected seed 195 with value = 0.5169 +Query 1/1: Action query time = 4.516 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5237 +t=58: Selected seed 195 with value = 0.5237 +Query 1/1: Action query time = 4.784 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5487 +t=74: Selected seed 195 with value = 0.5487 +Query 1/1: Action query time = 5.199 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7256 +t=90: Selected seed 195 with value = 0.7256 +Query 1/1: Action query time = 5.231 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8874 +t=106: Selected seed 195 with value = 0.8874 +Query 1/1: Action query time = 4.862 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9902 +t=122: Selected seed 195 with value = 0.9902 +Saved rollout MP4 at path ./rollouts/fcv2_175_t0_s01/2026_08_02-04_44_19--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.838 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3640 +t=10: Selected seed 195 with value = 0.3640 +Query 1/1: Action query time = 3.019 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4261 +t=26: Selected seed 195 with value = 0.4261 +Query 1/1: Action query time = 5.172 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5169 +t=42: Selected seed 195 with value = 0.5169 +Query 1/1: Action query time = 5.229 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6056 +t=58: Selected seed 195 with value = 0.6056 +Query 1/1: Action query time = 4.811 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7218 +t=74: Selected seed 195 with value = 0.7218 +Query 1/1: Action query time = 5.132 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8279 +t=90: Selected seed 195 with value = 0.8279 +Query 1/1: Action query time = 4.985 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9835 +t=106: Selected seed 195 with value = 0.9835 +Query 1/1: Action query time = 5.179 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9910 +t=122: Selected seed 195 with value = 0.9910 +Saved rollout MP4 at path ./rollouts/fcv2_175_t0_s01/2026_08_02-04_44_19--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 4... +Query 1/1: Action query time = 4.182 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3697 +t=10: Selected seed 195 with value = 0.3697 +Query 1/1: Action query time = 1.964 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4224 +t=26: Selected seed 195 with value = 0.4224 +Query 1/1: Action query time = 1.488 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5002 +t=42: Selected seed 195 with value = 0.5002 +Query 1/1: Action query time = 2.017 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6167 +t=58: Selected seed 195 with value = 0.6167 +Query 1/1: Action query time = 2.057 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6957 +t=74: Selected seed 195 with value = 0.6957 +Query 1/1: Action query time = 2.073 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8380 +t=90: Selected seed 195 with value = 0.8380 +Query 1/1: Action query time = 2.032 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9171 +t=106: Selected seed 195 with value = 0.9171 +Query 1/1: Action query time = 1.887 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t0_s01/2026_08_02-04_44_19--with_future_img--episode=4--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 4 +Total successes: 4 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_44_19--fcv2_175_t0_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_44_19--fcv2_175_t0_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..ed60b589c7614743933cc222e89ce882d2054a77 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_44_19--fcv2_175_t0_s07.txt @@ -0,0 +1,129 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t0_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.428 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3534 +t=10: Selected seed 195 with value = 0.3534 +Query 1/1: Action query time = 5.931 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4014 +t=26: Selected seed 195 with value = 0.4014 +Query 1/1: Action query time = 4.739 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4647 +t=42: Selected seed 195 with value = 0.4647 +Query 1/1: Action query time = 5.166 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5555 +t=58: Selected seed 195 with value = 0.5555 +Query 1/1: Action query time = 5.202 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6477 +t=74: Selected seed 195 with value = 0.6477 +Query 1/1: Action query time = 4.587 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7453 +t=90: Selected seed 195 with value = 0.7453 +Query 1/1: Action query time = 4.817 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9025 +t=106: Selected seed 195 with value = 0.9025 +Query 1/1: Action query time = 3.093 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t0_s07/2026_08_02-04_44_19--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.482 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3331 +t=10: Selected seed 195 with value = 0.3331 +Query 1/1: Action query time = 4.088 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3583 +t=26: Selected seed 195 with value = 0.3583 +Query 1/1: Action query time = 4.273 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4310 +t=42: Selected seed 195 with value = 0.4310 +Query 1/1: Action query time = 5.499 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5246 +t=58: Selected seed 195 with value = 0.5246 +Query 1/1: Action query time = 4.886 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6153 +t=74: Selected seed 195 with value = 0.6153 +Query 1/1: Action query time = 4.555 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7341 +t=90: Selected seed 195 with value = 0.7341 +Query 1/1: Action query time = 4.960 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8933 +t=106: Selected seed 195 with value = 0.8933 +Query 1/1: Action query time = 3.786 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=122: Selected seed 195 with value = 0.9982 +Saved rollout MP4 at path ./rollouts/fcv2_175_t0_s07/2026_08_02-04_44_19--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 2.416 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3717 +t=10: Selected seed 195 with value = 0.3717 +Query 1/1: Action query time = 4.582 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4222 +t=26: Selected seed 195 with value = 0.4222 +Query 1/1: Action query time = 4.802 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4802 +t=42: Selected seed 195 with value = 0.4802 +Query 1/1: Action query time = 5.573 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5672 +t=58: Selected seed 195 with value = 0.5672 +Query 1/1: Action query time = 5.379 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6573 +t=74: Selected seed 195 with value = 0.6573 +Query 1/1: Action query time = 4.978 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7985 +t=90: Selected seed 195 with value = 0.7985 +Query 1/1: Action query time = 5.129 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9241 +t=106: Selected seed 195 with value = 0.9241 +Query 1/1: Action query time = 4.316 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9959 +t=122: Selected seed 195 with value = 0.9959 +Saved rollout MP4 at path ./rollouts/fcv2_175_t0_s07/2026_08_02-04_44_19--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_40--fcv2_175_t1_s05.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_40--fcv2_175_t1_s05.txt new file mode 100644 index 0000000000000000000000000000000000000000..99e2c0f9848f00f93361f513e0f0ce65f6000e5d --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_40--fcv2_175_t1_s05.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t1_s05', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,21,37', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.914 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5011 +t=10: Selected seed 195 with value = 0.5011 +Query 1/1: Action query time = 4.675 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5829 +t=26: Selected seed 195 with value = 0.5829 +Query 1/1: Action query time = 4.913 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7236 +t=42: Selected seed 195 with value = 0.7236 +Query 1/1: Action query time = 4.570 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8021 +t=58: Selected seed 195 with value = 0.8021 +Query 1/1: Action query time = 3.250 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9474 +t=74: Selected seed 195 with value = 0.9474 +Query 1/1: Action query time = 4.474 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=90: Selected seed 195 with value = 0.9990 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s05/2026_08_02-04_47_40--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 4.192 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4496 +t=10: Selected seed 195 with value = 0.4496 +Query 1/1: Action query time = 5.326 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5550 +t=26: Selected seed 195 with value = 0.5550 +Query 1/1: Action query time = 5.578 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6556 +t=42: Selected seed 195 with value = 0.6556 +Query 1/1: Action query time = 5.001 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7481 +t=58: Selected seed 195 with value = 0.7481 +Query 1/1: Action query time = 4.047 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9181 +t=74: Selected seed 195 with value = 0.9181 +Query 1/1: Action query time = 5.102 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=90: Selected seed 195 with value = 0.9954 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s05/2026_08_02-04_47_40--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 3.898 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5175 +t=10: Selected seed 195 with value = 0.5175 +Query 1/1: Action query time = 5.227 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6120 +t=26: Selected seed 195 with value = 0.6120 +Query 1/1: Action query time = 5.159 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7046 +t=42: Selected seed 195 with value = 0.7046 +Query 1/1: Action query time = 4.790 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8212 +t=58: Selected seed 195 with value = 0.8212 +Query 1/1: Action query time = 4.586 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9410 +t=74: Selected seed 195 with value = 0.9410 +Query 1/1: Action query time = 3.810 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=90: Selected seed 195 with value = 0.9986 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s05/2026_08_02-04_47_40--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_41--fcv2_175_t1_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_41--fcv2_175_t1_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..5eb58fb5c8fbf1993919a695149aa21191712e91 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_41--fcv2_175_t1_s11.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t1_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.246 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5153 +t=10: Selected seed 195 with value = 0.5153 +Query 1/1: Action query time = 4.689 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5960 +t=26: Selected seed 195 with value = 0.5960 +Query 1/1: Action query time = 4.410 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7352 +t=42: Selected seed 195 with value = 0.7352 +Query 1/1: Action query time = 4.610 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8152 +t=58: Selected seed 195 with value = 0.8152 +Query 1/1: Action query time = 5.011 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9266 +t=74: Selected seed 195 with value = 0.9266 +Query 1/1: Action query time = 4.357 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9948 +t=90: Selected seed 195 with value = 0.9948 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s11/2026_08_02-04_47_41--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 2.864 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4857 +t=10: Selected seed 195 with value = 0.4857 +Query 1/1: Action query time = 4.769 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6168 +t=26: Selected seed 195 with value = 0.6168 +Query 1/1: Action query time = 4.695 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7292 +t=42: Selected seed 195 with value = 0.7292 +Query 1/1: Action query time = 4.851 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8231 +t=58: Selected seed 195 with value = 0.8231 +Query 1/1: Action query time = 4.693 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9622 +t=74: Selected seed 195 with value = 0.9622 +Query 1/1: Action query time = 4.835 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s11/2026_08_02-04_47_41--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 2.855 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5113 +t=10: Selected seed 195 with value = 0.5113 +Query 1/1: Action query time = 4.882 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5990 +t=26: Selected seed 195 with value = 0.5990 +Query 1/1: Action query time = 4.768 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7203 +t=42: Selected seed 195 with value = 0.7203 +Query 1/1: Action query time = 4.519 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7725 +t=58: Selected seed 195 with value = 0.7725 +Query 1/1: Action query time = 4.154 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9499 +t=74: Selected seed 195 with value = 0.9499 +Query 1/1: Action query time = 4.278 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=90: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s11/2026_08_02-04_47_41--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_41--fcv2_175_t1_s12.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_41--fcv2_175_t1_s12.txt new file mode 100644 index 0000000000000000000000000000000000000000..a6a68e3e3f8e3728d2418e50614467e0c5781c15 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_41--fcv2_175_t1_s12.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t1_s12', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='12,28,44', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.668 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5200 +t=10: Selected seed 195 with value = 0.5200 +Query 1/1: Action query time = 5.176 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5835 +t=26: Selected seed 195 with value = 0.5835 +Query 1/1: Action query time = 5.260 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7478 +t=42: Selected seed 195 with value = 0.7478 +Query 1/1: Action query time = 5.111 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8264 +t=58: Selected seed 195 with value = 0.8264 +Query 1/1: Action query time = 4.894 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9655 +t=74: Selected seed 195 with value = 0.9655 +Query 1/1: Action query time = 4.436 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9901 +t=90: Selected seed 195 with value = 0.9901 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s12/2026_08_02-04_47_41--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 4.651 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4977 +t=10: Selected seed 195 with value = 0.4977 +Query 1/1: Action query time = 5.125 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6027 +t=26: Selected seed 195 with value = 0.6027 +Query 1/1: Action query time = 5.284 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7299 +t=42: Selected seed 195 with value = 0.7299 +Query 1/1: Action query time = 4.609 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8070 +t=58: Selected seed 195 with value = 0.8070 +Query 1/1: Action query time = 3.894 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9469 +t=74: Selected seed 195 with value = 0.9469 +Query 1/1: Action query time = 4.632 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=90: Selected seed 195 with value = 0.9988 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s12/2026_08_02-04_47_41--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.566 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5113 +t=10: Selected seed 195 with value = 0.5113 +Query 1/1: Action query time = 4.832 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5919 +t=26: Selected seed 195 with value = 0.5919 +Query 1/1: Action query time = 4.233 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7128 +t=42: Selected seed 195 with value = 0.7128 +Query 1/1: Action query time = 4.159 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7746 +t=58: Selected seed 195 with value = 0.7746 +Query 1/1: Action query time = 3.594 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8934 +t=74: Selected seed 195 with value = 0.8934 +Query 1/1: Action query time = 3.198 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9757 +t=90: Selected seed 195 with value = 0.9757 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s12/2026_08_02-04_47_41--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_41--fcv2_175_t1_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_41--fcv2_175_t1_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..8aa2518eeee6bf8dd31245a568e4936ab46555f9 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_47_41--fcv2_175_t1_s14.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t1_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.817 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5230 +t=10: Selected seed 195 with value = 0.5230 +Query 1/1: Action query time = 4.876 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6261 +t=26: Selected seed 195 with value = 0.6261 +Query 1/1: Action query time = 4.849 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7440 +t=42: Selected seed 195 with value = 0.7440 +Query 1/1: Action query time = 4.838 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8743 +t=58: Selected seed 195 with value = 0.8743 +Query 1/1: Action query time = 4.336 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9953 +t=74: Selected seed 195 with value = 0.9953 +Query 1/1: Action query time = 4.338 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=90: Selected seed 195 with value = 0.9982 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s14/2026_08_02-04_47_41--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 4.844 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4862 +t=10: Selected seed 195 with value = 0.4862 +Query 1/1: Action query time = 5.047 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5773 +t=26: Selected seed 195 with value = 0.5773 +Query 1/1: Action query time = 5.326 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7021 +t=42: Selected seed 195 with value = 0.7021 +Query 1/1: Action query time = 4.938 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8175 +t=58: Selected seed 195 with value = 0.8175 +Query 1/1: Action query time = 4.547 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9270 +t=74: Selected seed 195 with value = 0.9270 +Query 1/1: Action query time = 4.004 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=90: Selected seed 195 with value = 0.9999 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s14/2026_08_02-04_47_41--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 4.833 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4783 +t=10: Selected seed 195 with value = 0.4783 +Query 1/1: Action query time = 5.130 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5771 +t=26: Selected seed 195 with value = 0.5771 +Query 1/1: Action query time = 5.182 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6931 +t=42: Selected seed 195 with value = 0.6931 +Query 1/1: Action query time = 4.188 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8084 +t=58: Selected seed 195 with value = 0.8084 +Query 1/1: Action query time = 3.047 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9260 +t=74: Selected seed 195 with value = 0.9260 +Query 1/1: Action query time = 2.749 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=90: Selected seed 195 with value = 0.9981 +Saved rollout MP4 at path ./rollouts/fcv2_175_t1_s14/2026_08_02-04_47_41--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_26--fcv2_175_t2_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_26--fcv2_175_t2_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..38751136f5fc340b477278a67707d462840814a5 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_26--fcv2_175_t2_s03.txt @@ -0,0 +1,161 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t2_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.312 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4941 +t=10: Selected seed 195 with value = 0.4941 +Query 1/1: Action query time = 3.024 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5548 +t=26: Selected seed 195 with value = 0.5548 +Query 1/1: Action query time = 4.183 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7207 +t=42: Selected seed 195 with value = 0.7207 +Query 1/1: Action query time = 3.500 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8210 +t=58: Selected seed 195 with value = 0.8210 +Query 1/1: Action query time = 5.917 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9842 +t=74: Selected seed 195 with value = 0.9842 +Query 1/1: Action query time = 5.277 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s03/2026_08_02-04_50_26--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.827 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4997 +t=10: Selected seed 195 with value = 0.4997 +Query 1/1: Action query time = 3.865 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5928 +t=26: Selected seed 195 with value = 0.5928 +Query 1/1: Action query time = 4.966 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7086 +t=42: Selected seed 195 with value = 0.7086 +Query 1/1: Action query time = 5.030 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7914 +t=58: Selected seed 195 with value = 0.7914 +Query 1/1: Action query time = 5.267 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9431 +t=74: Selected seed 195 with value = 0.9431 +Query 1/1: Action query time = 5.231 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9741 +t=90: Selected seed 195 with value = 0.9741 +Query 1/1: Action query time = 4.104 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9880 +t=106: Selected seed 195 with value = 0.9880 +Query 1/1: Action query time = 3.917 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.868 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.262 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.529 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.070 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.753 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.998 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.505 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9818 +t=234: Selected seed 195 with value = 0.9818 +Query 1/1: Action query time = 2.410 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9089 +t=250: Selected seed 195 with value = 0.9089 +Query 1/1: Action query time = 2.454 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8151 +t=266: Selected seed 195 with value = 0.8151 +Query 1/1: Action query time = 2.463 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8033 +t=282: Selected seed 195 with value = 0.8033 +Query 1/1: Action query time = 2.651 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6893 +t=298: Selected seed 195 with value = 0.6893 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s03/2026_08_02-04_50_26--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 1.368 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4769 +t=10: Selected seed 195 with value = 0.4769 +Query 1/1: Action query time = 1.367 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5732 +t=26: Selected seed 195 with value = 0.5732 +Query 1/1: Action query time = 2.218 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6617 +t=42: Selected seed 195 with value = 0.6617 +Query 1/1: Action query time = 1.999 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7578 +t=58: Selected seed 195 with value = 0.7578 +Query 1/1: Action query time = 2.038 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8955 +t=74: Selected seed 195 with value = 0.8955 +Query 1/1: Action query time = 1.935 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9936 +t=90: Selected seed 195 with value = 0.9936 +Query 1/1: Action query time = 1.899 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9898 +t=106: Selected seed 195 with value = 0.9898 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s03/2026_08_02-04_50_26--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_27--fcv2_175_t2_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_27--fcv2_175_t2_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..18f035a2c57d96d7ec98ed47b8718c7dff60d883 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_27--fcv2_175_t2_s08.txt @@ -0,0 +1,109 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t2_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.477 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5010 +t=10: Selected seed 195 with value = 0.5010 +Query 1/1: Action query time = 5.010 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6009 +t=26: Selected seed 195 with value = 0.6009 +Query 1/1: Action query time = 6.131 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7291 +t=42: Selected seed 195 with value = 0.7291 +Query 1/1: Action query time = 5.291 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8362 +t=58: Selected seed 195 with value = 0.8362 +Query 1/1: Action query time = 4.011 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9758 +t=74: Selected seed 195 with value = 0.9758 +Query 1/1: Action query time = 4.058 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s08/2026_08_02-04_50_27--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.264 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3904 +t=10: Selected seed 195 with value = 0.3904 +Query 1/1: Action query time = 5.017 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5761 +t=26: Selected seed 195 with value = 0.5761 +Query 1/1: Action query time = 5.035 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6462 +t=42: Selected seed 195 with value = 0.6462 +Query 1/1: Action query time = 5.443 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7409 +t=58: Selected seed 195 with value = 0.7409 +Query 1/1: Action query time = 4.517 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8702 +t=74: Selected seed 195 with value = 0.8702 +Query 1/1: Action query time = 5.039 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=90: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s08/2026_08_02-04_50_27--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 2.673 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4647 +t=10: Selected seed 195 with value = 0.4647 +Query 1/1: Action query time = 5.046 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5042 +t=26: Selected seed 195 with value = 0.5042 +Query 1/1: Action query time = 4.953 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5933 +t=42: Selected seed 195 with value = 0.5933 +Query 1/1: Action query time = 4.680 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7060 +t=58: Selected seed 195 with value = 0.7060 +Query 1/1: Action query time = 4.363 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8385 +t=74: Selected seed 195 with value = 0.8385 +Query 1/1: Action query time = 4.277 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9763 +t=90: Selected seed 195 with value = 0.9763 +Query 1/1: Action query time = 1.870 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s08/2026_08_02-04_50_27--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_28--fcv2_175_t2_s09.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_28--fcv2_175_t2_s09.txt new file mode 100644 index 0000000000000000000000000000000000000000..88cdc82e7c744c33cf3c30a1c279e85ffba16883 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_28--fcv2_175_t2_s09.txt @@ -0,0 +1,101 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t2_s09', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='9,25,41', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.672 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5116 +t=10: Selected seed 195 with value = 0.5116 +Query 1/1: Action query time = 5.076 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5887 +t=26: Selected seed 195 with value = 0.5887 +Query 1/1: Action query time = 3.911 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7246 +t=42: Selected seed 195 with value = 0.7246 +Query 1/1: Action query time = 4.578 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8383 +t=58: Selected seed 195 with value = 0.8383 +Query 1/1: Action query time = 5.116 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9798 +t=74: Selected seed 195 with value = 0.9798 +Query 1/1: Action query time = 3.805 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s09/2026_08_02-04_50_28--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.158 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5160 +t=10: Selected seed 195 with value = 0.5160 +Query 1/1: Action query time = 5.009 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5797 +t=26: Selected seed 195 with value = 0.5797 +Query 1/1: Action query time = 5.502 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7030 +t=42: Selected seed 195 with value = 0.7030 +Query 1/1: Action query time = 4.525 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8312 +t=58: Selected seed 195 with value = 0.8312 +Query 1/1: Action query time = 5.037 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9624 +t=74: Selected seed 195 with value = 0.9624 +Query 1/1: Action query time = 4.381 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s09/2026_08_02-04_50_28--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.932 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4820 +t=10: Selected seed 195 with value = 0.4820 +Query 1/1: Action query time = 5.003 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6082 +t=26: Selected seed 195 with value = 0.6082 +Query 1/1: Action query time = 4.959 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7143 +t=42: Selected seed 195 with value = 0.7143 +Query 1/1: Action query time = 4.261 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8522 +t=58: Selected seed 195 with value = 0.8522 +Query 1/1: Action query time = 4.010 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9389 +t=74: Selected seed 195 with value = 0.9389 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s09/2026_08_02-04_50_28--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_28--fcv2_175_t2_s13.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_28--fcv2_175_t2_s13.txt new file mode 100644 index 0000000000000000000000000000000000000000..335e3776796b0683c850598bf73431e996f7f665 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_50_28--fcv2_175_t2_s13.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t2_s13', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='13,29,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.378 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5001 +t=10: Selected seed 195 with value = 0.5001 +Query 1/1: Action query time = 4.775 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5755 +t=26: Selected seed 195 with value = 0.5755 +Query 1/1: Action query time = 5.498 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6697 +t=42: Selected seed 195 with value = 0.6697 +Query 1/1: Action query time = 3.931 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7741 +t=58: Selected seed 195 with value = 0.7741 +Query 1/1: Action query time = 3.208 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9090 +t=74: Selected seed 195 with value = 0.9090 +Query 1/1: Action query time = 4.956 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s13/2026_08_02-04_50_28--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.340 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5120 +t=10: Selected seed 195 with value = 0.5120 +Query 1/1: Action query time = 4.659 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6072 +t=26: Selected seed 195 with value = 0.6072 +Query 1/1: Action query time = 5.446 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7169 +t=42: Selected seed 195 with value = 0.7169 +Query 1/1: Action query time = 4.811 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8586 +t=58: Selected seed 195 with value = 0.8586 +Query 1/1: Action query time = 4.604 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.886 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s13/2026_08_02-04_50_28--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.861 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5291 +t=10: Selected seed 195 with value = 0.5291 +Query 1/1: Action query time = 3.174 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6007 +t=26: Selected seed 195 with value = 0.6007 +Query 1/1: Action query time = 5.028 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6978 +t=42: Selected seed 195 with value = 0.6978 +Query 1/1: Action query time = 5.199 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8078 +t=58: Selected seed 195 with value = 0.8078 +Query 1/1: Action query time = 4.370 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9764 +t=74: Selected seed 195 with value = 0.9764 +Query 1/1: Action query time = 4.692 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t2_s13/2026_08_02-04_50_28--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_29--fcv2_175_t3_s05.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_29--fcv2_175_t3_s05.txt new file mode 100644 index 0000000000000000000000000000000000000000..c10a38bbfa0214c830ca4014e9da4f29efa8aed5 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_29--fcv2_175_t3_s05.txt @@ -0,0 +1,217 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t3_s05', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,21,37', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 2.902 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2937 +t=10: Selected seed 195 with value = 0.2937 +Query 1/1: Action query time = 3.152 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4421 +t=26: Selected seed 195 with value = 0.4421 +Query 1/1: Action query time = 5.120 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4938 +t=42: Selected seed 195 with value = 0.4938 +Query 1/1: Action query time = 5.088 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4982 +t=58: Selected seed 195 with value = 0.4982 +Query 1/1: Action query time = 4.594 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5072 +t=74: Selected seed 195 with value = 0.5072 +Query 1/1: Action query time = 4.782 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6301 +t=90: Selected seed 195 with value = 0.6301 +Query 1/1: Action query time = 5.003 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6825 +t=106: Selected seed 195 with value = 0.6825 +Query 1/1: Action query time = 5.170 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7981 +t=122: Selected seed 195 with value = 0.7981 +Query 1/1: Action query time = 5.567 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9571 +t=138: Selected seed 195 with value = 0.9571 +Query 1/1: Action query time = 5.117 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.777 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8659 +t=170: Selected seed 195 with value = 0.8659 +Query 1/1: Action query time = 5.023 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6838 +t=186: Selected seed 195 with value = 0.6838 +Query 1/1: Action query time = 3.660 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7943 +t=202: Selected seed 195 with value = 0.7943 +Query 1/1: Action query time = 3.125 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9381 +t=218: Selected seed 195 with value = 0.9381 +Query 1/1: Action query time = 3.843 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=234: Selected seed 195 with value = 0.9945 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s05/2026_08_02-04_53_29--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.490 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3587 +t=10: Selected seed 195 with value = 0.3587 +Query 1/1: Action query time = 5.296 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4656 +t=26: Selected seed 195 with value = 0.4656 +Query 1/1: Action query time = 4.584 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5621 +t=42: Selected seed 195 with value = 0.5621 +Query 1/1: Action query time = 4.674 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6417 +t=58: Selected seed 195 with value = 0.6417 +Query 1/1: Action query time = 4.010 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6893 +t=74: Selected seed 195 with value = 0.6893 +Query 1/1: Action query time = 4.797 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8460 +t=90: Selected seed 195 with value = 0.8460 +Query 1/1: Action query time = 5.103 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9779 +t=106: Selected seed 195 with value = 0.9779 +Query 1/1: Action query time = 5.050 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6443 +t=122: Selected seed 195 with value = 0.6443 +Query 1/1: Action query time = 4.273 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7123 +t=138: Selected seed 195 with value = 0.7123 +Query 1/1: Action query time = 4.615 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8031 +t=154: Selected seed 195 with value = 0.8031 +Query 1/1: Action query time = 4.307 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9007 +t=170: Selected seed 195 with value = 0.9007 +Query 1/1: Action query time = 3.965 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9498 +t=186: Selected seed 195 with value = 0.9498 +Query 1/1: Action query time = 4.299 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9492 +t=202: Selected seed 195 with value = 0.9492 +Query 1/1: Action query time = 4.752 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6793 +t=218: Selected seed 195 with value = 0.6793 +Query 1/1: Action query time = 4.504 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7460 +t=234: Selected seed 195 with value = 0.7460 +Query 1/1: Action query time = 4.823 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8204 +t=250: Selected seed 195 with value = 0.8204 +Query 1/1: Action query time = 5.405 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8702 +t=266: Selected seed 195 with value = 0.8702 +Query 1/1: Action query time = 4.178 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8913 +t=282: Selected seed 195 with value = 0.8913 +Query 1/1: Action query time = 4.321 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8829 +t=298: Selected seed 195 with value = 0.8829 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s05/2026_08_02-04_53_29--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.823 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3262 +t=10: Selected seed 195 with value = 0.3262 +Query 1/1: Action query time = 4.587 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4953 +t=26: Selected seed 195 with value = 0.4953 +Query 1/1: Action query time = 4.943 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5643 +t=42: Selected seed 195 with value = 0.5643 +Query 1/1: Action query time = 4.918 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5228 +t=58: Selected seed 195 with value = 0.5228 +Query 1/1: Action query time = 4.823 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6772 +t=74: Selected seed 195 with value = 0.6772 +Query 1/1: Action query time = 3.599 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7051 +t=90: Selected seed 195 with value = 0.7051 +Query 1/1: Action query time = 3.790 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8627 +t=106: Selected seed 195 with value = 0.8627 +Query 1/1: Action query time = 4.148 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9554 +t=122: Selected seed 195 with value = 0.9554 +Query 1/1: Action query time = 4.118 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6205 +t=138: Selected seed 195 with value = 0.6205 +Query 1/1: Action query time = 4.385 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7199 +t=154: Selected seed 195 with value = 0.7199 +Query 1/1: Action query time = 3.061 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8575 +t=170: Selected seed 195 with value = 0.8575 +Query 1/1: Action query time = 3.222 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9834 +t=186: Selected seed 195 with value = 0.9834 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s05/2026_08_02-04_53_29--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_30--fcv2_175_t3_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_30--fcv2_175_t3_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..40b4daccd7c4f8c1a873daa1c69c957e46e44025 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_30--fcv2_175_t3_s06.txt @@ -0,0 +1,205 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t3_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 4.742 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4011 +t=10: Selected seed 195 with value = 0.4011 +Query 1/1: Action query time = 4.462 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4697 +t=26: Selected seed 195 with value = 0.4697 +Query 1/1: Action query time = 4.799 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5285 +t=42: Selected seed 195 with value = 0.5285 +Query 1/1: Action query time = 5.126 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6311 +t=58: Selected seed 195 with value = 0.6311 +Query 1/1: Action query time = 5.373 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6731 +t=74: Selected seed 195 with value = 0.6731 +Query 1/1: Action query time = 5.123 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8829 +t=90: Selected seed 195 with value = 0.8829 +Query 1/1: Action query time = 4.097 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9623 +t=106: Selected seed 195 with value = 0.9623 +Query 1/1: Action query time = 4.885 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6549 +t=122: Selected seed 195 with value = 0.6549 +Query 1/1: Action query time = 5.218 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7524 +t=138: Selected seed 195 with value = 0.7524 +Query 1/1: Action query time = 5.078 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9107 +t=154: Selected seed 195 with value = 0.9107 +Query 1/1: Action query time = 4.664 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9888 +t=170: Selected seed 195 with value = 0.9888 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s06/2026_08_02-04_53_30--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 3.741 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3675 +t=10: Selected seed 195 with value = 0.3675 +Query 1/1: Action query time = 4.069 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4750 +t=26: Selected seed 195 with value = 0.4750 +Query 1/1: Action query time = 3.960 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5686 +t=42: Selected seed 195 with value = 0.5686 +Query 1/1: Action query time = 4.766 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6528 +t=58: Selected seed 195 with value = 0.6528 +Query 1/1: Action query time = 5.111 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7412 +t=74: Selected seed 195 with value = 0.7412 +Query 1/1: Action query time = 5.293 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8598 +t=90: Selected seed 195 with value = 0.8598 +Query 1/1: Action query time = 5.015 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=106: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 4.249 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8662 +t=122: Selected seed 195 with value = 0.8662 +Query 1/1: Action query time = 5.255 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7673 +t=138: Selected seed 195 with value = 0.7673 +Query 1/1: Action query time = 4.785 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7207 +t=154: Selected seed 195 with value = 0.7207 +Query 1/1: Action query time = 4.971 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8808 +t=170: Selected seed 195 with value = 0.8808 +Query 1/1: Action query time = 4.588 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8809 +t=186: Selected seed 195 with value = 0.8809 +Query 1/1: Action query time = 5.427 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9818 +t=202: Selected seed 195 with value = 0.9818 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s06/2026_08_02-04_53_30--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 3.999 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3713 +t=10: Selected seed 195 with value = 0.3713 +Query 1/1: Action query time = 5.324 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4524 +t=26: Selected seed 195 with value = 0.4524 +Query 1/1: Action query time = 4.429 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5691 +t=42: Selected seed 195 with value = 0.5691 +Query 1/1: Action query time = 4.790 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6537 +t=58: Selected seed 195 with value = 0.6537 +Query 1/1: Action query time = 5.236 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7093 +t=74: Selected seed 195 with value = 0.7093 +Query 1/1: Action query time = 4.086 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8504 +t=90: Selected seed 195 with value = 0.8504 +Query 1/1: Action query time = 4.196 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9797 +t=106: Selected seed 195 with value = 0.9797 +Query 1/1: Action query time = 4.534 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9944 +t=122: Selected seed 195 with value = 0.9944 +Query 1/1: Action query time = 3.280 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6574 +t=138: Selected seed 195 with value = 0.6574 +Query 1/1: Action query time = 5.196 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7477 +t=154: Selected seed 195 with value = 0.7477 +Query 1/1: Action query time = 4.464 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7723 +t=170: Selected seed 195 with value = 0.7723 +Query 1/1: Action query time = 5.399 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9381 +t=186: Selected seed 195 with value = 0.9381 +Query 1/1: Action query time = 4.987 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9335 +t=202: Selected seed 195 with value = 0.9335 +Query 1/1: Action query time = 4.959 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9510 +t=218: Selected seed 195 with value = 0.9510 +Query 1/1: Action query time = 3.398 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9891 +t=234: Selected seed 195 with value = 0.9891 +Query 1/1: Action query time = 4.393 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=250: Selected seed 195 with value = 0.9962 +Query 1/1: Action query time = 4.161 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.896 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.363 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s06/2026_08_02-04_53_30--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_30--fcv2_175_t3_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_30--fcv2_175_t3_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..b36d3d3f9f138662e4e82608ea966dff1e653fac --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_30--fcv2_175_t3_s07.txt @@ -0,0 +1,241 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t3_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 5.753 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3679 +t=10: Selected seed 195 with value = 0.3679 +Query 1/1: Action query time = 5.320 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4703 +t=26: Selected seed 195 with value = 0.4703 +Query 1/1: Action query time = 5.180 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5775 +t=42: Selected seed 195 with value = 0.5775 +Query 1/1: Action query time = 4.909 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6636 +t=58: Selected seed 195 with value = 0.6636 +Query 1/1: Action query time = 4.721 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7537 +t=74: Selected seed 195 with value = 0.7537 +Query 1/1: Action query time = 5.243 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8756 +t=90: Selected seed 195 with value = 0.8756 +Query 1/1: Action query time = 4.957 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9913 +t=106: Selected seed 195 with value = 0.9913 +Query 1/1: Action query time = 4.640 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6391 +t=122: Selected seed 195 with value = 0.6391 +Query 1/1: Action query time = 4.833 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6986 +t=138: Selected seed 195 with value = 0.6986 +Query 1/1: Action query time = 4.778 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8484 +t=154: Selected seed 195 with value = 0.8484 +Query 1/1: Action query time = 3.743 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9619 +t=170: Selected seed 195 with value = 0.9619 +Query 1/1: Action query time = 3.914 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9392 +t=186: Selected seed 195 with value = 0.9392 +Query 1/1: Action query time = 3.827 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9385 +t=202: Selected seed 195 with value = 0.9385 +Query 1/1: Action query time = 4.265 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8056 +t=218: Selected seed 195 with value = 0.8056 +Query 1/1: Action query time = 3.675 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8888 +t=234: Selected seed 195 with value = 0.8888 +Query 1/1: Action query time = 4.605 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9357 +t=250: Selected seed 195 with value = 0.9357 +Query 1/1: Action query time = 5.181 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9514 +t=266: Selected seed 195 with value = 0.9514 +Query 1/1: Action query time = 4.575 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9637 +t=282: Selected seed 195 with value = 0.9637 +Query 1/1: Action query time = 4.500 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9739 +t=298: Selected seed 195 with value = 0.9739 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s07/2026_08_02-04_53_30--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 5.211 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3928 +t=10: Selected seed 195 with value = 0.3928 +Query 1/1: Action query time = 4.746 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4916 +t=26: Selected seed 195 with value = 0.4916 +Query 1/1: Action query time = 4.947 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5653 +t=42: Selected seed 195 with value = 0.5653 +Query 1/1: Action query time = 4.641 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6472 +t=58: Selected seed 195 with value = 0.6472 +Query 1/1: Action query time = 5.387 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7127 +t=74: Selected seed 195 with value = 0.7127 +Query 1/1: Action query time = 3.232 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8312 +t=90: Selected seed 195 with value = 0.8312 +Query 1/1: Action query time = 3.659 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9797 +t=106: Selected seed 195 with value = 0.9797 +Query 1/1: Action query time = 4.959 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8015 +t=122: Selected seed 195 with value = 0.8015 +Query 1/1: Action query time = 4.482 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5314 +t=138: Selected seed 195 with value = 0.5314 +Query 1/1: Action query time = 4.963 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6252 +t=154: Selected seed 195 with value = 0.6252 +Query 1/1: Action query time = 4.699 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7467 +t=170: Selected seed 195 with value = 0.7467 +Query 1/1: Action query time = 4.641 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8351 +t=186: Selected seed 195 with value = 0.8351 +Query 1/1: Action query time = 4.770 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9649 +t=202: Selected seed 195 with value = 0.9649 +Query 1/1: Action query time = 5.278 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=218: Selected seed 195 with value = 0.9989 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s07/2026_08_02-04_53_30--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.850 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4279 +t=10: Selected seed 195 with value = 0.4279 +Query 1/1: Action query time = 4.061 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4984 +t=26: Selected seed 195 with value = 0.4984 +Query 1/1: Action query time = 5.751 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5550 +t=42: Selected seed 195 with value = 0.5550 +Query 1/1: Action query time = 5.355 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6472 +t=58: Selected seed 195 with value = 0.6472 +Query 1/1: Action query time = 4.599 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7791 +t=74: Selected seed 195 with value = 0.7791 +Query 1/1: Action query time = 3.482 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9018 +t=90: Selected seed 195 with value = 0.9018 +Query 1/1: Action query time = 4.340 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=106: Selected seed 195 with value = 0.9903 +Query 1/1: Action query time = 4.235 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6436 +t=122: Selected seed 195 with value = 0.6436 +Query 1/1: Action query time = 4.036 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7452 +t=138: Selected seed 195 with value = 0.7452 +Query 1/1: Action query time = 3.836 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8459 +t=154: Selected seed 195 with value = 0.8459 +Query 1/1: Action query time = 3.027 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9500 +t=170: Selected seed 195 with value = 0.9500 +Query 1/1: Action query time = 2.900 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9370 +t=186: Selected seed 195 with value = 0.9370 +Query 1/1: Action query time = 2.536 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9328 +t=202: Selected seed 195 with value = 0.9328 +Query 1/1: Action query time = 2.619 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=218: Selected seed 195 with value = 0.9871 +Query 1/1: Action query time = 2.374 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9934 +t=234: Selected seed 195 with value = 0.9934 +Query 1/1: Action query time = 2.303 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.924 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.916 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.772 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s07/2026_08_02-04_53_30--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_30--fcv2_175_t3_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_30--fcv2_175_t3_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..a199111f73a652d0f9032e76305fea06ee23aece --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_53_30--fcv2_175_t3_s08.txt @@ -0,0 +1,205 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t3_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 3.978 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3979 +t=10: Selected seed 195 with value = 0.3979 +Query 1/1: Action query time = 4.489 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4604 +t=26: Selected seed 195 with value = 0.4604 +Query 1/1: Action query time = 5.514 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5689 +t=42: Selected seed 195 with value = 0.5689 +Query 1/1: Action query time = 5.135 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6391 +t=58: Selected seed 195 with value = 0.6391 +Query 1/1: Action query time = 4.762 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7234 +t=74: Selected seed 195 with value = 0.7234 +Query 1/1: Action query time = 4.518 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9101 +t=90: Selected seed 195 with value = 0.9101 +Query 1/1: Action query time = 4.500 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=106: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 5.337 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8072 +t=122: Selected seed 195 with value = 0.8072 +Query 1/1: Action query time = 5.087 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7714 +t=138: Selected seed 195 with value = 0.7714 +Query 1/1: Action query time = 5.031 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8763 +t=154: Selected seed 195 with value = 0.8763 +Query 1/1: Action query time = 5.432 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9499 +t=170: Selected seed 195 with value = 0.9499 +Query 1/1: Action query time = 4.297 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9528 +t=186: Selected seed 195 with value = 0.9528 +Query 1/1: Action query time = 3.499 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8952 +t=202: Selected seed 195 with value = 0.8952 +Query 1/1: Action query time = 4.172 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8936 +t=218: Selected seed 195 with value = 0.8936 +Query 1/1: Action query time = 5.056 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8947 +t=234: Selected seed 195 with value = 0.8947 +Query 1/1: Action query time = 5.038 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8950 +t=250: Selected seed 195 with value = 0.8950 +Query 1/1: Action query time = 5.467 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8873 +t=266: Selected seed 195 with value = 0.8873 +Query 1/1: Action query time = 4.585 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8900 +t=282: Selected seed 195 with value = 0.8900 +Query 1/1: Action query time = 4.365 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8920 +t=298: Selected seed 195 with value = 0.8920 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s08/2026_08_02-04_53_30--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 3.829 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3556 +t=10: Selected seed 195 with value = 0.3556 +Query 1/1: Action query time = 4.830 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4707 +t=26: Selected seed 195 with value = 0.4707 +Query 1/1: Action query time = 5.102 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5442 +t=42: Selected seed 195 with value = 0.5442 +Query 1/1: Action query time = 5.243 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6292 +t=58: Selected seed 195 with value = 0.6292 +Query 1/1: Action query time = 5.681 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7426 +t=74: Selected seed 195 with value = 0.7426 +Query 1/1: Action query time = 4.714 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8588 +t=90: Selected seed 195 with value = 0.8588 +Query 1/1: Action query time = 4.504 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9708 +t=106: Selected seed 195 with value = 0.9708 +Query 1/1: Action query time = 5.293 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9898 +t=122: Selected seed 195 with value = 0.9898 +Query 1/1: Action query time = 4.646 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6343 +t=138: Selected seed 195 with value = 0.6343 +Query 1/1: Action query time = 3.548 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7665 +t=154: Selected seed 195 with value = 0.7665 +Query 1/1: Action query time = 5.313 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8334 +t=170: Selected seed 195 with value = 0.8334 +Query 1/1: Action query time = 3.261 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9385 +t=186: Selected seed 195 with value = 0.9385 +Query 1/1: Action query time = 3.649 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=202: Selected seed 195 with value = 0.9928 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s08/2026_08_02-04_53_30--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.783 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3813 +t=10: Selected seed 195 with value = 0.3813 +Query 1/1: Action query time = 5.164 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4448 +t=26: Selected seed 195 with value = 0.4448 +Query 1/1: Action query time = 5.174 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4884 +t=42: Selected seed 195 with value = 0.4884 +Query 1/1: Action query time = 5.305 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6444 +t=58: Selected seed 195 with value = 0.6444 +Query 1/1: Action query time = 5.408 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6859 +t=74: Selected seed 195 with value = 0.6859 +Query 1/1: Action query time = 3.637 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9296 +t=90: Selected seed 195 with value = 0.9296 +Query 1/1: Action query time = 3.508 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9919 +t=106: Selected seed 195 with value = 0.9919 +Query 1/1: Action query time = 5.090 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7567 +t=122: Selected seed 195 with value = 0.7567 +Query 1/1: Action query time = 4.554 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7511 +t=138: Selected seed 195 with value = 0.7511 +Query 1/1: Action query time = 4.716 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8708 +t=154: Selected seed 195 with value = 0.8708 +Query 1/1: Action query time = 4.487 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9855 +t=170: Selected seed 195 with value = 0.9855 +Saved rollout MP4 at path ./rollouts/fcv2_175_t3_s08/2026_08_02-04_53_30--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_44--fcv2_175_t4_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_44--fcv2_175_t4_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..8fa59198eefcae46e4f7259dd65bd1b1711865b1 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_44--fcv2_175_t4_s02.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t4_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 1.663 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4923 +t=10: Selected seed 195 with value = 0.4923 +Query 1/1: Action query time = 2.950 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5965 +t=26: Selected seed 195 with value = 0.5965 +Query 1/1: Action query time = 3.410 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6878 +t=42: Selected seed 195 with value = 0.6878 +Query 1/1: Action query time = 5.127 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8088 +t=58: Selected seed 195 with value = 0.8088 +Query 1/1: Action query time = 5.706 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9743 +t=74: Selected seed 195 with value = 0.9743 +Query 1/1: Action query time = 5.745 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s02/2026_08_02-04_58_44--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.924 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4678 +t=10: Selected seed 195 with value = 0.4678 +Query 1/1: Action query time = 3.592 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5259 +t=26: Selected seed 195 with value = 0.5259 +Query 1/1: Action query time = 5.066 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6339 +t=42: Selected seed 195 with value = 0.6339 +Query 1/1: Action query time = 5.440 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7256 +t=58: Selected seed 195 with value = 0.7256 +Query 1/1: Action query time = 5.414 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9006 +t=74: Selected seed 195 with value = 0.9006 +Query 1/1: Action query time = 4.944 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=90: Selected seed 195 with value = 0.9958 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s02/2026_08_02-04_58_44--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.484 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4798 +t=10: Selected seed 195 with value = 0.4798 +Query 1/1: Action query time = 3.942 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6035 +t=26: Selected seed 195 with value = 0.6035 +Query 1/1: Action query time = 4.104 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7115 +t=42: Selected seed 195 with value = 0.7115 +Query 1/1: Action query time = 5.235 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8162 +t=58: Selected seed 195 with value = 0.8162 +Query 1/1: Action query time = 4.697 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9426 +t=74: Selected seed 195 with value = 0.9426 +Query 1/1: Action query time = 4.557 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9953 +t=90: Selected seed 195 with value = 0.9953 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s02/2026_08_02-04_58_44--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_45--fcv2_175_t4_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_45--fcv2_175_t4_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..c5a3a945f9377cbc60fd10e8e40f834bcce42c96 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_45--fcv2_175_t4_s01.txt @@ -0,0 +1,136 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t4_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 7.078 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5091 +t=10: Selected seed 195 with value = 0.5091 +Query 1/1: Action query time = 5.841 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6142 +t=26: Selected seed 195 with value = 0.6142 +Query 1/1: Action query time = 5.223 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7189 +t=42: Selected seed 195 with value = 0.7189 +Query 1/1: Action query time = 3.602 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8296 +t=58: Selected seed 195 with value = 0.8296 +Query 1/1: Action query time = 3.858 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9710 +t=74: Selected seed 195 with value = 0.9710 +Query 1/1: Action query time = 3.343 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9939 +t=90: Selected seed 195 with value = 0.9939 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s01/2026_08_02-04_58_45--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.764 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5073 +t=10: Selected seed 195 with value = 0.5073 +Query 1/1: Action query time = 4.287 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6182 +t=26: Selected seed 195 with value = 0.6182 +Query 1/1: Action query time = 4.389 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7184 +t=42: Selected seed 195 with value = 0.7184 +Query 1/1: Action query time = 4.758 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8514 +t=58: Selected seed 195 with value = 0.8514 +Query 1/1: Action query time = 3.832 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9622 +t=74: Selected seed 195 with value = 0.9622 +Query 1/1: Action query time = 2.930 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=90: Selected seed 195 with value = 0.9985 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s01/2026_08_02-04_58_45--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.394 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5058 +t=10: Selected seed 195 with value = 0.5058 +Query 1/1: Action query time = 5.589 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5897 +t=26: Selected seed 195 with value = 0.5897 +Query 1/1: Action query time = 5.334 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7085 +t=42: Selected seed 195 with value = 0.7085 +Query 1/1: Action query time = 4.584 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8170 +t=58: Selected seed 195 with value = 0.8170 +Query 1/1: Action query time = 2.152 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9547 +t=74: Selected seed 195 with value = 0.9547 +Query 1/1: Action query time = 1.675 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s01/2026_08_02-04_58_45--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 1.857 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4867 +t=10: Selected seed 195 with value = 0.4867 +Query 1/1: Action query time = 1.876 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5441 +t=26: Selected seed 195 with value = 0.5441 +Query 1/1: Action query time = 1.671 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6422 +t=42: Selected seed 195 with value = 0.6422 +Query 1/1: Action query time = 0.955 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7654 +t=58: Selected seed 195 with value = 0.7654 +Query 1/1: Action query time = 0.969 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9014 +t=74: Selected seed 195 with value = 0.9014 +Query 1/1: Action query time = 0.947 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s01/2026_08_02-04_58_45--with_future_img--episode=4--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 4 +Total successes: 4 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_45--fcv2_175_t4_s10.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_45--fcv2_175_t4_s10.txt new file mode 100644 index 0000000000000000000000000000000000000000..947fe5aeb2d9f97c22ffc1a0030adf0cb433b087 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_45--fcv2_175_t4_s10.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t4_s10', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='10,26,42', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.548 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5064 +t=10: Selected seed 195 with value = 0.5064 +Query 1/1: Action query time = 5.429 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6101 +t=26: Selected seed 195 with value = 0.6101 +Query 1/1: Action query time = 4.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6913 +t=42: Selected seed 195 with value = 0.6913 +Query 1/1: Action query time = 5.167 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8066 +t=58: Selected seed 195 with value = 0.8066 +Query 1/1: Action query time = 5.080 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9594 +t=74: Selected seed 195 with value = 0.9594 +Query 1/1: Action query time = 4.863 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s10/2026_08_02-04_58_45--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 1.796 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4859 +t=10: Selected seed 195 with value = 0.4859 +Query 1/1: Action query time = 3.212 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5743 +t=26: Selected seed 195 with value = 0.5743 +Query 1/1: Action query time = 5.192 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6654 +t=42: Selected seed 195 with value = 0.6654 +Query 1/1: Action query time = 4.964 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7756 +t=58: Selected seed 195 with value = 0.7756 +Query 1/1: Action query time = 4.026 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9357 +t=74: Selected seed 195 with value = 0.9357 +Query 1/1: Action query time = 5.398 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s10/2026_08_02-04_58_45--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.832 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4972 +t=10: Selected seed 195 with value = 0.4972 +Query 1/1: Action query time = 3.214 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5855 +t=26: Selected seed 195 with value = 0.5855 +Query 1/1: Action query time = 5.166 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6590 +t=42: Selected seed 195 with value = 0.6590 +Query 1/1: Action query time = 4.946 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8297 +t=58: Selected seed 195 with value = 0.8297 +Query 1/1: Action query time = 4.527 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9598 +t=74: Selected seed 195 with value = 0.9598 +Query 1/1: Action query time = 4.473 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=90: Selected seed 195 with value = 0.9870 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s10/2026_08_02-04_58_45--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_46--fcv2_175_t4_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_46--fcv2_175_t4_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..bbb256b1e87967a37b122ed13396085651420988 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-04_58_46--fcv2_175_t4_s11.txt @@ -0,0 +1,101 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t4_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.446 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5072 +t=10: Selected seed 195 with value = 0.5072 +Query 1/1: Action query time = 5.193 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6293 +t=26: Selected seed 195 with value = 0.6293 +Query 1/1: Action query time = 4.511 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6979 +t=42: Selected seed 195 with value = 0.6979 +Query 1/1: Action query time = 4.913 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8112 +t=58: Selected seed 195 with value = 0.8112 +Query 1/1: Action query time = 4.870 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9487 +t=74: Selected seed 195 with value = 0.9487 +Query 1/1: Action query time = 3.475 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s11/2026_08_02-04_58_46--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.056 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5154 +t=10: Selected seed 195 with value = 0.5154 +Query 1/1: Action query time = 5.211 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6216 +t=26: Selected seed 195 with value = 0.6216 +Query 1/1: Action query time = 5.481 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7343 +t=42: Selected seed 195 with value = 0.7343 +Query 1/1: Action query time = 4.491 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8261 +t=58: Selected seed 195 with value = 0.8261 +Query 1/1: Action query time = 4.823 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9872 +t=74: Selected seed 195 with value = 0.9872 +Query 1/1: Action query time = 4.336 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9944 +t=90: Selected seed 195 with value = 0.9944 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s11/2026_08_02-04_58_46--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.336 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4998 +t=10: Selected seed 195 with value = 0.4998 +Query 1/1: Action query time = 4.834 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5865 +t=26: Selected seed 195 with value = 0.5865 +Query 1/1: Action query time = 4.695 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6565 +t=42: Selected seed 195 with value = 0.6565 +Query 1/1: Action query time = 3.949 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7805 +t=58: Selected seed 195 with value = 0.7805 +Query 1/1: Action query time = 4.025 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9291 +t=74: Selected seed 195 with value = 0.9291 +Saved rollout MP4 at path ./rollouts/fcv2_175_t4_s11/2026_08_02-04_58_46--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_14--fcv2_175_t5_s05.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_14--fcv2_175_t5_s05.txt new file mode 100644 index 0000000000000000000000000000000000000000..924430601bc0bb7b085071447d1c7bb09e438525 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_14--fcv2_175_t5_s05.txt @@ -0,0 +1,137 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t5_s05', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,21,37', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 5.228 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3289 +t=10: Selected seed 195 with value = 0.3289 +Query 1/1: Action query time = 4.637 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3782 +t=26: Selected seed 195 with value = 0.3782 +Query 1/1: Action query time = 4.988 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4590 +t=42: Selected seed 195 with value = 0.4590 +Query 1/1: Action query time = 5.212 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5069 +t=58: Selected seed 195 with value = 0.5069 +Query 1/1: Action query time = 4.587 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6086 +t=74: Selected seed 195 with value = 0.6086 +Query 1/1: Action query time = 4.703 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7909 +t=90: Selected seed 195 with value = 0.7909 +Query 1/1: Action query time = 4.252 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9222 +t=106: Selected seed 195 with value = 0.9222 +Query 1/1: Action query time = 5.198 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9878 +t=122: Selected seed 195 with value = 0.9878 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s05/2026_08_02-05_01_14--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 6.440 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3184 +t=10: Selected seed 195 with value = 0.3184 +Query 1/1: Action query time = 5.121 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3607 +t=26: Selected seed 195 with value = 0.3607 +Query 1/1: Action query time = 5.163 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4158 +t=42: Selected seed 195 with value = 0.4158 +Query 1/1: Action query time = 5.155 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4982 +t=58: Selected seed 195 with value = 0.4982 +Query 1/1: Action query time = 4.459 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5572 +t=74: Selected seed 195 with value = 0.5572 +Query 1/1: Action query time = 4.329 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5983 +t=90: Selected seed 195 with value = 0.5983 +Query 1/1: Action query time = 3.739 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7964 +t=106: Selected seed 195 with value = 0.7964 +Query 1/1: Action query time = 4.650 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9899 +t=122: Selected seed 195 with value = 0.9899 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s05/2026_08_02-05_01_14--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 5.133 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2856 +t=10: Selected seed 195 with value = 0.2856 +Query 1/1: Action query time = 5.779 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3735 +t=26: Selected seed 195 with value = 0.3735 +Query 1/1: Action query time = 4.790 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4527 +t=42: Selected seed 195 with value = 0.4527 +Query 1/1: Action query time = 4.521 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4830 +t=58: Selected seed 195 with value = 0.4830 +Query 1/1: Action query time = 4.082 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5220 +t=74: Selected seed 195 with value = 0.5220 +Query 1/1: Action query time = 5.446 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6077 +t=90: Selected seed 195 with value = 0.6077 +Query 1/1: Action query time = 5.045 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7010 +t=106: Selected seed 195 with value = 0.7010 +Query 1/1: Action query time = 4.173 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8296 +t=122: Selected seed 195 with value = 0.8296 +Query 1/1: Action query time = 3.626 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9844 +t=138: Selected seed 195 with value = 0.9844 +Query 1/1: Action query time = 1.783 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9933 +t=154: Selected seed 195 with value = 0.9933 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s05/2026_08_02-05_01_14--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_15--fcv2_175_t5_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_15--fcv2_175_t5_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..f683acbda63e9116a8b040e1a33b7a6195efc3f5 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_15--fcv2_175_t5_s08.txt @@ -0,0 +1,137 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t5_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 1.993 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3752 +t=10: Selected seed 195 with value = 0.3752 +Query 1/1: Action query time = 2.357 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3951 +t=26: Selected seed 195 with value = 0.3951 +Query 1/1: Action query time = 5.852 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4666 +t=42: Selected seed 195 with value = 0.4666 +Query 1/1: Action query time = 4.505 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5063 +t=58: Selected seed 195 with value = 0.5063 +Query 1/1: Action query time = 4.614 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5016 +t=74: Selected seed 195 with value = 0.5016 +Query 1/1: Action query time = 5.157 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5856 +t=90: Selected seed 195 with value = 0.5856 +Query 1/1: Action query time = 5.550 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7045 +t=106: Selected seed 195 with value = 0.7045 +Query 1/1: Action query time = 4.676 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8505 +t=122: Selected seed 195 with value = 0.8505 +Query 1/1: Action query time = 3.676 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9624 +t=138: Selected seed 195 with value = 0.9624 +Query 1/1: Action query time = 5.060 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=154: Selected seed 195 with value = 0.9960 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s08/2026_08_02-05_01_15--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.383 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3437 +t=10: Selected seed 195 with value = 0.3437 +Query 1/1: Action query time = 4.944 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4066 +t=26: Selected seed 195 with value = 0.4066 +Query 1/1: Action query time = 4.450 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4792 +t=42: Selected seed 195 with value = 0.4792 +Query 1/1: Action query time = 4.704 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5695 +t=58: Selected seed 195 with value = 0.5695 +Query 1/1: Action query time = 5.038 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6563 +t=74: Selected seed 195 with value = 0.6563 +Query 1/1: Action query time = 4.853 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7489 +t=90: Selected seed 195 with value = 0.7489 +Query 1/1: Action query time = 5.120 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8844 +t=106: Selected seed 195 with value = 0.8844 +Query 1/1: Action query time = 5.279 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9858 +t=122: Selected seed 195 with value = 0.9858 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s08/2026_08_02-05_01_15--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 2.724 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2653 +t=10: Selected seed 195 with value = 0.2653 +Query 1/1: Action query time = 5.282 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2968 +t=26: Selected seed 195 with value = 0.2968 +Query 1/1: Action query time = 4.572 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4490 +t=42: Selected seed 195 with value = 0.4490 +Query 1/1: Action query time = 4.342 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5095 +t=58: Selected seed 195 with value = 0.5095 +Query 1/1: Action query time = 4.836 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6601 +t=74: Selected seed 195 with value = 0.6601 +Query 1/1: Action query time = 5.319 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7544 +t=90: Selected seed 195 with value = 0.7544 +Query 1/1: Action query time = 4.613 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8624 +t=106: Selected seed 195 with value = 0.8624 +Query 1/1: Action query time = 3.789 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9824 +t=122: Selected seed 195 with value = 0.9824 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s08/2026_08_02-05_01_15--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_16--fcv2_175_t5_s10.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_16--fcv2_175_t5_s10.txt new file mode 100644 index 0000000000000000000000000000000000000000..0ec27e6aa9e6a3382197fe4944e0d4b1a9ae0cfb --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_16--fcv2_175_t5_s10.txt @@ -0,0 +1,141 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t5_s10', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='10,26,42', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 6.199 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3527 +t=10: Selected seed 195 with value = 0.3527 +Query 1/1: Action query time = 4.766 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3900 +t=26: Selected seed 195 with value = 0.3900 +Query 1/1: Action query time = 4.716 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4541 +t=42: Selected seed 195 with value = 0.4541 +Query 1/1: Action query time = 5.097 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5698 +t=58: Selected seed 195 with value = 0.5698 +Query 1/1: Action query time = 5.458 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6876 +t=74: Selected seed 195 with value = 0.6876 +Query 1/1: Action query time = 5.231 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8131 +t=90: Selected seed 195 with value = 0.8131 +Query 1/1: Action query time = 3.802 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9349 +t=106: Selected seed 195 with value = 0.9349 +Query 1/1: Action query time = 4.774 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=122: Selected seed 195 with value = 0.9994 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s10/2026_08_02-05_01_16--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 3.953 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3796 +t=10: Selected seed 195 with value = 0.3796 +Query 1/1: Action query time = 4.623 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3882 +t=26: Selected seed 195 with value = 0.3882 +Query 1/1: Action query time = 4.521 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4355 +t=42: Selected seed 195 with value = 0.4355 +Query 1/1: Action query time = 5.110 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4828 +t=58: Selected seed 195 with value = 0.4828 +Query 1/1: Action query time = 5.192 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6028 +t=74: Selected seed 195 with value = 0.6028 +Query 1/1: Action query time = 5.068 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6933 +t=90: Selected seed 195 with value = 0.6933 +Query 1/1: Action query time = 4.973 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7656 +t=106: Selected seed 195 with value = 0.7656 +Query 1/1: Action query time = 5.193 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8948 +t=122: Selected seed 195 with value = 0.8948 +Query 1/1: Action query time = 2.804 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=138: Selected seed 195 with value = 0.9988 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s10/2026_08_02-05_01_16--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 4.849 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2875 +t=10: Selected seed 195 with value = 0.2875 +Query 1/1: Action query time = 5.615 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3833 +t=26: Selected seed 195 with value = 0.3833 +Query 1/1: Action query time = 4.718 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4071 +t=42: Selected seed 195 with value = 0.4071 +Query 1/1: Action query time = 4.334 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4920 +t=58: Selected seed 195 with value = 0.4920 +Query 1/1: Action query time = 4.570 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5606 +t=74: Selected seed 195 with value = 0.5606 +Query 1/1: Action query time = 4.851 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6097 +t=90: Selected seed 195 with value = 0.6097 +Query 1/1: Action query time = 3.579 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7253 +t=106: Selected seed 195 with value = 0.7253 +Query 1/1: Action query time = 2.546 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8271 +t=122: Selected seed 195 with value = 0.8271 +Query 1/1: Action query time = 1.762 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9258 +t=138: Selected seed 195 with value = 0.9258 +Query 1/1: Action query time = 1.167 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s10/2026_08_02-05_01_16--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_16--fcv2_175_t5_s15.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_16--fcv2_175_t5_s15.txt new file mode 100644 index 0000000000000000000000000000000000000000..6f62f32c980f2643a9491a04a5dafe759ad9a188 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_01_16--fcv2_175_t5_s15.txt @@ -0,0 +1,137 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t5_s15', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='15,31,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 7.562 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3872 +t=10: Selected seed 195 with value = 0.3872 +Query 1/1: Action query time = 5.728 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4606 +t=26: Selected seed 195 with value = 0.4606 +Query 1/1: Action query time = 5.435 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5436 +t=42: Selected seed 195 with value = 0.5436 +Query 1/1: Action query time = 4.280 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5107 +t=58: Selected seed 195 with value = 0.5107 +Query 1/1: Action query time = 4.741 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6015 +t=74: Selected seed 195 with value = 0.6015 +Query 1/1: Action query time = 4.875 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7758 +t=90: Selected seed 195 with value = 0.7758 +Query 1/1: Action query time = 4.844 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8599 +t=106: Selected seed 195 with value = 0.8599 +Query 1/1: Action query time = 2.893 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9679 +t=122: Selected seed 195 with value = 0.9679 +Query 1/1: Action query time = 1.965 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9929 +t=138: Selected seed 195 with value = 0.9929 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s15/2026_08_02-05_01_16--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 3.966 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3342 +t=10: Selected seed 195 with value = 0.3342 +Query 1/1: Action query time = 4.903 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3779 +t=26: Selected seed 195 with value = 0.3779 +Query 1/1: Action query time = 5.421 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4377 +t=42: Selected seed 195 with value = 0.4377 +Query 1/1: Action query time = 5.220 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5910 +t=58: Selected seed 195 with value = 0.5910 +Query 1/1: Action query time = 5.303 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7004 +t=74: Selected seed 195 with value = 0.7004 +Query 1/1: Action query time = 4.759 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8383 +t=90: Selected seed 195 with value = 0.8383 +Query 1/1: Action query time = 4.921 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9541 +t=106: Selected seed 195 with value = 0.9541 +Query 1/1: Action query time = 3.246 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=122: Selected seed 195 with value = 0.9986 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s15/2026_08_02-05_01_16--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 4.344 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3474 +t=10: Selected seed 195 with value = 0.3474 +Query 1/1: Action query time = 4.637 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4101 +t=26: Selected seed 195 with value = 0.4101 +Query 1/1: Action query time = 5.095 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4228 +t=42: Selected seed 195 with value = 0.4228 +Query 1/1: Action query time = 5.331 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4985 +t=58: Selected seed 195 with value = 0.4985 +Query 1/1: Action query time = 5.622 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6418 +t=74: Selected seed 195 with value = 0.6418 +Query 1/1: Action query time = 4.306 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8063 +t=90: Selected seed 195 with value = 0.8063 +Query 1/1: Action query time = 4.016 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7467 +t=106: Selected seed 195 with value = 0.7467 +Query 1/1: Action query time = 2.780 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8869 +t=122: Selected seed 195 with value = 0.8869 +Query 1/1: Action query time = 2.066 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9716 +t=138: Selected seed 195 with value = 0.9716 +Saved rollout MP4 at path ./rollouts/fcv2_175_t5_s15/2026_08_02-05_01_16--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_34--fcv2_175_t6_s00.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_34--fcv2_175_t6_s00.txt new file mode 100644 index 0000000000000000000000000000000000000000..aaf789198f8bf4516589bb5b9488347f53185029 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_34--fcv2_175_t6_s00.txt @@ -0,0 +1,132 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t6_s00', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,16,32,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 2.036 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4391 +t=10: Selected seed 195 with value = 0.4391 +Query 1/1: Action query time = 2.478 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5535 +t=26: Selected seed 195 with value = 0.5535 +Query 1/1: Action query time = 3.133 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6542 +t=42: Selected seed 195 with value = 0.6542 +Query 1/1: Action query time = 5.087 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5248 +t=58: Selected seed 195 with value = 0.5248 +Query 1/1: Action query time = 5.692 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6702 +t=74: Selected seed 195 with value = 0.6702 +Query 1/1: Action query time = 5.612 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8191 +t=90: Selected seed 195 with value = 0.8191 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s00/2026_08_02-05_04_34--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 4.184 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4561 +t=10: Selected seed 195 with value = 0.4561 +Query 1/1: Action query time = 4.513 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5367 +t=26: Selected seed 195 with value = 0.5367 +Query 1/1: Action query time = 4.663 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6253 +t=42: Selected seed 195 with value = 0.6253 +Query 1/1: Action query time = 5.953 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7168 +t=58: Selected seed 195 with value = 0.7168 +Query 1/1: Action query time = 3.885 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8527 +t=74: Selected seed 195 with value = 0.8527 +Query 1/1: Action query time = 4.319 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9674 +t=90: Selected seed 195 with value = 0.9674 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s00/2026_08_02-05_04_34--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.092 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4267 +t=10: Selected seed 195 with value = 0.4267 +Query 1/1: Action query time = 4.038 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4726 +t=26: Selected seed 195 with value = 0.4726 +Query 1/1: Action query time = 5.609 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5953 +t=42: Selected seed 195 with value = 0.5953 +Query 1/1: Action query time = 4.791 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6872 +t=58: Selected seed 195 with value = 0.6872 +Query 1/1: Action query time = 4.819 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8557 +t=74: Selected seed 195 with value = 0.8557 +Query 1/1: Action query time = 4.091 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9721 +t=90: Selected seed 195 with value = 0.9721 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s00/2026_08_02-05_04_34--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 4... +Query 1/1: Action query time = 3.120 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4214 +t=10: Selected seed 195 with value = 0.4214 +Query 1/1: Action query time = 2.553 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5030 +t=26: Selected seed 195 with value = 0.5030 +Query 1/1: Action query time = 3.477 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6214 +t=42: Selected seed 195 with value = 0.6214 +Query 1/1: Action query time = 2.866 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7392 +t=58: Selected seed 195 with value = 0.7392 +Query 1/1: Action query time = 2.246 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8957 +t=74: Selected seed 195 with value = 0.8957 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s00/2026_08_02-05_04_34--with_future_img--episode=4--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 4 +Total successes: 4 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_35--fcv2_175_t6_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_35--fcv2_175_t6_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..ef9103f6d8b7521f9148fa8a038156a94ee279ea --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_35--fcv2_175_t6_s08.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t6_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 2.799 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4497 +t=10: Selected seed 195 with value = 0.4497 +Query 1/1: Action query time = 5.662 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5212 +t=26: Selected seed 195 with value = 0.5212 +Query 1/1: Action query time = 4.271 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6283 +t=42: Selected seed 195 with value = 0.6283 +Query 1/1: Action query time = 3.989 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7160 +t=58: Selected seed 195 with value = 0.7160 +Query 1/1: Action query time = 5.881 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8693 +t=74: Selected seed 195 with value = 0.8693 +Query 1/1: Action query time = 4.698 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9865 +t=90: Selected seed 195 with value = 0.9865 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s08/2026_08_02-05_04_35--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 3.843 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4289 +t=10: Selected seed 195 with value = 0.4289 +Query 1/1: Action query time = 5.276 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5507 +t=26: Selected seed 195 with value = 0.5507 +Query 1/1: Action query time = 5.509 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6340 +t=42: Selected seed 195 with value = 0.6340 +Query 1/1: Action query time = 4.779 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7443 +t=58: Selected seed 195 with value = 0.7443 +Query 1/1: Action query time = 4.906 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8836 +t=74: Selected seed 195 with value = 0.8836 +Query 1/1: Action query time = 5.667 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=90: Selected seed 195 with value = 0.9993 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s08/2026_08_02-05_04_35--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 3.704 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4746 +t=10: Selected seed 195 with value = 0.4746 +Query 1/1: Action query time = 5.105 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5263 +t=26: Selected seed 195 with value = 0.5263 +Query 1/1: Action query time = 3.804 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5986 +t=42: Selected seed 195 with value = 0.5986 +Query 1/1: Action query time = 4.811 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6098 +t=58: Selected seed 195 with value = 0.6098 +Query 1/1: Action query time = 5.182 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6484 +t=74: Selected seed 195 with value = 0.6484 +Query 1/1: Action query time = 4.994 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7019 +t=90: Selected seed 195 with value = 0.7019 +Query 1/1: Action query time = 3.597 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7374 +t=106: Selected seed 195 with value = 0.7374 +Query 1/1: Action query time = 2.553 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7731 +t=122: Selected seed 195 with value = 0.7731 +Query 1/1: Action query time = 3.577 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8768 +t=138: Selected seed 195 with value = 0.8768 +Query 1/1: Action query time = 2.951 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8970 +t=154: Selected seed 195 with value = 0.8970 +Query 1/1: Action query time = 3.081 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9077 +t=170: Selected seed 195 with value = 0.9077 +Query 1/1: Action query time = 3.146 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9228 +t=186: Selected seed 195 with value = 0.9228 +Query 1/1: Action query time = 3.171 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=202: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 3.327 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9918 +t=218: Selected seed 195 with value = 0.9918 +Query 1/1: Action query time = 2.231 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=234: Selected seed 195 with value = 0.9928 +Query 1/1: Action query time = 1.728 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9811 +t=250: Selected seed 195 with value = 0.9811 +Query 1/1: Action query time = 2.819 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8167 +t=266: Selected seed 195 with value = 0.8167 +Query 1/1: Action query time = 3.283 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9006 +t=282: Selected seed 195 with value = 0.9006 +Query 1/1: Action query time = 3.353 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s08/2026_08_02-05_04_35--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_36--fcv2_175_t6_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_36--fcv2_175_t6_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..19d3e6ca3e34201aac064166af4c01cfc250eb71 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_36--fcv2_175_t6_s07.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t6_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 4.321 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4129 +t=10: Selected seed 195 with value = 0.4129 +Query 1/1: Action query time = 6.032 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5142 +t=26: Selected seed 195 with value = 0.5142 +Query 1/1: Action query time = 5.259 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6303 +t=42: Selected seed 195 with value = 0.6303 +Query 1/1: Action query time = 3.153 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7235 +t=58: Selected seed 195 with value = 0.7235 +Query 1/1: Action query time = 4.985 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8583 +t=74: Selected seed 195 with value = 0.8583 +Query 1/1: Action query time = 4.752 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s07/2026_08_02-05_04_36--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 3.794 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4633 +t=10: Selected seed 195 with value = 0.4633 +Query 1/1: Action query time = 4.224 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5235 +t=26: Selected seed 195 with value = 0.5235 +Query 1/1: Action query time = 3.796 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6237 +t=42: Selected seed 195 with value = 0.6237 +Query 1/1: Action query time = 3.477 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7500 +t=58: Selected seed 195 with value = 0.7500 +Query 1/1: Action query time = 5.871 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8824 +t=74: Selected seed 195 with value = 0.8824 +Query 1/1: Action query time = 5.033 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9755 +t=90: Selected seed 195 with value = 0.9755 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s07/2026_08_02-05_04_36--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.209 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4676 +t=10: Selected seed 195 with value = 0.4676 +Query 1/1: Action query time = 5.334 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5373 +t=26: Selected seed 195 with value = 0.5373 +Query 1/1: Action query time = 4.434 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6277 +t=42: Selected seed 195 with value = 0.6277 +Query 1/1: Action query time = 3.814 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6996 +t=58: Selected seed 195 with value = 0.6996 +Query 1/1: Action query time = 4.932 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8097 +t=74: Selected seed 195 with value = 0.8097 +Query 1/1: Action query time = 3.982 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9317 +t=90: Selected seed 195 with value = 0.9317 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s07/2026_08_02-05_04_36--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_36--fcv2_175_t6_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_36--fcv2_175_t6_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..063a449cb3543108400f932a831cbe742c5afd65 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-05_04_36--fcv2_175_t6_s11.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_175_t6_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 2.634 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4721 +t=10: Selected seed 195 with value = 0.4721 +Query 1/1: Action query time = 4.081 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5405 +t=26: Selected seed 195 with value = 0.5405 +Query 1/1: Action query time = 3.937 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6223 +t=42: Selected seed 195 with value = 0.6223 +Query 1/1: Action query time = 5.669 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5774 +t=58: Selected seed 195 with value = 0.5774 +Query 1/1: Action query time = 5.053 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6901 +t=74: Selected seed 195 with value = 0.6901 +Query 1/1: Action query time = 3.846 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7410 +t=90: Selected seed 195 with value = 0.7410 +Query 1/1: Action query time = 5.437 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8562 +t=106: Selected seed 195 with value = 0.8562 +Query 1/1: Action query time = 4.332 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9189 +t=122: Selected seed 195 with value = 0.9189 +Query 1/1: Action query time = 5.652 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9701 +t=138: Selected seed 195 with value = 0.9701 +Query 1/1: Action query time = 5.564 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9828 +t=154: Selected seed 195 with value = 0.9828 +Query 1/1: Action query time = 4.608 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9753 +t=170: Selected seed 195 with value = 0.9753 +Query 1/1: Action query time = 5.212 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=186: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 5.640 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9845 +t=202: Selected seed 195 with value = 0.9845 +Query 1/1: Action query time = 4.100 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9672 +t=218: Selected seed 195 with value = 0.9672 +Query 1/1: Action query time = 4.517 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9533 +t=234: Selected seed 195 with value = 0.9533 +Query 1/1: Action query time = 5.256 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9470 +t=250: Selected seed 195 with value = 0.9470 +Query 1/1: Action query time = 4.783 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9413 +t=266: Selected seed 195 with value = 0.9413 +Query 1/1: Action query time = 4.957 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9386 +t=282: Selected seed 195 with value = 0.9386 +Query 1/1: Action query time = 4.583 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9371 +t=298: Selected seed 195 with value = 0.9371 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s11/2026_08_02-05_04_36--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 3.553 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4352 +t=10: Selected seed 195 with value = 0.4352 +Query 1/1: Action query time = 4.069 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5035 +t=26: Selected seed 195 with value = 0.5035 +Query 1/1: Action query time = 3.037 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5880 +t=42: Selected seed 195 with value = 0.5880 +Query 1/1: Action query time = 2.820 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6935 +t=58: Selected seed 195 with value = 0.6935 +Query 1/1: Action query time = 2.729 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8086 +t=74: Selected seed 195 with value = 0.8086 +Query 1/1: Action query time = 2.265 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9614 +t=90: Selected seed 195 with value = 0.9614 +Query 1/1: Action query time = 2.976 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9956 +t=106: Selected seed 195 with value = 0.9956 +Query 1/1: Action query time = 2.022 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9897 +t=122: Selected seed 195 with value = 0.9897 +Query 1/1: Action query time = 1.803 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9915 +t=138: Selected seed 195 with value = 0.9915 +Query 1/1: Action query time = 2.876 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9324 +t=154: Selected seed 195 with value = 0.9324 +Query 1/1: Action query time = 3.095 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8880 +t=170: Selected seed 195 with value = 0.8880 +Query 1/1: Action query time = 3.263 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.522 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=202: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 2.518 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.185 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.337 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.141 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.094 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.067 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9015 +t=298: Selected seed 195 with value = 0.9015 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s11/2026_08_02-05_04_36--with_future_img--episode=2--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 0.966 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4419 +t=10: Selected seed 195 with value = 0.4419 +Query 1/1: Action query time = 0.972 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4171 +t=26: Selected seed 195 with value = 0.4171 +Query 1/1: Action query time = 0.985 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6151 +t=42: Selected seed 195 with value = 0.6151 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6702 +t=58: Selected seed 195 with value = 0.6702 +Query 1/1: Action query time = 0.987 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6141 +t=74: Selected seed 195 with value = 0.6141 +Query 1/1: Action query time = 0.954 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7071 +t=90: Selected seed 195 with value = 0.7071 +Query 1/1: Action query time = 0.975 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7769 +t=106: Selected seed 195 with value = 0.7769 +Query 1/1: Action query time = 0.950 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8472 +t=122: Selected seed 195 with value = 0.8472 +Query 1/1: Action query time = 0.968 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9224 +t=138: Selected seed 195 with value = 0.9224 +Query 1/1: Action query time = 0.953 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9688 +t=154: Selected seed 195 with value = 0.9688 +Query 1/1: Action query time = 0.985 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9699 +t=170: Selected seed 195 with value = 0.9699 +Query 1/1: Action query time = 0.971 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9769 +t=186: Selected seed 195 with value = 0.9769 +Query 1/1: Action query time = 0.967 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.964 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9952 +t=218: Selected seed 195 with value = 0.9952 +Query 1/1: Action query time = 0.969 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9846 +t=234: Selected seed 195 with value = 0.9846 +Query 1/1: Action query time = 0.970 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9763 +t=250: Selected seed 195 with value = 0.9763 +Query 1/1: Action query time = 0.965 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9710 +t=266: Selected seed 195 with value = 0.9710 +Query 1/1: Action query time = 0.973 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9681 +t=282: Selected seed 195 with value = 0.9681 +Query 1/1: Action query time = 0.975 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9659 +t=298: Selected seed 195 with value = 0.9659 +Saved rollout MP4 at path ./rollouts/fcv2_175_t6_s11/2026_08_02-05_04_36--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_12--fcv2_350_t0_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_12--fcv2_350_t0_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..f4a0bd1c978470b094ed6bb1a7eff0f3e0d77cb1 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_12--fcv2_350_t0_s03.txt @@ -0,0 +1,133 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t0_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.537 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3419 +t=10: Selected seed 195 with value = 0.3419 +Query 1/1: Action query time = 5.258 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3548 +t=26: Selected seed 195 with value = 0.3548 +Query 1/1: Action query time = 4.613 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5062 +t=42: Selected seed 195 with value = 0.5062 +Query 1/1: Action query time = 3.939 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5665 +t=58: Selected seed 195 with value = 0.5665 +Query 1/1: Action query time = 4.316 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6475 +t=74: Selected seed 195 with value = 0.6475 +Query 1/1: Action query time = 4.929 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8910 +t=90: Selected seed 195 with value = 0.8910 +Query 1/1: Action query time = 5.516 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.843 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s03/2026_08_02-08_22_12--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.798 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4128 +t=10: Selected seed 195 with value = 0.4128 +Query 1/1: Action query time = 4.078 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4821 +t=26: Selected seed 195 with value = 0.4821 +Query 1/1: Action query time = 5.699 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5513 +t=42: Selected seed 195 with value = 0.5513 +Query 1/1: Action query time = 4.948 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6341 +t=58: Selected seed 195 with value = 0.6341 +Query 1/1: Action query time = 4.512 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7129 +t=74: Selected seed 195 with value = 0.7129 +Query 1/1: Action query time = 4.633 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7630 +t=90: Selected seed 195 with value = 0.7630 +Query 1/1: Action query time = 4.063 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8542 +t=106: Selected seed 195 with value = 0.8542 +Query 1/1: Action query time = 4.216 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.510 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s03/2026_08_02-08_22_12--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.788 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3635 +t=10: Selected seed 195 with value = 0.3635 +Query 1/1: Action query time = 6.098 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3882 +t=26: Selected seed 195 with value = 0.3882 +Query 1/1: Action query time = 5.242 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4332 +t=42: Selected seed 195 with value = 0.4332 +Query 1/1: Action query time = 4.679 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5239 +t=58: Selected seed 195 with value = 0.5239 +Query 1/1: Action query time = 4.465 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5801 +t=74: Selected seed 195 with value = 0.5801 +Query 1/1: Action query time = 4.277 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7553 +t=90: Selected seed 195 with value = 0.7553 +Query 1/1: Action query time = 3.821 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9145 +t=106: Selected seed 195 with value = 0.9145 +Query 1/1: Action query time = 3.005 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s03/2026_08_02-08_22_12--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_12--fcv2_350_t0_s05.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_12--fcv2_350_t0_s05.txt new file mode 100644 index 0000000000000000000000000000000000000000..787195f767aa5612485bf1306bdd916be6975058 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_12--fcv2_350_t0_s05.txt @@ -0,0 +1,137 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t0_s05', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,21,37', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.419 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3449 +t=10: Selected seed 195 with value = 0.3449 +Query 1/1: Action query time = 4.298 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4242 +t=26: Selected seed 195 with value = 0.4242 +Query 1/1: Action query time = 5.242 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4966 +t=42: Selected seed 195 with value = 0.4966 +Query 1/1: Action query time = 5.388 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5941 +t=58: Selected seed 195 with value = 0.5941 +Query 1/1: Action query time = 5.638 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7058 +t=74: Selected seed 195 with value = 0.7058 +Query 1/1: Action query time = 5.310 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7687 +t=90: Selected seed 195 with value = 0.7687 +Query 1/1: Action query time = 3.522 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8799 +t=106: Selected seed 195 with value = 0.8799 +Query 1/1: Action query time = 4.158 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9846 +t=122: Selected seed 195 with value = 0.9846 +Query 1/1: Action query time = 3.536 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s05/2026_08_02-08_22_12--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.501 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3542 +t=10: Selected seed 195 with value = 0.3542 +Query 1/1: Action query time = 4.499 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3678 +t=26: Selected seed 195 with value = 0.3678 +Query 1/1: Action query time = 5.703 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4810 +t=42: Selected seed 195 with value = 0.4810 +Query 1/1: Action query time = 5.279 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5974 +t=58: Selected seed 195 with value = 0.5974 +Query 1/1: Action query time = 4.426 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7063 +t=74: Selected seed 195 with value = 0.7063 +Query 1/1: Action query time = 4.383 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8254 +t=90: Selected seed 195 with value = 0.8254 +Query 1/1: Action query time = 5.304 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9594 +t=106: Selected seed 195 with value = 0.9594 +Query 1/1: Action query time = 3.286 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9937 +t=122: Selected seed 195 with value = 0.9937 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s05/2026_08_02-08_22_12--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.584 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3488 +t=10: Selected seed 195 with value = 0.3488 +Query 1/1: Action query time = 4.475 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3904 +t=26: Selected seed 195 with value = 0.3904 +Query 1/1: Action query time = 5.244 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4261 +t=42: Selected seed 195 with value = 0.4261 +Query 1/1: Action query time = 5.329 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5143 +t=58: Selected seed 195 with value = 0.5143 +Query 1/1: Action query time = 4.542 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5762 +t=74: Selected seed 195 with value = 0.5762 +Query 1/1: Action query time = 4.467 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7039 +t=90: Selected seed 195 with value = 0.7039 +Query 1/1: Action query time = 2.908 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8939 +t=106: Selected seed 195 with value = 0.8939 +Query 1/1: Action query time = 3.338 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=122: Selected seed 195 with value = 0.9980 +Query 1/1: Action query time = 2.123 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s05/2026_08_02-08_22_12--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_13--fcv2_350_t0_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_13--fcv2_350_t0_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..b6d2fec5ba946b678126207e97b7b918fd8de2bc --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_13--fcv2_350_t0_s11.txt @@ -0,0 +1,129 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t0_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.709 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3612 +t=10: Selected seed 195 with value = 0.3612 +Query 1/1: Action query time = 4.711 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4113 +t=26: Selected seed 195 with value = 0.4113 +Query 1/1: Action query time = 4.506 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5133 +t=42: Selected seed 195 with value = 0.5133 +Query 1/1: Action query time = 4.735 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6071 +t=58: Selected seed 195 with value = 0.6071 +Query 1/1: Action query time = 4.781 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7433 +t=74: Selected seed 195 with value = 0.7433 +Query 1/1: Action query time = 4.919 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8331 +t=90: Selected seed 195 with value = 0.8331 +Query 1/1: Action query time = 4.590 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.597 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=122: Selected seed 195 with value = 0.9985 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s11/2026_08_02-08_22_13--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.866 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3840 +t=10: Selected seed 195 with value = 0.3840 +Query 1/1: Action query time = 4.250 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4676 +t=26: Selected seed 195 with value = 0.4676 +Query 1/1: Action query time = 4.290 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5406 +t=42: Selected seed 195 with value = 0.5406 +Query 1/1: Action query time = 5.302 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6282 +t=58: Selected seed 195 with value = 0.6282 +Query 1/1: Action query time = 4.743 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7191 +t=74: Selected seed 195 with value = 0.7191 +Query 1/1: Action query time = 5.118 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8665 +t=90: Selected seed 195 with value = 0.8665 +Query 1/1: Action query time = 5.099 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.324 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s11/2026_08_02-08_22_13--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.786 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3438 +t=10: Selected seed 195 with value = 0.3438 +Query 1/1: Action query time = 4.094 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3572 +t=26: Selected seed 195 with value = 0.3572 +Query 1/1: Action query time = 2.971 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5074 +t=42: Selected seed 195 with value = 0.5074 +Query 1/1: Action query time = 4.573 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5875 +t=58: Selected seed 195 with value = 0.5875 +Query 1/1: Action query time = 4.379 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6811 +t=74: Selected seed 195 with value = 0.6811 +Query 1/1: Action query time = 4.989 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8927 +t=90: Selected seed 195 with value = 0.8927 +Query 1/1: Action query time = 5.129 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.135 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s11/2026_08_02-08_22_13--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_13--fcv2_350_t0_s13.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_13--fcv2_350_t0_s13.txt new file mode 100644 index 0000000000000000000000000000000000000000..5d8ee96240c928975d7a368adc55cd40686d71ef --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_22_13--fcv2_350_t0_s13.txt @@ -0,0 +1,129 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t0_s13', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='13,29,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.784 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3702 +t=10: Selected seed 195 with value = 0.3702 +Query 1/1: Action query time = 5.310 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4150 +t=26: Selected seed 195 with value = 0.4150 +Query 1/1: Action query time = 5.595 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4854 +t=42: Selected seed 195 with value = 0.4854 +Query 1/1: Action query time = 5.435 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5638 +t=58: Selected seed 195 with value = 0.5638 +Query 1/1: Action query time = 5.113 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6883 +t=74: Selected seed 195 with value = 0.6883 +Query 1/1: Action query time = 4.406 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7916 +t=90: Selected seed 195 with value = 0.7916 +Query 1/1: Action query time = 4.917 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9062 +t=106: Selected seed 195 with value = 0.9062 +Query 1/1: Action query time = 4.687 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s13/2026_08_02-08_22_13--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.223 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3601 +t=10: Selected seed 195 with value = 0.3601 +Query 1/1: Action query time = 4.050 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3977 +t=26: Selected seed 195 with value = 0.3977 +Query 1/1: Action query time = 5.405 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4501 +t=42: Selected seed 195 with value = 0.4501 +Query 1/1: Action query time = 4.892 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5267 +t=58: Selected seed 195 with value = 0.5267 +Query 1/1: Action query time = 5.143 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6123 +t=74: Selected seed 195 with value = 0.6123 +Query 1/1: Action query time = 5.140 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7955 +t=90: Selected seed 195 with value = 0.7955 +Query 1/1: Action query time = 5.484 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9383 +t=106: Selected seed 195 with value = 0.9383 +Query 1/1: Action query time = 4.376 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s13/2026_08_02-08_22_13--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.833 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3575 +t=10: Selected seed 195 with value = 0.3575 +Query 1/1: Action query time = 3.138 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4135 +t=26: Selected seed 195 with value = 0.4135 +Query 1/1: Action query time = 4.561 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4329 +t=42: Selected seed 195 with value = 0.4329 +Query 1/1: Action query time = 5.506 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5277 +t=58: Selected seed 195 with value = 0.5277 +Query 1/1: Action query time = 5.058 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=74: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 4.832 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7723 +t=90: Selected seed 195 with value = 0.7723 +Query 1/1: Action query time = 4.364 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9195 +t=106: Selected seed 195 with value = 0.9195 +Query 1/1: Action query time = 3.754 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t0_s13/2026_08_02-08_22_13--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_33--fcv2_350_t1_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_33--fcv2_350_t1_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..b50b26e246166d60a99c0d5d88f59828a16b5469 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_33--fcv2_350_t1_s02.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t1_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 4.886 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5137 +t=10: Selected seed 195 with value = 0.5137 +Query 1/1: Action query time = 5.019 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6323 +t=26: Selected seed 195 with value = 0.6323 +Query 1/1: Action query time = 5.869 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6958 +t=42: Selected seed 195 with value = 0.6958 +Query 1/1: Action query time = 5.498 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8576 +t=58: Selected seed 195 with value = 0.8576 +Query 1/1: Action query time = 4.258 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=74: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 4.252 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s02/2026_08_02-08_25_33--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 2.897 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4956 +t=10: Selected seed 195 with value = 0.4956 +Query 1/1: Action query time = 5.322 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5890 +t=26: Selected seed 195 with value = 0.5890 +Query 1/1: Action query time = 5.444 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6922 +t=42: Selected seed 195 with value = 0.6922 +Query 1/1: Action query time = 4.770 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7955 +t=58: Selected seed 195 with value = 0.7955 +Query 1/1: Action query time = 4.496 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9338 +t=74: Selected seed 195 with value = 0.9338 +Query 1/1: Action query time = 4.231 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s02/2026_08_02-08_25_33--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 3.321 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5097 +t=10: Selected seed 195 with value = 0.5097 +Query 1/1: Action query time = 5.819 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5991 +t=26: Selected seed 195 with value = 0.5991 +Query 1/1: Action query time = 5.241 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7172 +t=42: Selected seed 195 with value = 0.7172 +Query 1/1: Action query time = 4.790 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8253 +t=58: Selected seed 195 with value = 0.8253 +Query 1/1: Action query time = 4.170 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9633 +t=74: Selected seed 195 with value = 0.9633 +Query 1/1: Action query time = 3.548 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s02/2026_08_02-08_25_33--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_33--fcv2_350_t1_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_33--fcv2_350_t1_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..c97abbec5892e8e0311d4376c59962741a2b398b --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_33--fcv2_350_t1_s07.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t1_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 6.413 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4940 +t=10: Selected seed 195 with value = 0.4940 +Query 1/1: Action query time = 5.130 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5990 +t=26: Selected seed 195 with value = 0.5990 +Query 1/1: Action query time = 4.343 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7134 +t=42: Selected seed 195 with value = 0.7134 +Query 1/1: Action query time = 4.958 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8183 +t=58: Selected seed 195 with value = 0.8183 +Query 1/1: Action query time = 4.796 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9609 +t=74: Selected seed 195 with value = 0.9609 +Query 1/1: Action query time = 3.436 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s07/2026_08_02-08_25_33--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 4.333 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5109 +t=10: Selected seed 195 with value = 0.5109 +Query 1/1: Action query time = 4.988 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6027 +t=26: Selected seed 195 with value = 0.6027 +Query 1/1: Action query time = 5.031 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7108 +t=42: Selected seed 195 with value = 0.7108 +Query 1/1: Action query time = 5.030 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8115 +t=58: Selected seed 195 with value = 0.8115 +Query 1/1: Action query time = 5.264 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9427 +t=74: Selected seed 195 with value = 0.9427 +Query 1/1: Action query time = 3.285 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=90: Selected seed 195 with value = 0.9924 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s07/2026_08_02-08_25_33--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 4.643 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5287 +t=10: Selected seed 195 with value = 0.5287 +Query 1/1: Action query time = 4.792 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5941 +t=26: Selected seed 195 with value = 0.5941 +Query 1/1: Action query time = 5.195 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6950 +t=42: Selected seed 195 with value = 0.6950 +Query 1/1: Action query time = 4.442 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7999 +t=58: Selected seed 195 with value = 0.7999 +Query 1/1: Action query time = 4.032 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9331 +t=74: Selected seed 195 with value = 0.9331 +Query 1/1: Action query time = 1.964 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s07/2026_08_02-08_25_33--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_33--fcv2_350_t1_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_33--fcv2_350_t1_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..a734d444f3b5bb10656dc742674d16e4916c239f --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_33--fcv2_350_t1_s08.txt @@ -0,0 +1,109 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t1_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 4.325 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4965 +t=10: Selected seed 195 with value = 0.4965 +Query 1/1: Action query time = 4.480 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5797 +t=26: Selected seed 195 with value = 0.5797 +Query 1/1: Action query time = 5.030 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6767 +t=42: Selected seed 195 with value = 0.6767 +Query 1/1: Action query time = 5.288 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7892 +t=58: Selected seed 195 with value = 0.7892 +Query 1/1: Action query time = 5.111 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9061 +t=74: Selected seed 195 with value = 0.9061 +Query 1/1: Action query time = 4.379 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s08/2026_08_02-08_25_33--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 5.048 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4929 +t=10: Selected seed 195 with value = 0.4929 +Query 1/1: Action query time = 5.055 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5782 +t=26: Selected seed 195 with value = 0.5782 +Query 1/1: Action query time = 4.909 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6842 +t=42: Selected seed 195 with value = 0.6842 +Query 1/1: Action query time = 5.053 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7801 +t=58: Selected seed 195 with value = 0.7801 +Query 1/1: Action query time = 4.540 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9063 +t=74: Selected seed 195 with value = 0.9063 +Query 1/1: Action query time = 3.391 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9950 +t=90: Selected seed 195 with value = 0.9950 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s08/2026_08_02-08_25_33--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 4.953 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4846 +t=10: Selected seed 195 with value = 0.4846 +Query 1/1: Action query time = 4.414 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5596 +t=26: Selected seed 195 with value = 0.5596 +Query 1/1: Action query time = 4.584 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6746 +t=42: Selected seed 195 with value = 0.6746 +Query 1/1: Action query time = 4.348 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7470 +t=58: Selected seed 195 with value = 0.7470 +Query 1/1: Action query time = 4.160 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8703 +t=74: Selected seed 195 with value = 0.8703 +Query 1/1: Action query time = 2.670 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9944 +t=90: Selected seed 195 with value = 0.9944 +Query 1/1: Action query time = 1.213 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s08/2026_08_02-08_25_33--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_34--fcv2_350_t1_s09.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_34--fcv2_350_t1_s09.txt new file mode 100644 index 0000000000000000000000000000000000000000..79642896b57e530d4abe6a683fc9ebd0e4beb1e6 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_34--fcv2_350_t1_s09.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t1_s09', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='9,25,41', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 3.628 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5144 +t=10: Selected seed 195 with value = 0.5144 +Query 1/1: Action query time = 4.940 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6015 +t=26: Selected seed 195 with value = 0.6015 +Query 1/1: Action query time = 4.743 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7327 +t=42: Selected seed 195 with value = 0.7327 +Query 1/1: Action query time = 5.080 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8427 +t=58: Selected seed 195 with value = 0.8427 +Query 1/1: Action query time = 4.871 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9762 +t=74: Selected seed 195 with value = 0.9762 +Query 1/1: Action query time = 5.041 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s09/2026_08_02-08_25_34--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 1.893 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5099 +t=10: Selected seed 195 with value = 0.5099 +Query 1/1: Action query time = 3.333 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5851 +t=26: Selected seed 195 with value = 0.5851 +Query 1/1: Action query time = 5.280 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7142 +t=42: Selected seed 195 with value = 0.7142 +Query 1/1: Action query time = 5.274 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8200 +t=58: Selected seed 195 with value = 0.8200 +Query 1/1: Action query time = 5.179 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9620 +t=74: Selected seed 195 with value = 0.9620 +Query 1/1: Action query time = 4.795 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s09/2026_08_02-08_25_34--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 4.117 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4795 +t=10: Selected seed 195 with value = 0.4795 +Query 1/1: Action query time = 1.662 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5819 +t=26: Selected seed 195 with value = 0.5819 +Query 1/1: Action query time = 4.364 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6990 +t=42: Selected seed 195 with value = 0.6990 +Query 1/1: Action query time = 4.580 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7884 +t=58: Selected seed 195 with value = 0.7884 +Query 1/1: Action query time = 5.163 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9511 +t=74: Selected seed 195 with value = 0.9511 +Query 1/1: Action query time = 4.975 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s09/2026_08_02-08_25_34--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_34--fcv2_350_t1_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_34--fcv2_350_t1_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..6d92c40216f79794c1ded3fb42a16148da3d87a0 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_25_34--fcv2_350_t1_s14.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t1_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 3.760 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5252 +t=10: Selected seed 195 with value = 0.5252 +Query 1/1: Action query time = 5.062 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6236 +t=26: Selected seed 195 with value = 0.6236 +Query 1/1: Action query time = 5.091 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7466 +t=42: Selected seed 195 with value = 0.7466 +Query 1/1: Action query time = 4.626 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8759 +t=58: Selected seed 195 with value = 0.8759 +Query 1/1: Action query time = 4.871 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.197 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s14/2026_08_02-08_25_34--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 1.470 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4924 +t=10: Selected seed 195 with value = 0.4924 +Query 1/1: Action query time = 3.335 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5753 +t=26: Selected seed 195 with value = 0.5753 +Query 1/1: Action query time = 5.348 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6902 +t=42: Selected seed 195 with value = 0.6902 +Query 1/1: Action query time = 5.250 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7965 +t=58: Selected seed 195 with value = 0.7965 +Query 1/1: Action query time = 4.944 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8619 +t=74: Selected seed 195 with value = 0.8619 +Query 1/1: Action query time = 4.910 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9675 +t=90: Selected seed 195 with value = 0.9675 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s14/2026_08_02-08_25_34--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 3.466 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4836 +t=10: Selected seed 195 with value = 0.4836 +Query 1/1: Action query time = 2.111 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5839 +t=26: Selected seed 195 with value = 0.5839 +Query 1/1: Action query time = 5.308 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6922 +t=42: Selected seed 195 with value = 0.6922 +Query 1/1: Action query time = 5.454 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7956 +t=58: Selected seed 195 with value = 0.7956 +Query 1/1: Action query time = 5.543 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9090 +t=74: Selected seed 195 with value = 0.9090 +Query 1/1: Action query time = 4.790 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t1_s14/2026_08_02-08_25_34--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_28_02--fcv2_350_t2_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_28_02--fcv2_350_t2_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..0ea30230a2f647c38183ef7da4164dd4ed33396c --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_28_02--fcv2_350_t2_s02.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t2_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.549 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4749 +t=10: Selected seed 195 with value = 0.4749 +Query 1/1: Action query time = 5.055 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5468 +t=26: Selected seed 195 with value = 0.5468 +Query 1/1: Action query time = 5.179 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6457 +t=42: Selected seed 195 with value = 0.6457 +Query 1/1: Action query time = 4.074 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8364 +t=58: Selected seed 195 with value = 0.8364 +Query 1/1: Action query time = 4.409 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9907 +t=74: Selected seed 195 with value = 0.9907 +Query 1/1: Action query time = 4.278 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t2_s02/2026_08_02-08_28_02--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.473 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4936 +t=10: Selected seed 195 with value = 0.4936 +Query 1/1: Action query time = 5.236 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5836 +t=26: Selected seed 195 with value = 0.5836 +Query 1/1: Action query time = 4.947 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7007 +t=42: Selected seed 195 with value = 0.7007 +Query 1/1: Action query time = 2.751 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8447 +t=58: Selected seed 195 with value = 0.8447 +Query 1/1: Action query time = 4.768 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=74: Selected seed 195 with value = 0.9946 +Query 1/1: Action query time = 4.225 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t2_s02/2026_08_02-08_28_02--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.437 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4169 +t=10: Selected seed 195 with value = 0.4169 +Query 1/1: Action query time = 4.636 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5661 +t=26: Selected seed 195 with value = 0.5661 +Query 1/1: Action query time = 5.181 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6984 +t=42: Selected seed 195 with value = 0.6984 +Query 1/1: Action query time = 5.040 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7930 +t=58: Selected seed 195 with value = 0.7930 +Query 1/1: Action query time = 2.269 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9575 +t=74: Selected seed 195 with value = 0.9575 +Query 1/1: Action query time = 3.002 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t2_s02/2026_08_02-08_28_02--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_28_03--fcv2_350_t2_s09.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_28_03--fcv2_350_t2_s09.txt new file mode 100644 index 0000000000000000000000000000000000000000..f96ce435cbe249c775b494b261e5d18db4f58842 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_28_03--fcv2_350_t2_s09.txt @@ -0,0 +1,101 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t2_s09', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='9,25,41', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.662 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5093 +t=10: Selected seed 195 with value = 0.5093 +Query 1/1: Action query time = 5.770 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5936 +t=26: Selected seed 195 with value = 0.5936 +Query 1/1: Action query time = 4.896 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7230 +t=42: Selected seed 195 with value = 0.7230 +Query 1/1: Action query time = 4.924 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8297 +t=58: Selected seed 195 with value = 0.8297 +Query 1/1: Action query time = 4.521 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.305 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t2_s09/2026_08_02-08_28_03--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.825 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5126 +t=10: Selected seed 195 with value = 0.5126 +Query 1/1: Action query time = 3.516 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5712 +t=26: Selected seed 195 with value = 0.5712 +Query 1/1: Action query time = 6.124 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7004 +t=42: Selected seed 195 with value = 0.7004 +Query 1/1: Action query time = 5.888 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8268 +t=58: Selected seed 195 with value = 0.8268 +Query 1/1: Action query time = 5.164 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9638 +t=74: Selected seed 195 with value = 0.9638 +Query 1/1: Action query time = 4.549 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t2_s09/2026_08_02-08_28_03--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.066 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4771 +t=10: Selected seed 195 with value = 0.4771 +Query 1/1: Action query time = 4.601 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5981 +t=26: Selected seed 195 with value = 0.5981 +Query 1/1: Action query time = 4.661 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7201 +t=42: Selected seed 195 with value = 0.7201 +Query 1/1: Action query time = 5.025 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8899 +t=58: Selected seed 195 with value = 0.8899 +Query 1/1: Action query time = 5.014 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t2_s09/2026_08_02-08_28_03--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_28_04--fcv2_350_t2_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_28_04--fcv2_350_t2_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..759a4654d24327ef5456e13fcfc7a8c5b2f4e312 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_28_04--fcv2_350_t2_s14.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t2_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.510 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4969 +t=10: Selected seed 195 with value = 0.4969 +Query 1/1: Action query time = 4.570 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5952 +t=26: Selected seed 195 with value = 0.5952 +Query 1/1: Action query time = 4.784 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6978 +t=42: Selected seed 195 with value = 0.6978 +Query 1/1: Action query time = 4.964 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7939 +t=58: Selected seed 195 with value = 0.7939 +Query 1/1: Action query time = 5.394 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9736 +t=74: Selected seed 195 with value = 0.9736 +Query 1/1: Action query time = 3.492 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t2_s14/2026_08_02-08_28_04--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.787 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5060 +t=10: Selected seed 195 with value = 0.5060 +Query 1/1: Action query time = 4.017 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5709 +t=26: Selected seed 195 with value = 0.5709 +Query 1/1: Action query time = 4.340 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6740 +t=42: Selected seed 195 with value = 0.6740 +Query 1/1: Action query time = 5.154 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8646 +t=58: Selected seed 195 with value = 0.8646 +Query 1/1: Action query time = 5.025 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9974 +t=74: Selected seed 195 with value = 0.9974 +Query 1/1: Action query time = 4.266 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t2_s14/2026_08_02-08_28_04--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.855 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4997 +t=10: Selected seed 195 with value = 0.4997 +Query 1/1: Action query time = 4.686 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6054 +t=26: Selected seed 195 with value = 0.6054 +Query 1/1: Action query time = 5.045 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6850 +t=42: Selected seed 195 with value = 0.6850 +Query 1/1: Action query time = 5.232 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8120 +t=58: Selected seed 195 with value = 0.8120 +Query 1/1: Action query time = 4.791 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=74: Selected seed 195 with value = 0.9886 +Query 1/1: Action query time = 3.559 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t2_s14/2026_08_02-08_28_04--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_02--fcv2_350_t3_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_02--fcv2_350_t3_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..b4e490f711ba9b6b7870ec3ef2901eb84bf4c990 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_02--fcv2_350_t3_s02.txt @@ -0,0 +1,173 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t3_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 3.034 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3847 +t=10: Selected seed 195 with value = 0.3847 +Query 1/1: Action query time = 4.703 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4584 +t=26: Selected seed 195 with value = 0.4584 +Query 1/1: Action query time = 4.445 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5458 +t=42: Selected seed 195 with value = 0.5458 +Query 1/1: Action query time = 5.416 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6178 +t=58: Selected seed 195 with value = 0.6178 +Query 1/1: Action query time = 5.484 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7986 +t=74: Selected seed 195 with value = 0.7986 +Query 1/1: Action query time = 5.286 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9364 +t=90: Selected seed 195 with value = 0.9364 +Query 1/1: Action query time = 4.534 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=106: Selected seed 195 with value = 0.9983 +Query 1/1: Action query time = 4.634 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6256 +t=122: Selected seed 195 with value = 0.6256 +Query 1/1: Action query time = 4.908 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6739 +t=138: Selected seed 195 with value = 0.6739 +Query 1/1: Action query time = 5.105 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7965 +t=154: Selected seed 195 with value = 0.7965 +Query 1/1: Action query time = 4.838 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9120 +t=170: Selected seed 195 with value = 0.9120 +Query 1/1: Action query time = 4.721 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s02/2026_08_02-08_31_02--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.895 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3822 +t=10: Selected seed 195 with value = 0.3822 +Query 1/1: Action query time = 4.884 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4549 +t=26: Selected seed 195 with value = 0.4549 +Query 1/1: Action query time = 3.809 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5435 +t=42: Selected seed 195 with value = 0.5435 +Query 1/1: Action query time = 4.049 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6062 +t=58: Selected seed 195 with value = 0.6062 +Query 1/1: Action query time = 5.189 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7314 +t=74: Selected seed 195 with value = 0.7314 +Query 1/1: Action query time = 4.217 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8814 +t=90: Selected seed 195 with value = 0.8814 +Query 1/1: Action query time = 4.909 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.371 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6455 +t=122: Selected seed 195 with value = 0.6455 +Query 1/1: Action query time = 4.624 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6915 +t=138: Selected seed 195 with value = 0.6915 +Query 1/1: Action query time = 4.825 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8038 +t=154: Selected seed 195 with value = 0.8038 +Query 1/1: Action query time = 4.795 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9401 +t=170: Selected seed 195 with value = 0.9401 +Query 1/1: Action query time = 4.874 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s02/2026_08_02-08_31_02--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.362 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4142 +t=10: Selected seed 195 with value = 0.4142 +Query 1/1: Action query time = 4.288 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4621 +t=26: Selected seed 195 with value = 0.4621 +Query 1/1: Action query time = 5.122 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5647 +t=42: Selected seed 195 with value = 0.5647 +Query 1/1: Action query time = 4.408 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6667 +t=58: Selected seed 195 with value = 0.6667 +Query 1/1: Action query time = 5.369 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7339 +t=74: Selected seed 195 with value = 0.7339 +Query 1/1: Action query time = 5.180 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9466 +t=90: Selected seed 195 with value = 0.9466 +Query 1/1: Action query time = 4.520 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9772 +t=106: Selected seed 195 with value = 0.9772 +Query 1/1: Action query time = 4.354 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6520 +t=122: Selected seed 195 with value = 0.6520 +Query 1/1: Action query time = 4.317 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7646 +t=138: Selected seed 195 with value = 0.7646 +Query 1/1: Action query time = 3.689 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8874 +t=154: Selected seed 195 with value = 0.8874 +Query 1/1: Action query time = 4.654 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s02/2026_08_02-08_31_02--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_03--fcv2_350_t3_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_03--fcv2_350_t3_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..839686ce6cb4c1fc2405a1286d6bf2b60b86d6de --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_03--fcv2_350_t3_s06.txt @@ -0,0 +1,209 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t3_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 6.472 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3904 +t=10: Selected seed 195 with value = 0.3904 +Query 1/1: Action query time = 4.588 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4590 +t=26: Selected seed 195 with value = 0.4590 +Query 1/1: Action query time = 4.570 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5036 +t=42: Selected seed 195 with value = 0.5036 +Query 1/1: Action query time = 4.960 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6050 +t=58: Selected seed 195 with value = 0.6050 +Query 1/1: Action query time = 5.339 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7021 +t=74: Selected seed 195 with value = 0.7021 +Query 1/1: Action query time = 5.462 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9404 +t=90: Selected seed 195 with value = 0.9404 +Query 1/1: Action query time = 5.288 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9730 +t=106: Selected seed 195 with value = 0.9730 +Query 1/1: Action query time = 5.085 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6683 +t=122: Selected seed 195 with value = 0.6683 +Query 1/1: Action query time = 4.908 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7678 +t=138: Selected seed 195 with value = 0.7678 +Query 1/1: Action query time = 4.782 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9237 +t=154: Selected seed 195 with value = 0.9237 +Query 1/1: Action query time = 3.713 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9813 +t=170: Selected seed 195 with value = 0.9813 +Query 1/1: Action query time = 2.458 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9604 +t=186: Selected seed 195 with value = 0.9604 +Query 1/1: Action query time = 4.366 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6901 +t=202: Selected seed 195 with value = 0.6901 +Query 1/1: Action query time = 4.729 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6339 +t=218: Selected seed 195 with value = 0.6339 +Query 1/1: Action query time = 4.226 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9068 +t=234: Selected seed 195 with value = 0.9068 +Query 1/1: Action query time = 5.081 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7838 +t=250: Selected seed 195 with value = 0.7838 +Query 1/1: Action query time = 5.691 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8956 +t=266: Selected seed 195 with value = 0.8956 +Query 1/1: Action query time = 4.536 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8523 +t=282: Selected seed 195 with value = 0.8523 +Query 1/1: Action query time = 4.951 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8964 +t=298: Selected seed 195 with value = 0.8964 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s06/2026_08_02-08_31_03--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 5.393 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3547 +t=10: Selected seed 195 with value = 0.3547 +Query 1/1: Action query time = 4.945 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4617 +t=26: Selected seed 195 with value = 0.4617 +Query 1/1: Action query time = 4.788 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5755 +t=42: Selected seed 195 with value = 0.5755 +Query 1/1: Action query time = 4.433 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6611 +t=58: Selected seed 195 with value = 0.6611 +Query 1/1: Action query time = 3.659 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7876 +t=74: Selected seed 195 with value = 0.7876 +Query 1/1: Action query time = 4.830 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8985 +t=90: Selected seed 195 with value = 0.8985 +Query 1/1: Action query time = 3.853 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.743 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6401 +t=122: Selected seed 195 with value = 0.6401 +Query 1/1: Action query time = 4.480 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7572 +t=138: Selected seed 195 with value = 0.7572 +Query 1/1: Action query time = 4.846 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8170 +t=154: Selected seed 195 with value = 0.8170 +Query 1/1: Action query time = 4.841 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9534 +t=170: Selected seed 195 with value = 0.9534 +Query 1/1: Action query time = 4.711 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s06/2026_08_02-08_31_03--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 5.318 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3593 +t=10: Selected seed 195 with value = 0.3593 +Query 1/1: Action query time = 5.035 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4586 +t=26: Selected seed 195 with value = 0.4586 +Query 1/1: Action query time = 4.498 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5701 +t=42: Selected seed 195 with value = 0.5701 +Query 1/1: Action query time = 4.643 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6369 +t=58: Selected seed 195 with value = 0.6369 +Query 1/1: Action query time = 3.780 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7437 +t=74: Selected seed 195 with value = 0.7437 +Query 1/1: Action query time = 3.179 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8736 +t=90: Selected seed 195 with value = 0.8736 +Query 1/1: Action query time = 3.607 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9972 +t=106: Selected seed 195 with value = 0.9972 +Query 1/1: Action query time = 3.479 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6002 +t=122: Selected seed 195 with value = 0.6002 +Query 1/1: Action query time = 4.087 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6839 +t=138: Selected seed 195 with value = 0.6839 +Query 1/1: Action query time = 3.895 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7793 +t=154: Selected seed 195 with value = 0.7793 +Query 1/1: Action query time = 3.868 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8473 +t=170: Selected seed 195 with value = 0.8473 +Query 1/1: Action query time = 3.196 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9691 +t=186: Selected seed 195 with value = 0.9691 +Query 1/1: Action query time = 3.020 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s06/2026_08_02-08_31_03--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_03--fcv2_350_t3_s09.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_03--fcv2_350_t3_s09.txt new file mode 100644 index 0000000000000000000000000000000000000000..dd01514094fc130561ecc77f8844507527d21da0 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_03--fcv2_350_t3_s09.txt @@ -0,0 +1,209 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t3_s09', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='9,25,41', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 6.450 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4009 +t=10: Selected seed 195 with value = 0.4009 +Query 1/1: Action query time = 5.750 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4956 +t=26: Selected seed 195 with value = 0.4956 +Query 1/1: Action query time = 5.145 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5826 +t=42: Selected seed 195 with value = 0.5826 +Query 1/1: Action query time = 4.931 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6564 +t=58: Selected seed 195 with value = 0.6564 +Query 1/1: Action query time = 4.772 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7567 +t=74: Selected seed 195 with value = 0.7567 +Query 1/1: Action query time = 4.928 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8995 +t=90: Selected seed 195 with value = 0.8995 +Query 1/1: Action query time = 4.992 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.739 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6722 +t=122: Selected seed 195 with value = 0.6722 +Query 1/1: Action query time = 4.943 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7912 +t=138: Selected seed 195 with value = 0.7912 +Query 1/1: Action query time = 4.869 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8180 +t=154: Selected seed 195 with value = 0.8180 +Query 1/1: Action query time = 3.595 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9768 +t=170: Selected seed 195 with value = 0.9768 +Query 1/1: Action query time = 4.467 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9240 +t=186: Selected seed 195 with value = 0.9240 +Query 1/1: Action query time = 3.949 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9573 +t=202: Selected seed 195 with value = 0.9573 +Query 1/1: Action query time = 4.785 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9045 +t=218: Selected seed 195 with value = 0.9045 +Query 1/1: Action query time = 4.657 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8208 +t=234: Selected seed 195 with value = 0.8208 +Query 1/1: Action query time = 5.083 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9037 +t=250: Selected seed 195 with value = 0.9037 +Query 1/1: Action query time = 5.083 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8525 +t=266: Selected seed 195 with value = 0.8525 +Query 1/1: Action query time = 5.246 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9044 +t=282: Selected seed 195 with value = 0.9044 +Query 1/1: Action query time = 5.125 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9031 +t=298: Selected seed 195 with value = 0.9031 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s09/2026_08_02-08_31_03--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 5.152 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3755 +t=10: Selected seed 195 with value = 0.3755 +Query 1/1: Action query time = 4.796 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4905 +t=26: Selected seed 195 with value = 0.4905 +Query 1/1: Action query time = 4.145 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5546 +t=42: Selected seed 195 with value = 0.5546 +Query 1/1: Action query time = 4.773 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6171 +t=58: Selected seed 195 with value = 0.6171 +Query 1/1: Action query time = 3.855 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6994 +t=74: Selected seed 195 with value = 0.6994 +Query 1/1: Action query time = 5.266 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8638 +t=90: Selected seed 195 with value = 0.8638 +Query 1/1: Action query time = 4.192 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=106: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 4.631 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6668 +t=122: Selected seed 195 with value = 0.6668 +Query 1/1: Action query time = 4.837 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7676 +t=138: Selected seed 195 with value = 0.7676 +Query 1/1: Action query time = 4.684 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8601 +t=154: Selected seed 195 with value = 0.8601 +Query 1/1: Action query time = 4.935 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s09/2026_08_02-08_31_03--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.634 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3978 +t=10: Selected seed 195 with value = 0.3978 +Query 1/1: Action query time = 4.783 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4672 +t=26: Selected seed 195 with value = 0.4672 +Query 1/1: Action query time = 5.007 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5493 +t=42: Selected seed 195 with value = 0.5493 +Query 1/1: Action query time = 4.846 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6512 +t=58: Selected seed 195 with value = 0.6512 +Query 1/1: Action query time = 4.189 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7533 +t=74: Selected seed 195 with value = 0.7533 +Query 1/1: Action query time = 3.951 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9055 +t=90: Selected seed 195 with value = 0.9055 +Query 1/1: Action query time = 3.840 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.026 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8343 +t=122: Selected seed 195 with value = 0.8343 +Query 1/1: Action query time = 3.953 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6930 +t=138: Selected seed 195 with value = 0.6930 +Query 1/1: Action query time = 3.178 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6340 +t=154: Selected seed 195 with value = 0.6340 +Query 1/1: Action query time = 2.383 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7059 +t=170: Selected seed 195 with value = 0.7059 +Query 1/1: Action query time = 2.404 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8132 +t=186: Selected seed 195 with value = 0.8132 +Query 1/1: Action query time = 2.238 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9459 +t=202: Selected seed 195 with value = 0.9459 +Query 1/1: Action query time = 2.267 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9829 +t=218: Selected seed 195 with value = 0.9829 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s09/2026_08_02-08_31_03--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_03--fcv2_350_t3_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_03--fcv2_350_t3_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..fddfcb94ae947a3fb0f772b35ffdf62d245c257c --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_31_03--fcv2_350_t3_s11.txt @@ -0,0 +1,197 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t3_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 4.488 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3840 +t=10: Selected seed 195 with value = 0.3840 +Query 1/1: Action query time = 4.612 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4663 +t=26: Selected seed 195 with value = 0.4663 +Query 1/1: Action query time = 5.305 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5242 +t=42: Selected seed 195 with value = 0.5242 +Query 1/1: Action query time = 4.964 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5931 +t=58: Selected seed 195 with value = 0.5931 +Query 1/1: Action query time = 5.373 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6708 +t=74: Selected seed 195 with value = 0.6708 +Query 1/1: Action query time = 4.953 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8755 +t=90: Selected seed 195 with value = 0.8755 +Query 1/1: Action query time = 5.344 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9842 +t=106: Selected seed 195 with value = 0.9842 +Query 1/1: Action query time = 5.075 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7055 +t=122: Selected seed 195 with value = 0.7055 +Query 1/1: Action query time = 4.282 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7720 +t=138: Selected seed 195 with value = 0.7720 +Query 1/1: Action query time = 4.822 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9445 +t=154: Selected seed 195 with value = 0.9445 +Query 1/1: Action query time = 5.139 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s11/2026_08_02-08_31_03--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 3.693 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3557 +t=10: Selected seed 195 with value = 0.3557 +Query 1/1: Action query time = 4.314 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4775 +t=26: Selected seed 195 with value = 0.4775 +Query 1/1: Action query time = 5.066 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5518 +t=42: Selected seed 195 with value = 0.5518 +Query 1/1: Action query time = 5.311 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6121 +t=58: Selected seed 195 with value = 0.6121 +Query 1/1: Action query time = 5.278 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6755 +t=74: Selected seed 195 with value = 0.6755 +Query 1/1: Action query time = 4.716 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8532 +t=90: Selected seed 195 with value = 0.8532 +Query 1/1: Action query time = 4.866 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9899 +t=106: Selected seed 195 with value = 0.9899 +Query 1/1: Action query time = 4.386 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6341 +t=122: Selected seed 195 with value = 0.6341 +Query 1/1: Action query time = 2.206 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7332 +t=138: Selected seed 195 with value = 0.7332 +Query 1/1: Action query time = 3.221 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7381 +t=154: Selected seed 195 with value = 0.7381 +Query 1/1: Action query time = 4.855 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8348 +t=170: Selected seed 195 with value = 0.8348 +Query 1/1: Action query time = 4.812 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7846 +t=186: Selected seed 195 with value = 0.7846 +Query 1/1: Action query time = 4.158 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7697 +t=202: Selected seed 195 with value = 0.7697 +Query 1/1: Action query time = 4.983 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8545 +t=218: Selected seed 195 with value = 0.8545 +Query 1/1: Action query time = 3.656 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9655 +t=234: Selected seed 195 with value = 0.9655 +Query 1/1: Action query time = 5.521 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s11/2026_08_02-08_31_03--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 5.857 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3955 +t=10: Selected seed 195 with value = 0.3955 +Query 1/1: Action query time = 4.588 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4738 +t=26: Selected seed 195 with value = 0.4738 +Query 1/1: Action query time = 4.727 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5426 +t=42: Selected seed 195 with value = 0.5426 +Query 1/1: Action query time = 4.886 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5827 +t=58: Selected seed 195 with value = 0.5827 +Query 1/1: Action query time = 3.908 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6738 +t=74: Selected seed 195 with value = 0.6738 +Query 1/1: Action query time = 4.796 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8502 +t=90: Selected seed 195 with value = 0.8502 +Query 1/1: Action query time = 5.080 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=106: Selected seed 195 with value = 0.9867 +Query 1/1: Action query time = 5.208 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7093 +t=122: Selected seed 195 with value = 0.7093 +Query 1/1: Action query time = 5.097 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8151 +t=138: Selected seed 195 with value = 0.8151 +Query 1/1: Action query time = 4.224 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5642 +t=154: Selected seed 195 with value = 0.5642 +Query 1/1: Action query time = 3.588 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6952 +t=170: Selected seed 195 with value = 0.6952 +Query 1/1: Action query time = 3.851 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7384 +t=186: Selected seed 195 with value = 0.7384 +Query 1/1: Action query time = 3.186 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9476 +t=202: Selected seed 195 with value = 0.9476 +Query 1/1: Action query time = 4.173 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=218: Selected seed 195 with value = 0.9985 +Saved rollout MP4 at path ./rollouts/fcv2_350_t3_s11/2026_08_02-08_31_03--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_39_10--fcv2_350_t5_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_39_10--fcv2_350_t5_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..d04516e109b583e24ee411153151f346c28422f8 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_39_10--fcv2_350_t5_s01.txt @@ -0,0 +1,180 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t5_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 6.079 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2745 +t=10: Selected seed 195 with value = 0.2745 +Query 1/1: Action query time = 5.530 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3991 +t=26: Selected seed 195 with value = 0.3991 +Query 1/1: Action query time = 5.080 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4961 +t=42: Selected seed 195 with value = 0.4961 +Query 1/1: Action query time = 4.244 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5379 +t=58: Selected seed 195 with value = 0.5379 +Query 1/1: Action query time = 4.932 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6105 +t=74: Selected seed 195 with value = 0.6105 +Query 1/1: Action query time = 5.277 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7176 +t=90: Selected seed 195 with value = 0.7176 +Query 1/1: Action query time = 4.685 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8174 +t=106: Selected seed 195 with value = 0.8174 +Query 1/1: Action query time = 2.931 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8870 +t=122: Selected seed 195 with value = 0.8870 +Query 1/1: Action query time = 3.572 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t5_s01/2026_08_02-08_39_10--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 3.736 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3105 +t=10: Selected seed 195 with value = 0.3105 +Query 1/1: Action query time = 5.043 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4146 +t=26: Selected seed 195 with value = 0.4146 +Query 1/1: Action query time = 5.431 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4979 +t=42: Selected seed 195 with value = 0.4979 +Query 1/1: Action query time = 5.104 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5146 +t=58: Selected seed 195 with value = 0.5146 +Query 1/1: Action query time = 5.000 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5546 +t=74: Selected seed 195 with value = 0.5546 +Query 1/1: Action query time = 4.989 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6922 +t=90: Selected seed 195 with value = 0.6922 +Query 1/1: Action query time = 3.804 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7725 +t=106: Selected seed 195 with value = 0.7725 +Query 1/1: Action query time = 3.461 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9341 +t=122: Selected seed 195 with value = 0.9341 +Query 1/1: Action query time = 4.735 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t5_s01/2026_08_02-08_39_10--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 4.456 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2431 +t=10: Selected seed 195 with value = 0.2431 +Query 1/1: Action query time = 4.479 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4393 +t=26: Selected seed 195 with value = 0.4393 +Query 1/1: Action query time = 5.004 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5097 +t=42: Selected seed 195 with value = 0.5097 +Query 1/1: Action query time = 4.806 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5663 +t=58: Selected seed 195 with value = 0.5663 +Query 1/1: Action query time = 4.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6759 +t=74: Selected seed 195 with value = 0.6759 +Query 1/1: Action query time = 4.824 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7735 +t=90: Selected seed 195 with value = 0.7735 +Query 1/1: Action query time = 3.597 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8641 +t=106: Selected seed 195 with value = 0.8641 +Query 1/1: Action query time = 2.946 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9671 +t=122: Selected seed 195 with value = 0.9671 +Saved rollout MP4 at path ./rollouts/fcv2_350_t5_s01/2026_08_02-08_39_10--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 1.341 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2286 +t=10: Selected seed 195 with value = 0.2286 +Query 1/1: Action query time = 1.474 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3762 +t=26: Selected seed 195 with value = 0.3762 +Query 1/1: Action query time = 1.478 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4416 +t=42: Selected seed 195 with value = 0.4416 +Query 1/1: Action query time = 1.523 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4891 +t=58: Selected seed 195 with value = 0.4891 +Query 1/1: Action query time = 1.380 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5824 +t=74: Selected seed 195 with value = 0.5824 +Query 1/1: Action query time = 1.366 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6875 +t=90: Selected seed 195 with value = 0.6875 +Query 1/1: Action query time = 0.959 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7909 +t=106: Selected seed 195 with value = 0.7909 +Query 1/1: Action query time = 0.949 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9569 +t=122: Selected seed 195 with value = 0.9569 +Query 1/1: Action query time = 0.962 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t5_s01/2026_08_02-08_39_10--with_future_img--episode=4--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 4 +Total successes: 4 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_39_10--fcv2_350_t5_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_39_10--fcv2_350_t5_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..ccb27e5a398588b8b5b5ec1e54f5fa4d96c9ada2 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_39_10--fcv2_350_t5_s02.txt @@ -0,0 +1,145 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t5_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 6.361 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3193 +t=10: Selected seed 195 with value = 0.3193 +Query 1/1: Action query time = 5.336 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3772 +t=26: Selected seed 195 with value = 0.3772 +Query 1/1: Action query time = 5.443 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4082 +t=42: Selected seed 195 with value = 0.4082 +Query 1/1: Action query time = 4.431 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4622 +t=58: Selected seed 195 with value = 0.4622 +Query 1/1: Action query time = 4.778 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5313 +t=74: Selected seed 195 with value = 0.5313 +Query 1/1: Action query time = 4.718 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6139 +t=90: Selected seed 195 with value = 0.6139 +Query 1/1: Action query time = 4.711 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7088 +t=106: Selected seed 195 with value = 0.7088 +Query 1/1: Action query time = 2.857 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8115 +t=122: Selected seed 195 with value = 0.8115 +Query 1/1: Action query time = 3.714 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9975 +t=138: Selected seed 195 with value = 0.9975 +Query 1/1: Action query time = 3.757 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=154: Selected seed 195 with value = 0.9999 +Saved rollout MP4 at path ./rollouts/fcv2_350_t5_s02/2026_08_02-08_39_10--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 5.453 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2967 +t=10: Selected seed 195 with value = 0.2967 +Query 1/1: Action query time = 5.315 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3755 +t=26: Selected seed 195 with value = 0.3755 +Query 1/1: Action query time = 5.263 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3475 +t=42: Selected seed 195 with value = 0.3475 +Query 1/1: Action query time = 4.953 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4786 +t=58: Selected seed 195 with value = 0.4786 +Query 1/1: Action query time = 4.899 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5864 +t=74: Selected seed 195 with value = 0.5864 +Query 1/1: Action query time = 3.251 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6662 +t=90: Selected seed 195 with value = 0.6662 +Query 1/1: Action query time = 3.919 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7470 +t=106: Selected seed 195 with value = 0.7470 +Query 1/1: Action query time = 4.742 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8840 +t=122: Selected seed 195 with value = 0.8840 +Query 1/1: Action query time = 4.934 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=138: Selected seed 195 with value = 0.9997 +Saved rollout MP4 at path ./rollouts/fcv2_350_t5_s02/2026_08_02-08_39_10--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 5.115 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2441 +t=10: Selected seed 195 with value = 0.2441 +Query 1/1: Action query time = 4.564 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3747 +t=26: Selected seed 195 with value = 0.3747 +Query 1/1: Action query time = 4.575 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4487 +t=42: Selected seed 195 with value = 0.4487 +Query 1/1: Action query time = 5.178 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4937 +t=58: Selected seed 195 with value = 0.4937 +Query 1/1: Action query time = 4.314 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5759 +t=74: Selected seed 195 with value = 0.5759 +Query 1/1: Action query time = 2.183 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6371 +t=90: Selected seed 195 with value = 0.6371 +Query 1/1: Action query time = 2.491 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7285 +t=106: Selected seed 195 with value = 0.7285 +Query 1/1: Action query time = 1.785 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9240 +t=122: Selected seed 195 with value = 0.9240 +Query 1/1: Action query time = 1.366 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=138: Selected seed 195 with value = 0.9960 +Saved rollout MP4 at path ./rollouts/fcv2_350_t5_s02/2026_08_02-08_39_10--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_39_11--fcv2_350_t5_s13.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_39_11--fcv2_350_t5_s13.txt new file mode 100644 index 0000000000000000000000000000000000000000..7c7eb6c160dfabc63527807c431b10dfae8bca81 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_39_11--fcv2_350_t5_s13.txt @@ -0,0 +1,133 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t5_s13', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='13,29,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 5.435 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2803 +t=10: Selected seed 195 with value = 0.2803 +Query 1/1: Action query time = 5.084 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2684 +t=26: Selected seed 195 with value = 0.2684 +Query 1/1: Action query time = 4.933 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3590 +t=42: Selected seed 195 with value = 0.3590 +Query 1/1: Action query time = 4.448 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4812 +t=58: Selected seed 195 with value = 0.4812 +Query 1/1: Action query time = 5.016 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5748 +t=74: Selected seed 195 with value = 0.5748 +Query 1/1: Action query time = 5.369 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7220 +t=90: Selected seed 195 with value = 0.7220 +Query 1/1: Action query time = 5.557 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8366 +t=106: Selected seed 195 with value = 0.8366 +Query 1/1: Action query time = 4.894 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9695 +t=122: Selected seed 195 with value = 0.9695 +Query 1/1: Action query time = 3.058 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t5_s13/2026_08_02-08_39_11--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.162 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3510 +t=10: Selected seed 195 with value = 0.3510 +Query 1/1: Action query time = 4.796 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4232 +t=26: Selected seed 195 with value = 0.4232 +Query 1/1: Action query time = 5.269 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4737 +t=42: Selected seed 195 with value = 0.4737 +Query 1/1: Action query time = 5.428 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6613 +t=58: Selected seed 195 with value = 0.6613 +Query 1/1: Action query time = 5.295 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7216 +t=74: Selected seed 195 with value = 0.7216 +Query 1/1: Action query time = 4.783 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7791 +t=90: Selected seed 195 with value = 0.7791 +Query 1/1: Action query time = 4.819 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9592 +t=106: Selected seed 195 with value = 0.9592 +Query 1/1: Action query time = 4.675 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=122: Selected seed 195 with value = 0.9914 +Saved rollout MP4 at path ./rollouts/fcv2_350_t5_s13/2026_08_02-08_39_11--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 3.650 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3109 +t=10: Selected seed 195 with value = 0.3109 +Query 1/1: Action query time = 4.820 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4218 +t=26: Selected seed 195 with value = 0.4218 +Query 1/1: Action query time = 4.865 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4842 +t=42: Selected seed 195 with value = 0.4842 +Query 1/1: Action query time = 4.860 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5537 +t=58: Selected seed 195 with value = 0.5537 +Query 1/1: Action query time = 5.251 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6926 +t=74: Selected seed 195 with value = 0.6926 +Query 1/1: Action query time = 4.735 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7702 +t=90: Selected seed 195 with value = 0.7702 +Query 1/1: Action query time = 4.750 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8618 +t=106: Selected seed 195 with value = 0.8618 +Query 1/1: Action query time = 4.585 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9785 +t=122: Selected seed 195 with value = 0.9785 +Saved rollout MP4 at path ./rollouts/fcv2_350_t5_s13/2026_08_02-08_39_11--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_27--fcv2_350_t6_s04.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_27--fcv2_350_t6_s04.txt new file mode 100644 index 0000000000000000000000000000000000000000..175907676ca9a59dad163cc570e1eca78965d673 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_27--fcv2_350_t6_s04.txt @@ -0,0 +1,201 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t6_s04', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='4,20,36', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 1.993 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4171 +t=10: Selected seed 195 with value = 0.4171 +Query 1/1: Action query time = 3.419 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5139 +t=26: Selected seed 195 with value = 0.5139 +Query 1/1: Action query time = 5.116 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5602 +t=42: Selected seed 195 with value = 0.5602 +Query 1/1: Action query time = 4.467 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4966 +t=58: Selected seed 195 with value = 0.4966 +Query 1/1: Action query time = 4.454 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5627 +t=74: Selected seed 195 with value = 0.5627 +Query 1/1: Action query time = 5.147 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6625 +t=90: Selected seed 195 with value = 0.6625 +Query 1/1: Action query time = 5.217 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6429 +t=106: Selected seed 195 with value = 0.6429 +Query 1/1: Action query time = 3.766 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6824 +t=122: Selected seed 195 with value = 0.6824 +Query 1/1: Action query time = 4.637 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8158 +t=138: Selected seed 195 with value = 0.8158 +Query 1/1: Action query time = 5.061 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9215 +t=154: Selected seed 195 with value = 0.9215 +Query 1/1: Action query time = 4.750 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9437 +t=170: Selected seed 195 with value = 0.9437 +Query 1/1: Action query time = 5.073 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.046 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.990 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.380 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.153 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=250: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 4.684 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.612 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7587 +t=282: Selected seed 195 with value = 0.7587 +Query 1/1: Action query time = 5.213 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7866 +t=298: Selected seed 195 with value = 0.7866 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s04/2026_08_02-08_42_27--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 2.720 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4240 +t=10: Selected seed 195 with value = 0.4240 +Query 1/1: Action query time = 2.851 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4986 +t=26: Selected seed 195 with value = 0.4986 +Query 1/1: Action query time = 2.795 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6546 +t=42: Selected seed 195 with value = 0.6546 +Query 1/1: Action query time = 2.994 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7516 +t=58: Selected seed 195 with value = 0.7516 +Query 1/1: Action query time = 3.511 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8361 +t=74: Selected seed 195 with value = 0.8361 +Query 1/1: Action query time = 3.690 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9724 +t=90: Selected seed 195 with value = 0.9724 +Query 1/1: Action query time = 3.632 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.934 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.345 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=138: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 3.139 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9933 +t=154: Selected seed 195 with value = 0.9933 +Query 1/1: Action query time = 3.083 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9949 +t=170: Selected seed 195 with value = 0.9949 +Query 1/1: Action query time = 3.032 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.037 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.047 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8566 +t=218: Selected seed 195 with value = 0.8566 +Query 1/1: Action query time = 1.858 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8700 +t=234: Selected seed 195 with value = 0.8700 +Query 1/1: Action query time = 1.644 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9386 +t=250: Selected seed 195 with value = 0.9386 +Query 1/1: Action query time = 1.542 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s04/2026_08_02-08_42_27--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 1.191 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4529 +t=10: Selected seed 195 with value = 0.4529 +Query 1/1: Action query time = 1.144 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5030 +t=26: Selected seed 195 with value = 0.5030 +Query 1/1: Action query time = 1.138 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6354 +t=42: Selected seed 195 with value = 0.6354 +Query 1/1: Action query time = 1.254 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7435 +t=58: Selected seed 195 with value = 0.7435 +Query 1/1: Action query time = 1.217 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8626 +t=74: Selected seed 195 with value = 0.8626 +Query 1/1: Action query time = 1.218 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=90: Selected seed 195 with value = 0.9958 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s04/2026_08_02-08_42_27--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_28--fcv2_350_t6_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_28--fcv2_350_t6_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..04d900d1c491247841e0bcad9ca65a73f1a84eff --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_28--fcv2_350_t6_s06.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t6_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 4.478 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4854 +t=10: Selected seed 195 with value = 0.4854 +Query 1/1: Action query time = 5.328 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5521 +t=26: Selected seed 195 with value = 0.5521 +Query 1/1: Action query time = 5.561 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6598 +t=42: Selected seed 195 with value = 0.6598 +Query 1/1: Action query time = 4.652 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7410 +t=58: Selected seed 195 with value = 0.7410 +Query 1/1: Action query time = 4.775 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8803 +t=74: Selected seed 195 with value = 0.8803 +Query 1/1: Action query time = 4.748 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9754 +t=90: Selected seed 195 with value = 0.9754 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s06/2026_08_02-08_42_28--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 4.786 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4639 +t=10: Selected seed 195 with value = 0.4639 +Query 1/1: Action query time = 5.102 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4728 +t=26: Selected seed 195 with value = 0.4728 +Query 1/1: Action query time = 5.010 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6108 +t=42: Selected seed 195 with value = 0.6108 +Query 1/1: Action query time = 4.912 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6979 +t=58: Selected seed 195 with value = 0.6979 +Query 1/1: Action query time = 4.781 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8184 +t=74: Selected seed 195 with value = 0.8184 +Query 1/1: Action query time = 4.988 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9590 +t=90: Selected seed 195 with value = 0.9590 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s06/2026_08_02-08_42_28--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.451 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4355 +t=10: Selected seed 195 with value = 0.4355 +Query 1/1: Action query time = 4.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5362 +t=26: Selected seed 195 with value = 0.5362 +Query 1/1: Action query time = 4.844 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6256 +t=42: Selected seed 195 with value = 0.6256 +Query 1/1: Action query time = 4.947 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7380 +t=58: Selected seed 195 with value = 0.7380 +Query 1/1: Action query time = 4.153 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8630 +t=74: Selected seed 195 with value = 0.8630 +Query 1/1: Action query time = 2.858 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s06/2026_08_02-08_42_28--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_28--fcv2_350_t6_s09.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_28--fcv2_350_t6_s09.txt new file mode 100644 index 0000000000000000000000000000000000000000..ea3c05f41a7d3f9a65a826f4d8861a17c33742d7 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_28--fcv2_350_t6_s09.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t6_s09', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='9,25,41', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 6.769 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4228 +t=10: Selected seed 195 with value = 0.4228 +Query 1/1: Action query time = 5.093 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4973 +t=26: Selected seed 195 with value = 0.4973 +Query 1/1: Action query time = 4.602 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5700 +t=42: Selected seed 195 with value = 0.5700 +Query 1/1: Action query time = 4.855 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6621 +t=58: Selected seed 195 with value = 0.6621 +Query 1/1: Action query time = 4.533 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7812 +t=74: Selected seed 195 with value = 0.7812 +Query 1/1: Action query time = 4.393 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9051 +t=90: Selected seed 195 with value = 0.9051 +Query 1/1: Action query time = 2.868 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s09/2026_08_02-08_42_28--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 5.043 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4231 +t=10: Selected seed 195 with value = 0.4231 +Query 1/1: Action query time = 5.257 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3977 +t=26: Selected seed 195 with value = 0.3977 +Query 1/1: Action query time = 5.398 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5671 +t=42: Selected seed 195 with value = 0.5671 +Query 1/1: Action query time = 4.374 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6983 +t=58: Selected seed 195 with value = 0.6983 +Query 1/1: Action query time = 4.577 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8414 +t=74: Selected seed 195 with value = 0.8414 +Query 1/1: Action query time = 4.013 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9940 +t=90: Selected seed 195 with value = 0.9940 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s09/2026_08_02-08_42_28--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 5.424 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4137 +t=10: Selected seed 195 with value = 0.4137 +Query 1/1: Action query time = 5.021 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5209 +t=26: Selected seed 195 with value = 0.5209 +Query 1/1: Action query time = 4.581 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6058 +t=42: Selected seed 195 with value = 0.6058 +Query 1/1: Action query time = 3.827 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6972 +t=58: Selected seed 195 with value = 0.6972 +Query 1/1: Action query time = 2.641 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8383 +t=74: Selected seed 195 with value = 0.8383 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s09/2026_08_02-08_42_28--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_29--fcv2_350_t6_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_29--fcv2_350_t6_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..669d668f7587d21d1471267fb66a9d3db38e8a3c --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_29--fcv2_350_t6_s07.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t6_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 6.655 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4110 +t=10: Selected seed 195 with value = 0.4110 +Query 1/1: Action query time = 4.339 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5271 +t=26: Selected seed 195 with value = 0.5271 +Query 1/1: Action query time = 4.565 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6285 +t=42: Selected seed 195 with value = 0.6285 +Query 1/1: Action query time = 5.276 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6938 +t=58: Selected seed 195 with value = 0.6938 +Query 1/1: Action query time = 5.270 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8213 +t=74: Selected seed 195 with value = 0.8213 +Query 1/1: Action query time = 3.699 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9566 +t=90: Selected seed 195 with value = 0.9566 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s07/2026_08_02-08_42_29--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 4.654 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4535 +t=10: Selected seed 195 with value = 0.4535 +Query 1/1: Action query time = 4.218 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5036 +t=26: Selected seed 195 with value = 0.5036 +Query 1/1: Action query time = 5.029 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6206 +t=42: Selected seed 195 with value = 0.6206 +Query 1/1: Action query time = 5.065 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7326 +t=58: Selected seed 195 with value = 0.7326 +Query 1/1: Action query time = 5.151 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8535 +t=74: Selected seed 195 with value = 0.8535 +Query 1/1: Action query time = 4.327 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9753 +t=90: Selected seed 195 with value = 0.9753 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s07/2026_08_02-08_42_29--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.894 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4701 +t=10: Selected seed 195 with value = 0.4701 +Query 1/1: Action query time = 4.538 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5321 +t=26: Selected seed 195 with value = 0.5321 +Query 1/1: Action query time = 4.827 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6340 +t=42: Selected seed 195 with value = 0.6340 +Query 1/1: Action query time = 4.499 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6536 +t=58: Selected seed 195 with value = 0.6536 +Query 1/1: Action query time = 3.825 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5711 +t=74: Selected seed 195 with value = 0.5711 +Query 1/1: Action query time = 2.945 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6154 +t=90: Selected seed 195 with value = 0.6154 +Query 1/1: Action query time = 2.992 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5968 +t=106: Selected seed 195 with value = 0.5968 +Query 1/1: Action query time = 3.144 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6571 +t=122: Selected seed 195 with value = 0.6571 +Query 1/1: Action query time = 3.219 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7462 +t=138: Selected seed 195 with value = 0.7462 +Query 1/1: Action query time = 3.332 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8007 +t=154: Selected seed 195 with value = 0.8007 +Query 1/1: Action query time = 3.626 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8485 +t=170: Selected seed 195 with value = 0.8485 +Query 1/1: Action query time = 3.494 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9192 +t=186: Selected seed 195 with value = 0.9192 +Query 1/1: Action query time = 3.600 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9807 +t=202: Selected seed 195 with value = 0.9807 +Query 1/1: Action query time = 3.112 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=218: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 3.385 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.109 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=250: Selected seed 195 with value = 0.9997 +Query 1/1: Action query time = 3.079 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.043 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.053 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s07/2026_08_02-08_42_29--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_29--fcv2_350_t6_s15.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_29--fcv2_350_t6_s15.txt new file mode 100644 index 0000000000000000000000000000000000000000..0e3495c09fe6fbdd162fe4661a32e30e732b3262 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_42_29--fcv2_350_t6_s15.txt @@ -0,0 +1,101 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_forget_cl_v2_from350_constlr_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fcv2_350_t6_s15', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='15,31,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 3.316 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4416 +t=10: Selected seed 195 with value = 0.4416 +Query 1/1: Action query time = 5.106 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5364 +t=26: Selected seed 195 with value = 0.5364 +Query 1/1: Action query time = 5.656 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6434 +t=42: Selected seed 195 with value = 0.6434 +Query 1/1: Action query time = 4.830 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7578 +t=58: Selected seed 195 with value = 0.7578 +Query 1/1: Action query time = 4.981 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8964 +t=74: Selected seed 195 with value = 0.8964 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s15/2026_08_02-08_42_29--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 4.465 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4605 +t=10: Selected seed 195 with value = 0.4605 +Query 1/1: Action query time = 2.625 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4924 +t=26: Selected seed 195 with value = 0.4924 +Query 1/1: Action query time = 3.354 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6353 +t=42: Selected seed 195 with value = 0.6353 +Query 1/1: Action query time = 5.148 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7309 +t=58: Selected seed 195 with value = 0.7309 +Query 1/1: Action query time = 5.071 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8461 +t=74: Selected seed 195 with value = 0.8461 +Query 1/1: Action query time = 4.957 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9800 +t=90: Selected seed 195 with value = 0.9800 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s15/2026_08_02-08_42_29--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.689 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4669 +t=10: Selected seed 195 with value = 0.4669 +Query 1/1: Action query time = 4.095 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5011 +t=26: Selected seed 195 with value = 0.5011 +Query 1/1: Action query time = 3.420 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5868 +t=42: Selected seed 195 with value = 0.5868 +Query 1/1: Action query time = 4.861 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7077 +t=58: Selected seed 195 with value = 0.7077 +Query 1/1: Action query time = 4.896 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8437 +t=74: Selected seed 195 with value = 0.8437 +Query 1/1: Action query time = 4.980 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9625 +t=90: Selected seed 195 with value = 0.9625 +Saved rollout MP4 at path ./rollouts/fcv2_350_t6_s15/2026_08_02-08_42_29--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_49_15--ft6100_t0_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_49_15--ft6100_t0_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..ef86c1d57ed52f0a44ed487cfbb7cd78f25d9a54 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_49_15--ft6100_t0_s02.txt @@ -0,0 +1,145 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t0_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.495 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3765 +t=10: Selected seed 195 with value = 0.3765 +Query 1/1: Action query time = 4.550 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3857 +t=26: Selected seed 195 with value = 0.3857 +Query 1/1: Action query time = 6.114 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4568 +t=42: Selected seed 195 with value = 0.4568 +Query 1/1: Action query time = 5.510 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5647 +t=58: Selected seed 195 with value = 0.5647 +Query 1/1: Action query time = 4.539 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6700 +t=74: Selected seed 195 with value = 0.6700 +Query 1/1: Action query time = 5.107 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7844 +t=90: Selected seed 195 with value = 0.7844 +Query 1/1: Action query time = 5.351 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6857 +t=106: Selected seed 195 with value = 0.6857 +Query 1/1: Action query time = 4.914 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8286 +t=122: Selected seed 195 with value = 0.8286 +Query 1/1: Action query time = 4.596 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9613 +t=138: Selected seed 195 with value = 0.9613 +Saved rollout MP4 at path ./rollouts/ft6100_t0_s02/2026_08_02-08_49_15--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.638 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2966 +t=10: Selected seed 195 with value = 0.2966 +Query 1/1: Action query time = 2.614 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3797 +t=26: Selected seed 195 with value = 0.3797 +Query 1/1: Action query time = 4.739 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4232 +t=42: Selected seed 195 with value = 0.4232 +Query 1/1: Action query time = 4.003 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5257 +t=58: Selected seed 195 with value = 0.5257 +Query 1/1: Action query time = 5.110 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5941 +t=74: Selected seed 195 with value = 0.5941 +Query 1/1: Action query time = 5.488 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5859 +t=90: Selected seed 195 with value = 0.5859 +Query 1/1: Action query time = 5.336 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7142 +t=106: Selected seed 195 with value = 0.7142 +Query 1/1: Action query time = 4.760 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8645 +t=122: Selected seed 195 with value = 0.8645 +Query 1/1: Action query time = 5.077 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6100_t0_s02/2026_08_02-08_49_15--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.615 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3052 +t=10: Selected seed 195 with value = 0.3052 +Query 1/1: Action query time = 4.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2936 +t=26: Selected seed 195 with value = 0.2936 +Query 1/1: Action query time = 3.281 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4094 +t=42: Selected seed 195 with value = 0.4094 +Query 1/1: Action query time = 4.241 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5883 +t=58: Selected seed 195 with value = 0.5883 +Query 1/1: Action query time = 5.796 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7162 +t=74: Selected seed 195 with value = 0.7162 +Query 1/1: Action query time = 5.150 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5824 +t=90: Selected seed 195 with value = 0.5824 +Query 1/1: Action query time = 5.021 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6032 +t=106: Selected seed 195 with value = 0.6032 +Query 1/1: Action query time = 4.010 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7495 +t=122: Selected seed 195 with value = 0.7495 +Query 1/1: Action query time = 5.301 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8709 +t=138: Selected seed 195 with value = 0.8709 +Query 1/1: Action query time = 5.404 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=154: Selected seed 195 with value = 0.9964 +Saved rollout MP4 at path ./rollouts/ft6100_t0_s02/2026_08_02-08_49_15--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_49_15--ft6100_t0_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_49_15--ft6100_t0_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..bdca672bd8193c7c0fda70e58607f3d20a38e5cf --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_49_15--ft6100_t0_s03.txt @@ -0,0 +1,221 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t0_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.747 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3211 +t=10: Selected seed 195 with value = 0.3211 +Query 1/1: Action query time = 5.007 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3199 +t=26: Selected seed 195 with value = 0.3199 +Query 1/1: Action query time = 5.454 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4622 +t=42: Selected seed 195 with value = 0.4622 +Query 1/1: Action query time = 4.479 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7051 +t=58: Selected seed 195 with value = 0.7051 +Query 1/1: Action query time = 4.630 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7483 +t=74: Selected seed 195 with value = 0.7483 +Query 1/1: Action query time = 5.441 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8413 +t=90: Selected seed 195 with value = 0.8413 +Query 1/1: Action query time = 5.216 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8982 +t=106: Selected seed 195 with value = 0.8982 +Query 1/1: Action query time = 5.199 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9132 +t=122: Selected seed 195 with value = 0.9132 +Query 1/1: Action query time = 5.011 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9449 +t=138: Selected seed 195 with value = 0.9449 +Query 1/1: Action query time = 4.117 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9582 +t=154: Selected seed 195 with value = 0.9582 +Query 1/1: Action query time = 1.918 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8734 +t=170: Selected seed 195 with value = 0.8734 +Query 1/1: Action query time = 2.802 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8483 +t=186: Selected seed 195 with value = 0.8483 +Query 1/1: Action query time = 5.466 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8233 +t=202: Selected seed 195 with value = 0.8233 +Query 1/1: Action query time = 3.965 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7703 +t=218: Selected seed 195 with value = 0.7703 +Query 1/1: Action query time = 5.255 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7654 +t=234: Selected seed 195 with value = 0.7654 +Query 1/1: Action query time = 5.481 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7655 +t=250: Selected seed 195 with value = 0.7655 +Query 1/1: Action query time = 4.848 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7933 +t=266: Selected seed 195 with value = 0.7933 +Query 1/1: Action query time = 4.590 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8078 +t=282: Selected seed 195 with value = 0.8078 +Query 1/1: Action query time = 5.089 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8137 +t=298: Selected seed 195 with value = 0.8137 +Saved rollout MP4 at path ./rollouts/ft6100_t0_s03/2026_08_02-08_49_15--with_future_img--episode=1--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.796 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3852 +t=10: Selected seed 195 with value = 0.3852 +Query 1/1: Action query time = 4.684 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3557 +t=26: Selected seed 195 with value = 0.3557 +Query 1/1: Action query time = 3.257 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4462 +t=42: Selected seed 195 with value = 0.4462 +Query 1/1: Action query time = 4.650 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5181 +t=58: Selected seed 195 with value = 0.5181 +Query 1/1: Action query time = 5.401 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6157 +t=74: Selected seed 195 with value = 0.6157 +Query 1/1: Action query time = 5.733 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6662 +t=90: Selected seed 195 with value = 0.6662 +Query 1/1: Action query time = 5.285 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7206 +t=106: Selected seed 195 with value = 0.7206 +Query 1/1: Action query time = 4.878 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8933 +t=122: Selected seed 195 with value = 0.8933 +Query 1/1: Action query time = 4.933 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7112 +t=138: Selected seed 195 with value = 0.7112 +Query 1/1: Action query time = 4.967 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7033 +t=154: Selected seed 195 with value = 0.7033 +Query 1/1: Action query time = 3.740 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7646 +t=170: Selected seed 195 with value = 0.7646 +Query 1/1: Action query time = 3.998 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8364 +t=186: Selected seed 195 with value = 0.8364 +Query 1/1: Action query time = 4.613 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9023 +t=202: Selected seed 195 with value = 0.9023 +Query 1/1: Action query time = 3.958 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8946 +t=218: Selected seed 195 with value = 0.8946 +Query 1/1: Action query time = 3.246 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8693 +t=234: Selected seed 195 with value = 0.8693 +Query 1/1: Action query time = 2.940 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8394 +t=250: Selected seed 195 with value = 0.8394 +Query 1/1: Action query time = 3.667 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8107 +t=266: Selected seed 195 with value = 0.8107 +Query 1/1: Action query time = 2.780 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7741 +t=282: Selected seed 195 with value = 0.7741 +Query 1/1: Action query time = 2.691 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7339 +t=298: Selected seed 195 with value = 0.7339 +Saved rollout MP4 at path ./rollouts/ft6100_t0_s03/2026_08_02-08_49_15--with_future_img--episode=2--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 1.227 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3200 +t=10: Selected seed 195 with value = 0.3200 +Query 1/1: Action query time = 1.018 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3821 +t=26: Selected seed 195 with value = 0.3821 +Query 1/1: Action query time = 1.082 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4421 +t=42: Selected seed 195 with value = 0.4421 +Query 1/1: Action query time = 1.227 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5364 +t=58: Selected seed 195 with value = 0.5364 +Query 1/1: Action query time = 1.022 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6186 +t=74: Selected seed 195 with value = 0.6186 +Query 1/1: Action query time = 1.758 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6977 +t=90: Selected seed 195 with value = 0.6977 +Query 1/1: Action query time = 1.574 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7715 +t=106: Selected seed 195 with value = 0.7715 +Query 1/1: Action query time = 1.401 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8187 +t=122: Selected seed 195 with value = 0.8187 +Query 1/1: Action query time = 1.657 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9402 +t=138: Selected seed 195 with value = 0.9402 +Saved rollout MP4 at path ./rollouts/ft6100_t0_s03/2026_08_02-08_49_15--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_49_16--ft6100_t0_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_49_16--ft6100_t0_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..0ef909657af9ed04b47a7787c362c038b972f5f8 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_49_16--ft6100_t0_s14.txt @@ -0,0 +1,149 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t0_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 7.276 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3251 +t=10: Selected seed 195 with value = 0.3251 +Query 1/1: Action query time = 6.451 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3691 +t=26: Selected seed 195 with value = 0.3691 +Query 1/1: Action query time = 6.218 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4567 +t=42: Selected seed 195 with value = 0.4567 +Query 1/1: Action query time = 3.941 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5335 +t=58: Selected seed 195 with value = 0.5335 +Query 1/1: Action query time = 5.026 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6265 +t=74: Selected seed 195 with value = 0.6265 +Query 1/1: Action query time = 5.041 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5764 +t=90: Selected seed 195 with value = 0.5764 +Query 1/1: Action query time = 4.906 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7438 +t=106: Selected seed 195 with value = 0.7438 +Query 1/1: Action query time = 4.877 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8338 +t=122: Selected seed 195 with value = 0.8338 +Query 1/1: Action query time = 3.651 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9471 +t=138: Selected seed 195 with value = 0.9471 +Saved rollout MP4 at path ./rollouts/ft6100_t0_s14/2026_08_02-08_49_16--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.028 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3482 +t=10: Selected seed 195 with value = 0.3482 +Query 1/1: Action query time = 4.506 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3485 +t=26: Selected seed 195 with value = 0.3485 +Query 1/1: Action query time = 4.623 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4451 +t=42: Selected seed 195 with value = 0.4451 +Query 1/1: Action query time = 4.880 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5229 +t=58: Selected seed 195 with value = 0.5229 +Query 1/1: Action query time = 5.012 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6183 +t=74: Selected seed 195 with value = 0.6183 +Query 1/1: Action query time = 5.096 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6238 +t=90: Selected seed 195 with value = 0.6238 +Query 1/1: Action query time = 5.424 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7251 +t=106: Selected seed 195 with value = 0.7251 +Query 1/1: Action query time = 4.797 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7330 +t=122: Selected seed 195 with value = 0.7330 +Query 1/1: Action query time = 3.976 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8417 +t=138: Selected seed 195 with value = 0.8417 +Query 1/1: Action query time = 2.567 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9878 +t=154: Selected seed 195 with value = 0.9878 +Saved rollout MP4 at path ./rollouts/ft6100_t0_s14/2026_08_02-08_49_16--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 1.693 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3562 +t=10: Selected seed 195 with value = 0.3562 +Query 1/1: Action query time = 5.263 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3608 +t=26: Selected seed 195 with value = 0.3608 +Query 1/1: Action query time = 4.807 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4513 +t=42: Selected seed 195 with value = 0.4513 +Query 1/1: Action query time = 4.828 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5393 +t=58: Selected seed 195 with value = 0.5393 +Query 1/1: Action query time = 5.190 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6386 +t=74: Selected seed 195 with value = 0.6386 +Query 1/1: Action query time = 5.220 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6479 +t=90: Selected seed 195 with value = 0.6479 +Query 1/1: Action query time = 5.365 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7316 +t=106: Selected seed 195 with value = 0.7316 +Query 1/1: Action query time = 5.075 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7961 +t=122: Selected seed 195 with value = 0.7961 +Query 1/1: Action query time = 5.103 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8950 +t=138: Selected seed 195 with value = 0.8950 +Query 1/1: Action query time = 4.325 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6100_t0_s14/2026_08_02-08_49_16--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_53_53--ft6100_t1_s00.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_53_53--ft6100_t1_s00.txt new file mode 100644 index 0000000000000000000000000000000000000000..9a4a9382ce0b94a38d0370a2d7cb04d761078df3 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_53_53--ft6100_t1_s00.txt @@ -0,0 +1,344 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t1_s00', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,16,32,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 3.134 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4988 +t=10: Selected seed 195 with value = 0.4988 +Query 1/1: Action query time = 5.029 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6134 +t=26: Selected seed 195 with value = 0.6134 +Query 1/1: Action query time = 4.507 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6228 +t=42: Selected seed 195 with value = 0.6228 +Query 1/1: Action query time = 4.849 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7461 +t=58: Selected seed 195 with value = 0.7461 +Query 1/1: Action query time = 5.158 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8528 +t=74: Selected seed 195 with value = 0.8528 +Query 1/1: Action query time = 5.199 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9770 +t=90: Selected seed 195 with value = 0.9770 +Query 1/1: Action query time = 5.071 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.891 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.322 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.049 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.957 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=170: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 5.067 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=186: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 4.999 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=202: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 5.138 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=218: Selected seed 195 with value = 0.9968 +Query 1/1: Action query time = 5.174 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=234: Selected seed 195 with value = 0.9963 +Query 1/1: Action query time = 5.279 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9956 +t=250: Selected seed 195 with value = 0.9956 +Query 1/1: Action query time = 5.345 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9887 +t=266: Selected seed 195 with value = 0.9887 +Query 1/1: Action query time = 5.348 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9695 +t=282: Selected seed 195 with value = 0.9695 +Query 1/1: Action query time = 4.894 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9682 +t=298: Selected seed 195 with value = 0.9682 +Saved rollout MP4 at path ./rollouts/ft6100_t1_s00/2026_08_02-08_53_53--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 1.534 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4955 +t=10: Selected seed 195 with value = 0.4955 +Query 1/1: Action query time = 3.140 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=26: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 6.148 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6299 +t=42: Selected seed 195 with value = 0.6299 +Query 1/1: Action query time = 5.117 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6724 +t=58: Selected seed 195 with value = 0.6724 +Query 1/1: Action query time = 5.275 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7829 +t=74: Selected seed 195 with value = 0.7829 +Query 1/1: Action query time = 5.552 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8115 +t=90: Selected seed 195 with value = 0.8115 +Query 1/1: Action query time = 5.045 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8700 +t=106: Selected seed 195 with value = 0.8700 +Query 1/1: Action query time = 4.956 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8966 +t=122: Selected seed 195 with value = 0.8966 +Query 1/1: Action query time = 5.047 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9472 +t=138: Selected seed 195 with value = 0.9472 +Query 1/1: Action query time = 4.975 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=154: Selected seed 195 with value = 0.9983 +Query 1/1: Action query time = 5.000 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9564 +t=170: Selected seed 195 with value = 0.9564 +Query 1/1: Action query time = 5.701 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9457 +t=186: Selected seed 195 with value = 0.9457 +Query 1/1: Action query time = 5.672 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9347 +t=202: Selected seed 195 with value = 0.9347 +Query 1/1: Action query time = 5.624 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9225 +t=218: Selected seed 195 with value = 0.9225 +Query 1/1: Action query time = 5.087 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9134 +t=234: Selected seed 195 with value = 0.9134 +Query 1/1: Action query time = 4.557 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9033 +t=250: Selected seed 195 with value = 0.9033 +Query 1/1: Action query time = 5.208 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8795 +t=266: Selected seed 195 with value = 0.8795 +Query 1/1: Action query time = 4.683 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8378 +t=282: Selected seed 195 with value = 0.8378 +Query 1/1: Action query time = 5.091 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7537 +t=298: Selected seed 195 with value = 0.7537 +Saved rollout MP4 at path ./rollouts/ft6100_t1_s00/2026_08_02-08_53_53--with_future_img--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 2.265 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5038 +t=10: Selected seed 195 with value = 0.5038 +Query 1/1: Action query time = 2.228 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5504 +t=26: Selected seed 195 with value = 0.5504 +Query 1/1: Action query time = 4.364 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6272 +t=42: Selected seed 195 with value = 0.6272 +Query 1/1: Action query time = 5.624 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7970 +t=58: Selected seed 195 with value = 0.7970 +Query 1/1: Action query time = 5.170 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8155 +t=74: Selected seed 195 with value = 0.8155 +Query 1/1: Action query time = 4.828 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9217 +t=90: Selected seed 195 with value = 0.9217 +Query 1/1: Action query time = 5.146 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9631 +t=106: Selected seed 195 with value = 0.9631 +Query 1/1: Action query time = 5.187 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.944 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.744 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.889 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=170: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 4.869 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9908 +t=186: Selected seed 195 with value = 0.9908 +Query 1/1: Action query time = 4.944 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=202: Selected seed 195 with value = 0.9903 +Query 1/1: Action query time = 5.055 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9836 +t=218: Selected seed 195 with value = 0.9836 +Query 1/1: Action query time = 4.608 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9732 +t=234: Selected seed 195 with value = 0.9732 +Query 1/1: Action query time = 4.531 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9618 +t=250: Selected seed 195 with value = 0.9618 +Query 1/1: Action query time = 4.079 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9007 +t=266: Selected seed 195 with value = 0.9007 +Query 1/1: Action query time = 4.299 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9222 +t=282: Selected seed 195 with value = 0.9222 +Query 1/1: Action query time = 3.884 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8793 +t=298: Selected seed 195 with value = 0.8793 +Saved rollout MP4 at path ./rollouts/ft6100_t1_s00/2026_08_02-08_53_53--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 1.219 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4931 +t=10: Selected seed 195 with value = 0.4931 +Query 1/1: Action query time = 0.992 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5485 +t=26: Selected seed 195 with value = 0.5485 +Query 1/1: Action query time = 0.991 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6243 +t=42: Selected seed 195 with value = 0.6243 +Query 1/1: Action query time = 1.021 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7564 +t=58: Selected seed 195 with value = 0.7564 +Query 1/1: Action query time = 1.022 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8336 +t=74: Selected seed 195 with value = 0.8336 +Query 1/1: Action query time = 1.226 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9571 +t=90: Selected seed 195 with value = 0.9571 +Query 1/1: Action query time = 1.233 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.233 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.199 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.214 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.207 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=170: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 1.219 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=186: Selected seed 195 with value = 0.9958 +Query 1/1: Action query time = 1.215 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9902 +t=202: Selected seed 195 with value = 0.9902 +Query 1/1: Action query time = 1.226 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9398 +t=218: Selected seed 195 with value = 0.9398 +Query 1/1: Action query time = 1.200 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8396 +t=234: Selected seed 195 with value = 0.8396 +Query 1/1: Action query time = 1.196 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8941 +t=250: Selected seed 195 with value = 0.8941 +Query 1/1: Action query time = 1.209 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.196 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.215 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6100_t1_s00/2026_08_02-08_53_53--with_future_img--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 4 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_53_54--ft6100_t1_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_53_54--ft6100_t1_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..65247576ab31d74c76444d6de8425d4c18e8a4be --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_53_54--ft6100_t1_s03.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t1_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.725 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4858 +t=10: Selected seed 195 with value = 0.4858 +Query 1/1: Action query time = 4.873 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6083 +t=26: Selected seed 195 with value = 0.6083 +Query 1/1: Action query time = 4.770 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6274 +t=42: Selected seed 195 with value = 0.6274 +Query 1/1: Action query time = 5.077 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7453 +t=58: Selected seed 195 with value = 0.7453 +Query 1/1: Action query time = 5.324 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8511 +t=74: Selected seed 195 with value = 0.8511 +Query 1/1: Action query time = 5.201 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9622 +t=90: Selected seed 195 with value = 0.9622 +Query 1/1: Action query time = 4.595 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.183 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.232 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.084 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.960 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.837 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=186: Selected seed 195 with value = 0.9962 +Query 1/1: Action query time = 5.001 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=202: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 5.065 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=218: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 5.166 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=234: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 5.168 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9943 +t=250: Selected seed 195 with value = 0.9943 +Query 1/1: Action query time = 5.153 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=266: Selected seed 195 with value = 0.9966 +Query 1/1: Action query time = 4.861 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=282: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 3.313 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=298: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/ft6100_t1_s03/2026_08_02-08_53_54--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 5.405 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5070 +t=10: Selected seed 195 with value = 0.5070 +Query 1/1: Action query time = 4.664 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=26: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 5.032 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6433 +t=42: Selected seed 195 with value = 0.6433 +Query 1/1: Action query time = 4.978 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7513 +t=58: Selected seed 195 with value = 0.7513 +Query 1/1: Action query time = 5.007 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8537 +t=74: Selected seed 195 with value = 0.8537 +Query 1/1: Action query time = 5.386 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9779 +t=90: Selected seed 195 with value = 0.9779 +Query 1/1: Action query time = 5.431 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.394 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.502 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.348 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9959 +t=154: Selected seed 195 with value = 0.9959 +Query 1/1: Action query time = 5.205 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9956 +t=170: Selected seed 195 with value = 0.9956 +Query 1/1: Action query time = 4.514 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=186: Selected seed 195 with value = 0.9963 +Query 1/1: Action query time = 4.887 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9931 +t=202: Selected seed 195 with value = 0.9931 +Query 1/1: Action query time = 4.678 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9934 +t=218: Selected seed 195 with value = 0.9934 +Query 1/1: Action query time = 5.429 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=234: Selected seed 195 with value = 0.9971 +Query 1/1: Action query time = 4.939 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.782 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9723 +t=266: Selected seed 195 with value = 0.9723 +Query 1/1: Action query time = 4.337 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9972 +t=282: Selected seed 195 with value = 0.9972 +Query 1/1: Action query time = 3.922 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9930 +t=298: Selected seed 195 with value = 0.9930 +Saved rollout MP4 at path ./rollouts/ft6100_t1_s03/2026_08_02-08_53_54--with_future_img--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.412 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5014 +t=10: Selected seed 195 with value = 0.5014 +Query 1/1: Action query time = 5.434 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6175 +t=26: Selected seed 195 with value = 0.6175 +Query 1/1: Action query time = 5.116 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6216 +t=42: Selected seed 195 with value = 0.6216 +Query 1/1: Action query time = 4.958 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7312 +t=58: Selected seed 195 with value = 0.7312 +Query 1/1: Action query time = 5.113 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8487 +t=74: Selected seed 195 with value = 0.8487 +Query 1/1: Action query time = 5.209 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9693 +t=90: Selected seed 195 with value = 0.9693 +Query 1/1: Action query time = 5.285 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.432 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.986 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.982 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=154: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 5.014 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9940 +t=170: Selected seed 195 with value = 0.9940 +Query 1/1: Action query time = 5.192 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9995 +t=186: Selected seed 195 with value = 0.9995 +Query 1/1: Action query time = 4.226 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.171 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=218: Selected seed 195 with value = 0.9989 +Query 1/1: Action query time = 3.405 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=234: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 4.187 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9996 +t=250: Selected seed 195 with value = 0.9996 +Query 1/1: Action query time = 4.068 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=266: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 3.116 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=282: Selected seed 195 with value = 0.9999 +Query 1/1: Action query time = 2.552 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=298: Selected seed 195 with value = 0.9991 +Saved rollout MP4 at path ./rollouts/ft6100_t1_s03/2026_08_02-08_53_54--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_53_54--ft6100_t1_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_53_54--ft6100_t1_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..68a253bef69e18bb5ca2d30939818496d39c0e79 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-08_53_54--ft6100_t1_s06.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t1_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.970 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5106 +t=10: Selected seed 195 with value = 0.5106 +Query 1/1: Action query time = 5.069 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5378 +t=26: Selected seed 195 with value = 0.5378 +Query 1/1: Action query time = 4.552 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6274 +t=42: Selected seed 195 with value = 0.6274 +Query 1/1: Action query time = 4.963 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7833 +t=58: Selected seed 195 with value = 0.7833 +Query 1/1: Action query time = 5.169 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8491 +t=74: Selected seed 195 with value = 0.8491 +Query 1/1: Action query time = 5.362 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9513 +t=90: Selected seed 195 with value = 0.9513 +Query 1/1: Action query time = 5.331 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.110 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.621 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.419 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=154: Selected seed 195 with value = 0.9968 +Query 1/1: Action query time = 5.336 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=170: Selected seed 195 with value = 0.9999 +Query 1/1: Action query time = 5.508 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.135 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=202: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 5.097 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=218: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 5.092 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=234: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 5.068 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=250: Selected seed 195 with value = 0.9997 +Query 1/1: Action query time = 5.134 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.823 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=282: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 1.943 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6100_t1_s06/2026_08_02-08_53_54--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 6.143 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4998 +t=10: Selected seed 195 with value = 0.4998 +Query 1/1: Action query time = 5.411 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5490 +t=26: Selected seed 195 with value = 0.5490 +Query 1/1: Action query time = 5.105 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6368 +t=42: Selected seed 195 with value = 0.6368 +Query 1/1: Action query time = 5.320 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7439 +t=58: Selected seed 195 with value = 0.7439 +Query 1/1: Action query time = 5.124 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8633 +t=74: Selected seed 195 with value = 0.8633 +Query 1/1: Action query time = 4.990 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9615 +t=90: Selected seed 195 with value = 0.9615 +Query 1/1: Action query time = 4.969 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.843 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.995 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9995 +t=138: Selected seed 195 with value = 0.9995 +Query 1/1: Action query time = 5.611 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9996 +t=154: Selected seed 195 with value = 0.9996 +Query 1/1: Action query time = 5.642 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=170: Selected seed 195 with value = 0.9958 +Query 1/1: Action query time = 5.391 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9787 +t=186: Selected seed 195 with value = 0.9787 +Query 1/1: Action query time = 4.860 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9325 +t=202: Selected seed 195 with value = 0.9325 +Query 1/1: Action query time = 3.961 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8424 +t=218: Selected seed 195 with value = 0.8424 +Query 1/1: Action query time = 5.187 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9176 +t=234: Selected seed 195 with value = 0.9176 +Query 1/1: Action query time = 4.725 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.082 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.034 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.206 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=298: Selected seed 195 with value = 0.9998 +Saved rollout MP4 at path ./rollouts/ft6100_t1_s06/2026_08_02-08_53_54--with_future_img--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 4.935 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4861 +t=10: Selected seed 195 with value = 0.4861 +Query 1/1: Action query time = 4.910 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5409 +t=26: Selected seed 195 with value = 0.5409 +Query 1/1: Action query time = 5.079 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6289 +t=42: Selected seed 195 with value = 0.6289 +Query 1/1: Action query time = 5.471 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8115 +t=58: Selected seed 195 with value = 0.8115 +Query 1/1: Action query time = 5.028 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8293 +t=74: Selected seed 195 with value = 0.8293 +Query 1/1: Action query time = 4.958 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9286 +t=90: Selected seed 195 with value = 0.9286 +Query 1/1: Action query time = 4.865 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9705 +t=106: Selected seed 195 with value = 0.9705 +Query 1/1: Action query time = 5.160 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.175 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.177 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.237 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.222 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=186: Selected seed 195 with value = 0.9989 +Query 1/1: Action query time = 4.881 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=202: Selected seed 195 with value = 0.9982 +Query 1/1: Action query time = 4.388 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=218: Selected seed 195 with value = 0.9958 +Query 1/1: Action query time = 3.014 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=234: Selected seed 195 with value = 0.9903 +Query 1/1: Action query time = 3.685 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9882 +t=250: Selected seed 195 with value = 0.9882 +Query 1/1: Action query time = 3.323 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=266: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 3.235 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9828 +t=282: Selected seed 195 with value = 0.9828 +Query 1/1: Action query time = 2.066 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9693 +t=298: Selected seed 195 with value = 0.9693 +Saved rollout MP4 at path ./rollouts/ft6100_t1_s06/2026_08_02-08_53_54--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s00.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s00.txt new file mode 100644 index 0000000000000000000000000000000000000000..c479beb521c3eaf5d18a7c82c0653b1bd71fd9c6 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s00.txt @@ -0,0 +1,192 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t2_s00', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,16,32,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.969 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5479 +t=10: Selected seed 195 with value = 0.5479 +Query 1/1: Action query time = 4.822 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6116 +t=26: Selected seed 195 with value = 0.6116 +Query 1/1: Action query time = 5.086 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8030 +t=42: Selected seed 195 with value = 0.8030 +Query 1/1: Action query time = 4.660 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7977 +t=58: Selected seed 195 with value = 0.7977 +Query 1/1: Action query time = 5.736 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8341 +t=74: Selected seed 195 with value = 0.8341 +Query 1/1: Action query time = 4.942 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9740 +t=90: Selected seed 195 with value = 0.9740 +Query 1/1: Action query time = 5.089 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.645 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.439 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.837 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.274 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.388 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=186: Selected seed 195 with value = 0.9997 +Query 1/1: Action query time = 5.018 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9925 +t=202: Selected seed 195 with value = 0.9925 +Query 1/1: Action query time = 5.218 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9785 +t=218: Selected seed 195 with value = 0.9785 +Query 1/1: Action query time = 5.209 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9482 +t=234: Selected seed 195 with value = 0.9482 +Query 1/1: Action query time = 5.343 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9365 +t=250: Selected seed 195 with value = 0.9365 +Query 1/1: Action query time = 5.351 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9188 +t=266: Selected seed 195 with value = 0.9188 +Query 1/1: Action query time = 2.890 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8985 +t=282: Selected seed 195 with value = 0.8985 +Query 1/1: Action query time = 2.919 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8785 +t=298: Selected seed 195 with value = 0.8785 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s00/2026_08_02-09_00_34--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.196 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4246 +t=10: Selected seed 195 with value = 0.4246 +Query 1/1: Action query time = 3.294 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4569 +t=26: Selected seed 195 with value = 0.4569 +Query 1/1: Action query time = 3.145 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6258 +t=42: Selected seed 195 with value = 0.6258 +Query 1/1: Action query time = 2.731 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6767 +t=58: Selected seed 195 with value = 0.6767 +Query 1/1: Action query time = 2.988 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7534 +t=74: Selected seed 195 with value = 0.7534 +Query 1/1: Action query time = 3.459 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8451 +t=90: Selected seed 195 with value = 0.8451 +Query 1/1: Action query time = 3.960 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9929 +t=106: Selected seed 195 with value = 0.9929 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s00/2026_08_02-09_00_34--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.660 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4091 +t=10: Selected seed 195 with value = 0.4091 +Query 1/1: Action query time = 4.103 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4508 +t=26: Selected seed 195 with value = 0.4508 +Query 1/1: Action query time = 3.490 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6965 +t=42: Selected seed 195 with value = 0.6965 +Query 1/1: Action query time = 2.997 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6926 +t=58: Selected seed 195 with value = 0.6926 +Query 1/1: Action query time = 2.830 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7829 +t=74: Selected seed 195 with value = 0.7829 +Query 1/1: Action query time = 2.578 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9052 +t=90: Selected seed 195 with value = 0.9052 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s00/2026_08_02-09_00_34--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 3.194 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4367 +t=10: Selected seed 195 with value = 0.4367 +Query 1/1: Action query time = 1.616 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4794 +t=26: Selected seed 195 with value = 0.4794 +Query 1/1: Action query time = 2.014 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6868 +t=42: Selected seed 195 with value = 0.6868 +Query 1/1: Action query time = 1.563 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7133 +t=58: Selected seed 195 with value = 0.7133 +Query 1/1: Action query time = 2.578 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8122 +t=74: Selected seed 195 with value = 0.8122 +Query 1/1: Action query time = 2.444 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9598 +t=90: Selected seed 195 with value = 0.9598 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s00/2026_08_02-09_00_34--with_future_img--episode=4--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 4 +# successes: 3 (75.0%) +Current task success rate: 0.75 +Current total success rate: 0.75 +Final results: +Total episodes: 4 +Total successes: 3 +Overall success rate: 0.7500 (75.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..f4d7f166bd14514928c71bb38bd9011c4e739f2d --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s02.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t2_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.312 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4419 +t=10: Selected seed 195 with value = 0.4419 +Query 1/1: Action query time = 5.102 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5020 +t=26: Selected seed 195 with value = 0.5020 +Query 1/1: Action query time = 5.049 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5894 +t=42: Selected seed 195 with value = 0.5894 +Query 1/1: Action query time = 5.217 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7319 +t=58: Selected seed 195 with value = 0.7319 +Query 1/1: Action query time = 4.932 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7809 +t=74: Selected seed 195 with value = 0.7809 +Query 1/1: Action query time = 4.061 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8448 +t=90: Selected seed 195 with value = 0.8448 +Query 1/1: Action query time = 5.068 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=106: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 3.767 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.837 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.206 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.379 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.498 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.886 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=202: Selected seed 195 with value = 0.9966 +Query 1/1: Action query time = 4.123 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9907 +t=218: Selected seed 195 with value = 0.9907 +Query 1/1: Action query time = 4.919 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9824 +t=234: Selected seed 195 with value = 0.9824 +Query 1/1: Action query time = 5.219 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9656 +t=250: Selected seed 195 with value = 0.9656 +Query 1/1: Action query time = 5.066 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9419 +t=266: Selected seed 195 with value = 0.9419 +Query 1/1: Action query time = 5.098 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9343 +t=282: Selected seed 195 with value = 0.9343 +Query 1/1: Action query time = 4.933 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9151 +t=298: Selected seed 195 with value = 0.9151 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s02/2026_08_02-09_00_34--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 1.468 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4781 +t=10: Selected seed 195 with value = 0.4781 +Query 1/1: Action query time = 1.771 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5764 +t=26: Selected seed 195 with value = 0.5764 +Query 1/1: Action query time = 2.594 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7484 +t=42: Selected seed 195 with value = 0.7484 +Query 1/1: Action query time = 4.052 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7970 +t=58: Selected seed 195 with value = 0.7970 +Query 1/1: Action query time = 4.127 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7846 +t=74: Selected seed 195 with value = 0.7846 +Query 1/1: Action query time = 4.066 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8773 +t=90: Selected seed 195 with value = 0.8773 +Query 1/1: Action query time = 3.195 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.517 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.988 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.987 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.221 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.892 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.816 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.000 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=218: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 3.092 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=234: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 2.239 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9813 +t=250: Selected seed 195 with value = 0.9813 +Query 1/1: Action query time = 2.925 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9697 +t=266: Selected seed 195 with value = 0.9697 +Query 1/1: Action query time = 1.980 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9462 +t=282: Selected seed 195 with value = 0.9462 +Query 1/1: Action query time = 3.225 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9287 +t=298: Selected seed 195 with value = 0.9287 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s02/2026_08_02-09_00_34--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 2.665 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3706 +t=10: Selected seed 195 with value = 0.3706 +Query 1/1: Action query time = 2.256 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3619 +t=26: Selected seed 195 with value = 0.3619 +Query 1/1: Action query time = 1.547 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4103 +t=42: Selected seed 195 with value = 0.4103 +Query 1/1: Action query time = 1.370 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6926 +t=58: Selected seed 195 with value = 0.6926 +Query 1/1: Action query time = 1.838 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7200 +t=74: Selected seed 195 with value = 0.7200 +Query 1/1: Action query time = 2.135 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7944 +t=90: Selected seed 195 with value = 0.7944 +Query 1/1: Action query time = 2.089 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7434 +t=106: Selected seed 195 with value = 0.7434 +Query 1/1: Action query time = 2.656 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7044 +t=122: Selected seed 195 with value = 0.7044 +Query 1/1: Action query time = 2.145 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7222 +t=138: Selected seed 195 with value = 0.7222 +Query 1/1: Action query time = 1.771 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6861 +t=154: Selected seed 195 with value = 0.6861 +Query 1/1: Action query time = 1.389 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7153 +t=170: Selected seed 195 with value = 0.7153 +Query 1/1: Action query time = 1.432 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7693 +t=186: Selected seed 195 with value = 0.7693 +Query 1/1: Action query time = 1.925 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7140 +t=202: Selected seed 195 with value = 0.7140 +Query 1/1: Action query time = 1.925 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7721 +t=218: Selected seed 195 with value = 0.7721 +Query 1/1: Action query time = 2.252 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7708 +t=234: Selected seed 195 with value = 0.7708 +Query 1/1: Action query time = 1.935 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8007 +t=250: Selected seed 195 with value = 0.8007 +Query 1/1: Action query time = 1.869 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7223 +t=266: Selected seed 195 with value = 0.7223 +Query 1/1: Action query time = 2.009 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8065 +t=282: Selected seed 195 with value = 0.8065 +Query 1/1: Action query time = 1.831 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7317 +t=298: Selected seed 195 with value = 0.7317 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s02/2026_08_02-09_00_34--with_future_img--episode=3--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s05.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s05.txt new file mode 100644 index 0000000000000000000000000000000000000000..c7c00d38ab2bb2608260bdd8f07b081c005a9a66 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s05.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t2_s05', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,21,37', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.399 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4766 +t=10: Selected seed 195 with value = 0.4766 +Query 1/1: Action query time = 4.221 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5098 +t=26: Selected seed 195 with value = 0.5098 +Query 1/1: Action query time = 5.104 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6246 +t=42: Selected seed 195 with value = 0.6246 +Query 1/1: Action query time = 5.444 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7417 +t=58: Selected seed 195 with value = 0.7417 +Query 1/1: Action query time = 5.368 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7881 +t=74: Selected seed 195 with value = 0.7881 +Query 1/1: Action query time = 3.696 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8753 +t=90: Selected seed 195 with value = 0.8753 +Query 1/1: Action query time = 5.588 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.336 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.030 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.389 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.348 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.910 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.329 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.290 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=218: Selected seed 195 with value = 0.9976 +Query 1/1: Action query time = 4.797 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=234: Selected seed 195 with value = 0.9946 +Query 1/1: Action query time = 5.152 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9895 +t=250: Selected seed 195 with value = 0.9895 +Query 1/1: Action query time = 5.213 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9861 +t=266: Selected seed 195 with value = 0.9861 +Query 1/1: Action query time = 5.546 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9786 +t=282: Selected seed 195 with value = 0.9786 +Query 1/1: Action query time = 5.298 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9651 +t=298: Selected seed 195 with value = 0.9651 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s05/2026_08_02-09_00_34--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.080 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4718 +t=10: Selected seed 195 with value = 0.4718 +Query 1/1: Action query time = 1.891 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6026 +t=26: Selected seed 195 with value = 0.6026 +Query 1/1: Action query time = 3.016 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8421 +t=42: Selected seed 195 with value = 0.8421 +Query 1/1: Action query time = 4.130 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8017 +t=58: Selected seed 195 with value = 0.8017 +Query 1/1: Action query time = 4.130 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7739 +t=74: Selected seed 195 with value = 0.7739 +Query 1/1: Action query time = 3.763 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8634 +t=90: Selected seed 195 with value = 0.8634 +Query 1/1: Action query time = 3.253 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9585 +t=106: Selected seed 195 with value = 0.9585 +Query 1/1: Action query time = 3.777 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.334 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.232 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.491 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.886 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.418 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.761 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.111 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.338 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9942 +t=250: Selected seed 195 with value = 0.9942 +Query 1/1: Action query time = 2.327 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9783 +t=266: Selected seed 195 with value = 0.9783 +Query 1/1: Action query time = 2.699 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9460 +t=282: Selected seed 195 with value = 0.9460 +Query 1/1: Action query time = 3.310 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9018 +t=298: Selected seed 195 with value = 0.9018 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s05/2026_08_02-09_00_34--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 1.532 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3172 +t=10: Selected seed 195 with value = 0.3172 +Query 1/1: Action query time = 2.248 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3816 +t=26: Selected seed 195 with value = 0.3816 +Query 1/1: Action query time = 1.892 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4520 +t=42: Selected seed 195 with value = 0.4520 +Query 1/1: Action query time = 2.186 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7705 +t=58: Selected seed 195 with value = 0.7705 +Query 1/1: Action query time = 2.135 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7937 +t=74: Selected seed 195 with value = 0.7937 +Query 1/1: Action query time = 1.452 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6874 +t=90: Selected seed 195 with value = 0.6874 +Query 1/1: Action query time = 1.627 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4375 +t=106: Selected seed 195 with value = 0.4375 +Query 1/1: Action query time = 2.093 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7284 +t=122: Selected seed 195 with value = 0.7284 +Query 1/1: Action query time = 2.263 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7380 +t=138: Selected seed 195 with value = 0.7380 +Query 1/1: Action query time = 1.936 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7769 +t=154: Selected seed 195 with value = 0.7769 +Query 1/1: Action query time = 2.115 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7555 +t=170: Selected seed 195 with value = 0.7555 +Query 1/1: Action query time = 2.244 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7259 +t=186: Selected seed 195 with value = 0.7259 +Query 1/1: Action query time = 2.007 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7781 +t=202: Selected seed 195 with value = 0.7781 +Query 1/1: Action query time = 2.090 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7746 +t=218: Selected seed 195 with value = 0.7746 +Query 1/1: Action query time = 1.758 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7590 +t=234: Selected seed 195 with value = 0.7590 +Query 1/1: Action query time = 1.843 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6714 +t=250: Selected seed 195 with value = 0.6714 +Query 1/1: Action query time = 1.715 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6651 +t=266: Selected seed 195 with value = 0.6651 +Query 1/1: Action query time = 1.922 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9617 +t=282: Selected seed 195 with value = 0.9617 +Query 1/1: Action query time = 1.431 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9642 +t=298: Selected seed 195 with value = 0.9642 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s05/2026_08_02-09_00_34--with_future_img--episode=3--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..cbd0fcf60f3660a46312224a4fc8546cc7d7ba52 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_34--ft6100_t2_s08.txt @@ -0,0 +1,161 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t2_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.182 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4738 +t=10: Selected seed 195 with value = 0.4738 +Query 1/1: Action query time = 5.316 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4772 +t=26: Selected seed 195 with value = 0.4772 +Query 1/1: Action query time = 5.177 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6197 +t=42: Selected seed 195 with value = 0.6197 +Query 1/1: Action query time = 4.789 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7709 +t=58: Selected seed 195 with value = 0.7709 +Query 1/1: Action query time = 4.870 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8310 +t=74: Selected seed 195 with value = 0.8310 +Query 1/1: Action query time = 5.037 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9503 +t=90: Selected seed 195 with value = 0.9503 +Query 1/1: Action query time = 3.866 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.374 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.540 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.129 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.533 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.438 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.164 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.655 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.780 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=234: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 4.989 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=250: Selected seed 195 with value = 0.9965 +Query 1/1: Action query time = 5.197 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9936 +t=266: Selected seed 195 with value = 0.9936 +Query 1/1: Action query time = 5.378 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9898 +t=282: Selected seed 195 with value = 0.9898 +Query 1/1: Action query time = 5.285 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9853 +t=298: Selected seed 195 with value = 0.9853 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s08/2026_08_02-09_00_34--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.081 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3485 +t=10: Selected seed 195 with value = 0.3485 +Query 1/1: Action query time = 4.701 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4035 +t=26: Selected seed 195 with value = 0.4035 +Query 1/1: Action query time = 5.325 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8913 +t=42: Selected seed 195 with value = 0.8913 +Query 1/1: Action query time = 5.575 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6906 +t=58: Selected seed 195 with value = 0.6906 +Query 1/1: Action query time = 5.027 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7492 +t=74: Selected seed 195 with value = 0.7492 +Query 1/1: Action query time = 3.845 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8232 +t=90: Selected seed 195 with value = 0.8232 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s08/2026_08_02-09_00_34--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 2.711 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4035 +t=10: Selected seed 195 with value = 0.4035 +Query 1/1: Action query time = 4.548 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4504 +t=26: Selected seed 195 with value = 0.4504 +Query 1/1: Action query time = 4.429 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8077 +t=42: Selected seed 195 with value = 0.8077 +Query 1/1: Action query time = 4.523 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6440 +t=58: Selected seed 195 with value = 0.6440 +Query 1/1: Action query time = 4.711 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6651 +t=74: Selected seed 195 with value = 0.6651 +Query 1/1: Action query time = 4.650 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7510 +t=90: Selected seed 195 with value = 0.7510 +Query 1/1: Action query time = 4.417 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9374 +t=106: Selected seed 195 with value = 0.9374 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s08/2026_08_02-09_00_34--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_35--ft6100_t2_s10.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_35--ft6100_t2_s10.txt new file mode 100644 index 0000000000000000000000000000000000000000..20c6970a1a0fb1eec0306d9ce4f620946521ecb2 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_00_35--ft6100_t2_s10.txt @@ -0,0 +1,209 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t2_s10', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='10,26,42', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.332 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4210 +t=10: Selected seed 195 with value = 0.4210 +Query 1/1: Action query time = 5.365 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4989 +t=26: Selected seed 195 with value = 0.4989 +Query 1/1: Action query time = 5.131 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7013 +t=42: Selected seed 195 with value = 0.7013 +Query 1/1: Action query time = 4.881 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7214 +t=58: Selected seed 195 with value = 0.7214 +Query 1/1: Action query time = 4.936 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8112 +t=74: Selected seed 195 with value = 0.8112 +Query 1/1: Action query time = 5.076 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9269 +t=90: Selected seed 195 with value = 0.9269 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s10/2026_08_02-09_00_35--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.826 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3018 +t=10: Selected seed 195 with value = 0.3018 +Query 1/1: Action query time = 4.680 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3344 +t=26: Selected seed 195 with value = 0.3344 +Query 1/1: Action query time = 4.811 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3997 +t=42: Selected seed 195 with value = 0.3997 +Query 1/1: Action query time = 5.028 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8106 +t=58: Selected seed 195 with value = 0.8106 +Query 1/1: Action query time = 5.431 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7452 +t=74: Selected seed 195 with value = 0.7452 +Query 1/1: Action query time = 5.248 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7963 +t=90: Selected seed 195 with value = 0.7963 +Query 1/1: Action query time = 5.049 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5453 +t=106: Selected seed 195 with value = 0.5453 +Query 1/1: Action query time = 4.270 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7402 +t=122: Selected seed 195 with value = 0.7402 +Query 1/1: Action query time = 5.015 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6403 +t=138: Selected seed 195 with value = 0.6403 +Query 1/1: Action query time = 5.076 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7433 +t=154: Selected seed 195 with value = 0.7433 +Query 1/1: Action query time = 5.282 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7113 +t=170: Selected seed 195 with value = 0.7113 +Query 1/1: Action query time = 5.232 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7370 +t=186: Selected seed 195 with value = 0.7370 +Query 1/1: Action query time = 3.845 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7020 +t=202: Selected seed 195 with value = 0.7020 +Query 1/1: Action query time = 2.397 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7269 +t=218: Selected seed 195 with value = 0.7269 +Query 1/1: Action query time = 2.958 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9071 +t=234: Selected seed 195 with value = 0.9071 +Query 1/1: Action query time = 5.130 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.360 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.254 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.110 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s10/2026_08_02-09_00_35--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 2.780 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4421 +t=10: Selected seed 195 with value = 0.4421 +Query 1/1: Action query time = 3.979 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4990 +t=26: Selected seed 195 with value = 0.4990 +Query 1/1: Action query time = 4.617 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7494 +t=42: Selected seed 195 with value = 0.7494 +Query 1/1: Action query time = 4.814 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7780 +t=58: Selected seed 195 with value = 0.7780 +Query 1/1: Action query time = 4.312 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7062 +t=74: Selected seed 195 with value = 0.7062 +Query 1/1: Action query time = 4.098 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9515 +t=90: Selected seed 195 with value = 0.9515 +Query 1/1: Action query time = 4.385 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9558 +t=106: Selected seed 195 with value = 0.9558 +Query 1/1: Action query time = 3.477 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9598 +t=122: Selected seed 195 with value = 0.9598 +Query 1/1: Action query time = 2.811 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9516 +t=138: Selected seed 195 with value = 0.9516 +Query 1/1: Action query time = 2.814 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9468 +t=154: Selected seed 195 with value = 0.9468 +Query 1/1: Action query time = 2.338 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9425 +t=170: Selected seed 195 with value = 0.9425 +Query 1/1: Action query time = 2.555 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9361 +t=186: Selected seed 195 with value = 0.9361 +Query 1/1: Action query time = 2.437 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9334 +t=202: Selected seed 195 with value = 0.9334 +Query 1/1: Action query time = 1.857 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9302 +t=218: Selected seed 195 with value = 0.9302 +Query 1/1: Action query time = 1.436 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9255 +t=234: Selected seed 195 with value = 0.9255 +Query 1/1: Action query time = 1.421 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9236 +t=250: Selected seed 195 with value = 0.9236 +Query 1/1: Action query time = 2.898 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9207 +t=266: Selected seed 195 with value = 0.9207 +Query 1/1: Action query time = 3.084 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9197 +t=282: Selected seed 195 with value = 0.9197 +Query 1/1: Action query time = 2.699 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9190 +t=298: Selected seed 195 with value = 0.9190 +Saved rollout MP4 at path ./rollouts/ft6100_t2_s10/2026_08_02-09_00_35--with_future_img--episode=3--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s00.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s00.txt new file mode 100644 index 0000000000000000000000000000000000000000..77334659e97fa7aa73db06226d0ecc508ea75192 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s00.txt @@ -0,0 +1,344 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t3_s00', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,16,32,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 6.414 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2049 +t=10: Selected seed 195 with value = 0.2049 +Query 1/1: Action query time = 3.626 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2156 +t=26: Selected seed 195 with value = 0.2156 +Query 1/1: Action query time = 4.689 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2318 +t=42: Selected seed 195 with value = 0.2318 +Query 1/1: Action query time = 6.004 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2986 +t=58: Selected seed 195 with value = 0.2986 +Query 1/1: Action query time = 5.686 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2693 +t=74: Selected seed 195 with value = 0.2693 +Query 1/1: Action query time = 4.756 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5317 +t=90: Selected seed 195 with value = 0.5317 +Query 1/1: Action query time = 4.628 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5476 +t=106: Selected seed 195 with value = 0.5476 +Query 1/1: Action query time = 5.561 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7008 +t=122: Selected seed 195 with value = 0.7008 +Query 1/1: Action query time = 5.428 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5793 +t=138: Selected seed 195 with value = 0.5793 +Query 1/1: Action query time = 5.242 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5696 +t=154: Selected seed 195 with value = 0.5696 +Query 1/1: Action query time = 5.163 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5808 +t=170: Selected seed 195 with value = 0.5808 +Query 1/1: Action query time = 5.135 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5837 +t=186: Selected seed 195 with value = 0.5837 +Query 1/1: Action query time = 5.033 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5626 +t=202: Selected seed 195 with value = 0.5626 +Query 1/1: Action query time = 5.150 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5578 +t=218: Selected seed 195 with value = 0.5578 +Query 1/1: Action query time = 5.057 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5615 +t=234: Selected seed 195 with value = 0.5615 +Query 1/1: Action query time = 4.867 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5777 +t=250: Selected seed 195 with value = 0.5777 +Query 1/1: Action query time = 4.985 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5714 +t=266: Selected seed 195 with value = 0.5714 +Query 1/1: Action query time = 4.682 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5642 +t=282: Selected seed 195 with value = 0.5642 +Query 1/1: Action query time = 2.442 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5795 +t=298: Selected seed 195 with value = 0.5795 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s00/2026_08_02-09_05_11--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.804 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2122 +t=10: Selected seed 195 with value = 0.2122 +Query 1/1: Action query time = 5.128 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2967 +t=26: Selected seed 195 with value = 0.2967 +Query 1/1: Action query time = 5.953 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2480 +t=42: Selected seed 195 with value = 0.2480 +Query 1/1: Action query time = 5.392 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3301 +t=58: Selected seed 195 with value = 0.3301 +Query 1/1: Action query time = 5.010 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4229 +t=74: Selected seed 195 with value = 0.4229 +Query 1/1: Action query time = 4.961 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5345 +t=90: Selected seed 195 with value = 0.5345 +Query 1/1: Action query time = 4.987 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5633 +t=106: Selected seed 195 with value = 0.5633 +Query 1/1: Action query time = 4.969 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3761 +t=122: Selected seed 195 with value = 0.3761 +Query 1/1: Action query time = 5.135 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5907 +t=138: Selected seed 195 with value = 0.5907 +Query 1/1: Action query time = 5.173 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7428 +t=154: Selected seed 195 with value = 0.7428 +Query 1/1: Action query time = 5.202 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8768 +t=170: Selected seed 195 with value = 0.8768 +Query 1/1: Action query time = 5.158 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9570 +t=186: Selected seed 195 with value = 0.9570 +Query 1/1: Action query time = 5.558 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9200 +t=202: Selected seed 195 with value = 0.9200 +Query 1/1: Action query time = 5.275 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8846 +t=218: Selected seed 195 with value = 0.8846 +Query 1/1: Action query time = 5.333 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9101 +t=234: Selected seed 195 with value = 0.9101 +Query 1/1: Action query time = 5.261 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9479 +t=250: Selected seed 195 with value = 0.9479 +Query 1/1: Action query time = 3.907 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9426 +t=266: Selected seed 195 with value = 0.9426 +Query 1/1: Action query time = 3.503 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9147 +t=282: Selected seed 195 with value = 0.9147 +Query 1/1: Action query time = 4.879 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9229 +t=298: Selected seed 195 with value = 0.9229 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s00/2026_08_02-09_05_11--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.638 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2248 +t=10: Selected seed 195 with value = 0.2248 +Query 1/1: Action query time = 4.734 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2963 +t=26: Selected seed 195 with value = 0.2963 +Query 1/1: Action query time = 4.709 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2257 +t=42: Selected seed 195 with value = 0.2257 +Query 1/1: Action query time = 5.084 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3176 +t=58: Selected seed 195 with value = 0.3176 +Query 1/1: Action query time = 5.376 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3676 +t=74: Selected seed 195 with value = 0.3676 +Query 1/1: Action query time = 5.191 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4502 +t=90: Selected seed 195 with value = 0.4502 +Query 1/1: Action query time = 5.220 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6164 +t=106: Selected seed 195 with value = 0.6164 +Query 1/1: Action query time = 5.091 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5962 +t=122: Selected seed 195 with value = 0.5962 +Query 1/1: Action query time = 5.200 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7103 +t=138: Selected seed 195 with value = 0.7103 +Query 1/1: Action query time = 5.205 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7823 +t=154: Selected seed 195 with value = 0.7823 +Query 1/1: Action query time = 5.137 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8825 +t=170: Selected seed 195 with value = 0.8825 +Query 1/1: Action query time = 5.215 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9473 +t=186: Selected seed 195 with value = 0.9473 +Query 1/1: Action query time = 5.297 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9470 +t=202: Selected seed 195 with value = 0.9470 +Query 1/1: Action query time = 5.264 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9719 +t=218: Selected seed 195 with value = 0.9719 +Query 1/1: Action query time = 5.233 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9490 +t=234: Selected seed 195 with value = 0.9490 +Query 1/1: Action query time = 4.254 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9699 +t=250: Selected seed 195 with value = 0.9699 +Query 1/1: Action query time = 3.907 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9338 +t=266: Selected seed 195 with value = 0.9338 +Query 1/1: Action query time = 3.688 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9251 +t=282: Selected seed 195 with value = 0.9251 +Query 1/1: Action query time = 3.226 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9315 +t=298: Selected seed 195 with value = 0.9315 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s00/2026_08_02-09_05_11--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 1.154 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2120 +t=10: Selected seed 195 with value = 0.2120 +Query 1/1: Action query time = 1.264 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2720 +t=26: Selected seed 195 with value = 0.2720 +Query 1/1: Action query time = 1.239 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3032 +t=42: Selected seed 195 with value = 0.3032 +Query 1/1: Action query time = 1.197 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3107 +t=58: Selected seed 195 with value = 0.3107 +Query 1/1: Action query time = 1.211 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4016 +t=74: Selected seed 195 with value = 0.4016 +Query 1/1: Action query time = 1.219 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4797 +t=90: Selected seed 195 with value = 0.4797 +Query 1/1: Action query time = 1.202 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3833 +t=106: Selected seed 195 with value = 0.3833 +Query 1/1: Action query time = 1.204 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6638 +t=122: Selected seed 195 with value = 0.6638 +Query 1/1: Action query time = 1.147 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9272 +t=138: Selected seed 195 with value = 0.9272 +Query 1/1: Action query time = 1.171 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9408 +t=154: Selected seed 195 with value = 0.9408 +Query 1/1: Action query time = 1.216 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8925 +t=170: Selected seed 195 with value = 0.8925 +Query 1/1: Action query time = 1.239 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7985 +t=186: Selected seed 195 with value = 0.7985 +Query 1/1: Action query time = 1.233 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5993 +t=202: Selected seed 195 with value = 0.5993 +Query 1/1: Action query time = 1.220 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6852 +t=218: Selected seed 195 with value = 0.6852 +Query 1/1: Action query time = 1.209 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7138 +t=234: Selected seed 195 with value = 0.7138 +Query 1/1: Action query time = 1.244 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8246 +t=250: Selected seed 195 with value = 0.8246 +Query 1/1: Action query time = 1.229 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9538 +t=266: Selected seed 195 with value = 0.9538 +Query 1/1: Action query time = 1.204 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9956 +t=282: Selected seed 195 with value = 0.9956 +Query 1/1: Action query time = 1.232 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9944 +t=298: Selected seed 195 with value = 0.9944 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s00/2026_08_02-09_05_11--with_future_img--episode=4--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 4 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..5a48e584225558bd5cc1645306385ce74d9ec6fa --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s01.txt @@ -0,0 +1,344 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t3_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 6.210 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2022 +t=10: Selected seed 195 with value = 0.2022 +Query 1/1: Action query time = 6.054 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2208 +t=26: Selected seed 195 with value = 0.2208 +Query 1/1: Action query time = 6.052 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2491 +t=42: Selected seed 195 with value = 0.2491 +Query 1/1: Action query time = 4.063 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2514 +t=58: Selected seed 195 with value = 0.2514 +Query 1/1: Action query time = 4.562 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2399 +t=74: Selected seed 195 with value = 0.2399 +Query 1/1: Action query time = 5.256 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2285 +t=90: Selected seed 195 with value = 0.2285 +Query 1/1: Action query time = 5.511 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2958 +t=106: Selected seed 195 with value = 0.2958 +Query 1/1: Action query time = 4.773 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4993 +t=122: Selected seed 195 with value = 0.4993 +Query 1/1: Action query time = 5.070 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5322 +t=138: Selected seed 195 with value = 0.5322 +Query 1/1: Action query time = 5.074 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7702 +t=154: Selected seed 195 with value = 0.7702 +Query 1/1: Action query time = 5.058 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9196 +t=170: Selected seed 195 with value = 0.9196 +Query 1/1: Action query time = 5.062 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9043 +t=186: Selected seed 195 with value = 0.9043 +Query 1/1: Action query time = 5.054 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9442 +t=202: Selected seed 195 with value = 0.9442 +Query 1/1: Action query time = 4.807 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9348 +t=218: Selected seed 195 with value = 0.9348 +Query 1/1: Action query time = 4.727 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7591 +t=234: Selected seed 195 with value = 0.7591 +Query 1/1: Action query time = 4.664 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6794 +t=250: Selected seed 195 with value = 0.6794 +Query 1/1: Action query time = 5.019 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6990 +t=266: Selected seed 195 with value = 0.6990 +Query 1/1: Action query time = 4.413 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8460 +t=282: Selected seed 195 with value = 0.8460 +Query 1/1: Action query time = 1.963 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8463 +t=298: Selected seed 195 with value = 0.8463 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s01/2026_08_02-09_05_11--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 5.913 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2388 +t=10: Selected seed 195 with value = 0.2388 +Query 1/1: Action query time = 4.753 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4208 +t=26: Selected seed 195 with value = 0.4208 +Query 1/1: Action query time = 4.457 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5103 +t=42: Selected seed 195 with value = 0.5103 +Query 1/1: Action query time = 5.282 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2366 +t=58: Selected seed 195 with value = 0.2366 +Query 1/1: Action query time = 5.209 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4714 +t=74: Selected seed 195 with value = 0.4714 +Query 1/1: Action query time = 5.485 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9371 +t=90: Selected seed 195 with value = 0.9371 +Query 1/1: Action query time = 5.458 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9386 +t=106: Selected seed 195 with value = 0.9386 +Query 1/1: Action query time = 5.592 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8079 +t=122: Selected seed 195 with value = 0.8079 +Query 1/1: Action query time = 5.104 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7341 +t=138: Selected seed 195 with value = 0.7341 +Query 1/1: Action query time = 5.158 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8784 +t=154: Selected seed 195 with value = 0.8784 +Query 1/1: Action query time = 5.072 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9394 +t=170: Selected seed 195 with value = 0.9394 +Query 1/1: Action query time = 4.956 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9655 +t=186: Selected seed 195 with value = 0.9655 +Query 1/1: Action query time = 5.041 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=202: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 4.820 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=218: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 5.037 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9707 +t=234: Selected seed 195 with value = 0.9707 +Query 1/1: Action query time = 4.893 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9376 +t=250: Selected seed 195 with value = 0.9376 +Query 1/1: Action query time = 4.694 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8964 +t=266: Selected seed 195 with value = 0.8964 +Query 1/1: Action query time = 3.310 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8949 +t=282: Selected seed 195 with value = 0.8949 +Query 1/1: Action query time = 4.267 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9135 +t=298: Selected seed 195 with value = 0.9135 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s01/2026_08_02-09_05_11--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 5.317 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2188 +t=10: Selected seed 195 with value = 0.2188 +Query 1/1: Action query time = 5.282 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3072 +t=26: Selected seed 195 with value = 0.3072 +Query 1/1: Action query time = 5.337 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2590 +t=42: Selected seed 195 with value = 0.2590 +Query 1/1: Action query time = 5.194 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3099 +t=58: Selected seed 195 with value = 0.3099 +Query 1/1: Action query time = 5.441 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4232 +t=74: Selected seed 195 with value = 0.4232 +Query 1/1: Action query time = 5.460 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5436 +t=90: Selected seed 195 with value = 0.5436 +Query 1/1: Action query time = 5.492 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5383 +t=106: Selected seed 195 with value = 0.5383 +Query 1/1: Action query time = 5.277 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3894 +t=122: Selected seed 195 with value = 0.3894 +Query 1/1: Action query time = 5.035 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6893 +t=138: Selected seed 195 with value = 0.6893 +Query 1/1: Action query time = 4.994 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7229 +t=154: Selected seed 195 with value = 0.7229 +Query 1/1: Action query time = 5.044 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8828 +t=170: Selected seed 195 with value = 0.8828 +Query 1/1: Action query time = 5.086 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8482 +t=186: Selected seed 195 with value = 0.8482 +Query 1/1: Action query time = 5.059 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9333 +t=202: Selected seed 195 with value = 0.9333 +Query 1/1: Action query time = 4.931 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9605 +t=218: Selected seed 195 with value = 0.9605 +Query 1/1: Action query time = 4.747 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.964 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9882 +t=250: Selected seed 195 with value = 0.9882 +Query 1/1: Action query time = 4.106 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9649 +t=266: Selected seed 195 with value = 0.9649 +Query 1/1: Action query time = 3.933 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9298 +t=282: Selected seed 195 with value = 0.9298 +Query 1/1: Action query time = 1.701 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8544 +t=298: Selected seed 195 with value = 0.8544 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s01/2026_08_02-09_05_11--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 1.257 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2261 +t=10: Selected seed 195 with value = 0.2261 +Query 1/1: Action query time = 1.184 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3279 +t=26: Selected seed 195 with value = 0.3279 +Query 1/1: Action query time = 1.197 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3123 +t=42: Selected seed 195 with value = 0.3123 +Query 1/1: Action query time = 1.206 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3633 +t=58: Selected seed 195 with value = 0.3633 +Query 1/1: Action query time = 1.208 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3966 +t=74: Selected seed 195 with value = 0.3966 +Query 1/1: Action query time = 1.176 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7469 +t=90: Selected seed 195 with value = 0.7469 +Query 1/1: Action query time = 1.223 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8267 +t=106: Selected seed 195 with value = 0.8267 +Query 1/1: Action query time = 1.218 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5986 +t=122: Selected seed 195 with value = 0.5986 +Query 1/1: Action query time = 1.182 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6981 +t=138: Selected seed 195 with value = 0.6981 +Query 1/1: Action query time = 1.289 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7797 +t=154: Selected seed 195 with value = 0.7797 +Query 1/1: Action query time = 1.241 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8861 +t=170: Selected seed 195 with value = 0.8861 +Query 1/1: Action query time = 1.236 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.171 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9717 +t=202: Selected seed 195 with value = 0.9717 +Query 1/1: Action query time = 1.220 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9798 +t=218: Selected seed 195 with value = 0.9798 +Query 1/1: Action query time = 1.290 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=234: Selected seed 195 with value = 0.9926 +Query 1/1: Action query time = 1.247 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9600 +t=250: Selected seed 195 with value = 0.9600 +Query 1/1: Action query time = 1.224 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9299 +t=266: Selected seed 195 with value = 0.9299 +Query 1/1: Action query time = 1.271 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9368 +t=282: Selected seed 195 with value = 0.9368 +Query 1/1: Action query time = 1.077 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9617 +t=298: Selected seed 195 with value = 0.9617 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s01/2026_08_02-09_05_11--with_future_img--episode=4--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 4 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s04.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s04.txt new file mode 100644 index 0000000000000000000000000000000000000000..b8e0f61e465031931e50eeeb9c8a9f4bf423f112 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s04.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t3_s04', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='4,20,36', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 2.709 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2126 +t=10: Selected seed 195 with value = 0.2126 +Query 1/1: Action query time = 3.442 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2087 +t=26: Selected seed 195 with value = 0.2087 +Query 1/1: Action query time = 5.070 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2304 +t=42: Selected seed 195 with value = 0.2304 +Query 1/1: Action query time = 5.874 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2559 +t=58: Selected seed 195 with value = 0.2559 +Query 1/1: Action query time = 5.292 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2671 +t=74: Selected seed 195 with value = 0.2671 +Query 1/1: Action query time = 3.810 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2476 +t=90: Selected seed 195 with value = 0.2476 +Query 1/1: Action query time = 4.839 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2487 +t=106: Selected seed 195 with value = 0.2487 +Query 1/1: Action query time = 5.469 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3251 +t=122: Selected seed 195 with value = 0.3251 +Query 1/1: Action query time = 5.643 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5373 +t=138: Selected seed 195 with value = 0.5373 +Query 1/1: Action query time = 4.995 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6139 +t=154: Selected seed 195 with value = 0.6139 +Query 1/1: Action query time = 5.137 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5538 +t=170: Selected seed 195 with value = 0.5538 +Query 1/1: Action query time = 5.110 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8826 +t=186: Selected seed 195 with value = 0.8826 +Query 1/1: Action query time = 5.113 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9610 +t=202: Selected seed 195 with value = 0.9610 +Query 1/1: Action query time = 5.440 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9523 +t=218: Selected seed 195 with value = 0.9523 +Query 1/1: Action query time = 5.464 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7939 +t=234: Selected seed 195 with value = 0.7939 +Query 1/1: Action query time = 4.950 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8037 +t=250: Selected seed 195 with value = 0.8037 +Query 1/1: Action query time = 4.840 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7683 +t=266: Selected seed 195 with value = 0.7683 +Query 1/1: Action query time = 4.895 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8700 +t=282: Selected seed 195 with value = 0.8700 +Query 1/1: Action query time = 4.877 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8760 +t=298: Selected seed 195 with value = 0.8760 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s04/2026_08_02-09_05_11--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 1.311 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2160 +t=10: Selected seed 195 with value = 0.2160 +Query 1/1: Action query time = 1.605 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2324 +t=26: Selected seed 195 with value = 0.2324 +Query 1/1: Action query time = 1.674 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2245 +t=42: Selected seed 195 with value = 0.2245 +Query 1/1: Action query time = 4.685 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2014 +t=58: Selected seed 195 with value = 0.2014 +Query 1/1: Action query time = 5.125 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2842 +t=74: Selected seed 195 with value = 0.2842 +Query 1/1: Action query time = 5.448 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2887 +t=90: Selected seed 195 with value = 0.2887 +Query 1/1: Action query time = 4.697 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3395 +t=106: Selected seed 195 with value = 0.3395 +Query 1/1: Action query time = 5.071 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4440 +t=122: Selected seed 195 with value = 0.4440 +Query 1/1: Action query time = 5.155 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5906 +t=138: Selected seed 195 with value = 0.5906 +Query 1/1: Action query time = 5.134 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5482 +t=154: Selected seed 195 with value = 0.5482 +Query 1/1: Action query time = 4.894 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3502 +t=170: Selected seed 195 with value = 0.3502 +Query 1/1: Action query time = 5.102 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5044 +t=186: Selected seed 195 with value = 0.5044 +Query 1/1: Action query time = 5.092 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8702 +t=202: Selected seed 195 with value = 0.8702 +Query 1/1: Action query time = 5.218 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8311 +t=218: Selected seed 195 with value = 0.8311 +Query 1/1: Action query time = 5.184 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7216 +t=234: Selected seed 195 with value = 0.7216 +Query 1/1: Action query time = 5.283 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8922 +t=250: Selected seed 195 with value = 0.8922 +Query 1/1: Action query time = 5.079 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9455 +t=266: Selected seed 195 with value = 0.9455 +Query 1/1: Action query time = 5.031 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8921 +t=282: Selected seed 195 with value = 0.8921 +Query 1/1: Action query time = 5.078 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9107 +t=298: Selected seed 195 with value = 0.9107 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s04/2026_08_02-09_05_11--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 5.324 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2187 +t=10: Selected seed 195 with value = 0.2187 +Query 1/1: Action query time = 2.557 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3519 +t=26: Selected seed 195 with value = 0.3519 +Query 1/1: Action query time = 1.996 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3243 +t=42: Selected seed 195 with value = 0.3243 +Query 1/1: Action query time = 2.094 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3944 +t=58: Selected seed 195 with value = 0.3944 +Query 1/1: Action query time = 4.202 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5549 +t=74: Selected seed 195 with value = 0.5549 +Query 1/1: Action query time = 5.673 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6792 +t=90: Selected seed 195 with value = 0.6792 +Query 1/1: Action query time = 5.464 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6006 +t=106: Selected seed 195 with value = 0.6006 +Query 1/1: Action query time = 5.350 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7603 +t=122: Selected seed 195 with value = 0.7603 +Query 1/1: Action query time = 5.349 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8920 +t=138: Selected seed 195 with value = 0.8920 +Query 1/1: Action query time = 5.234 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9121 +t=154: Selected seed 195 with value = 0.9121 +Query 1/1: Action query time = 5.216 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9618 +t=170: Selected seed 195 with value = 0.9618 +Query 1/1: Action query time = 5.373 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9720 +t=186: Selected seed 195 with value = 0.9720 +Query 1/1: Action query time = 5.504 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9542 +t=202: Selected seed 195 with value = 0.9542 +Query 1/1: Action query time = 5.560 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9224 +t=218: Selected seed 195 with value = 0.9224 +Query 1/1: Action query time = 5.356 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9107 +t=234: Selected seed 195 with value = 0.9107 +Query 1/1: Action query time = 5.234 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9155 +t=250: Selected seed 195 with value = 0.9155 +Query 1/1: Action query time = 5.058 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9341 +t=266: Selected seed 195 with value = 0.9341 +Query 1/1: Action query time = 5.163 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9609 +t=282: Selected seed 195 with value = 0.9609 +Query 1/1: Action query time = 5.301 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9594 +t=298: Selected seed 195 with value = 0.9594 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s04/2026_08_02-09_05_11--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..c43cb11f33f211bb90d6a629f0be520594c8bad6 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_05_11--ft6100_t3_s07.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t3_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 2.949 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2062 +t=10: Selected seed 195 with value = 0.2062 +Query 1/1: Action query time = 3.491 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3029 +t=26: Selected seed 195 with value = 0.3029 +Query 1/1: Action query time = 5.007 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2957 +t=42: Selected seed 195 with value = 0.2957 +Query 1/1: Action query time = 5.884 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3780 +t=58: Selected seed 195 with value = 0.3780 +Query 1/1: Action query time = 6.215 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4514 +t=74: Selected seed 195 with value = 0.4514 +Query 1/1: Action query time = 5.624 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5690 +t=90: Selected seed 195 with value = 0.5690 +Query 1/1: Action query time = 4.779 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6622 +t=106: Selected seed 195 with value = 0.6622 +Query 1/1: Action query time = 4.828 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7986 +t=122: Selected seed 195 with value = 0.7986 +Query 1/1: Action query time = 5.312 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8495 +t=138: Selected seed 195 with value = 0.8495 +Query 1/1: Action query time = 5.214 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8933 +t=154: Selected seed 195 with value = 0.8933 +Query 1/1: Action query time = 4.999 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9200 +t=170: Selected seed 195 with value = 0.9200 +Query 1/1: Action query time = 5.039 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9841 +t=186: Selected seed 195 with value = 0.9841 +Query 1/1: Action query time = 5.087 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=202: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 4.948 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9660 +t=218: Selected seed 195 with value = 0.9660 +Query 1/1: Action query time = 5.197 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9259 +t=234: Selected seed 195 with value = 0.9259 +Query 1/1: Action query time = 5.197 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9103 +t=250: Selected seed 195 with value = 0.9103 +Query 1/1: Action query time = 5.227 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9323 +t=266: Selected seed 195 with value = 0.9323 +Query 1/1: Action query time = 5.234 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9606 +t=282: Selected seed 195 with value = 0.9606 +Query 1/1: Action query time = 4.973 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=298: Selected seed 195 with value = 0.9870 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s07/2026_08_02-09_05_11--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 1.474 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2038 +t=10: Selected seed 195 with value = 0.2038 +Query 1/1: Action query time = 2.252 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2321 +t=26: Selected seed 195 with value = 0.2321 +Query 1/1: Action query time = 4.860 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2176 +t=42: Selected seed 195 with value = 0.2176 +Query 1/1: Action query time = 4.569 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2402 +t=58: Selected seed 195 with value = 0.2402 +Query 1/1: Action query time = 5.477 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2569 +t=74: Selected seed 195 with value = 0.2569 +Query 1/1: Action query time = 5.813 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2560 +t=90: Selected seed 195 with value = 0.2560 +Query 1/1: Action query time = 5.290 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2509 +t=106: Selected seed 195 with value = 0.2509 +Query 1/1: Action query time = 4.985 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2810 +t=122: Selected seed 195 with value = 0.2810 +Query 1/1: Action query time = 4.962 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4070 +t=138: Selected seed 195 with value = 0.4070 +Query 1/1: Action query time = 5.047 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5582 +t=154: Selected seed 195 with value = 0.5582 +Query 1/1: Action query time = 5.114 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6160 +t=170: Selected seed 195 with value = 0.6160 +Query 1/1: Action query time = 5.238 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1890 +t=186: Selected seed 195 with value = 0.1890 +Query 1/1: Action query time = 5.238 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7893 +t=202: Selected seed 195 with value = 0.7893 +Query 1/1: Action query time = 5.399 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8080 +t=218: Selected seed 195 with value = 0.8080 +Query 1/1: Action query time = 5.348 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8809 +t=234: Selected seed 195 with value = 0.8809 +Query 1/1: Action query time = 5.217 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7383 +t=250: Selected seed 195 with value = 0.7383 +Query 1/1: Action query time = 5.427 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8697 +t=266: Selected seed 195 with value = 0.8697 +Query 1/1: Action query time = 5.177 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8027 +t=282: Selected seed 195 with value = 0.8027 +Query 1/1: Action query time = 4.311 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9253 +t=298: Selected seed 195 with value = 0.9253 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s07/2026_08_02-09_05_11--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 2.400 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1898 +t=10: Selected seed 195 with value = 0.1898 +Query 1/1: Action query time = 1.953 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2046 +t=26: Selected seed 195 with value = 0.2046 +Query 1/1: Action query time = 1.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2012 +t=42: Selected seed 195 with value = 0.2012 +Query 1/1: Action query time = 4.257 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2423 +t=58: Selected seed 195 with value = 0.2423 +Query 1/1: Action query time = 5.625 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2946 +t=74: Selected seed 195 with value = 0.2946 +Query 1/1: Action query time = 5.356 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3500 +t=90: Selected seed 195 with value = 0.3500 +Query 1/1: Action query time = 5.322 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5075 +t=106: Selected seed 195 with value = 0.5075 +Query 1/1: Action query time = 5.279 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5882 +t=122: Selected seed 195 with value = 0.5882 +Query 1/1: Action query time = 4.992 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6089 +t=138: Selected seed 195 with value = 0.6089 +Query 1/1: Action query time = 5.099 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7536 +t=154: Selected seed 195 with value = 0.7536 +Query 1/1: Action query time = 5.288 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8529 +t=170: Selected seed 195 with value = 0.8529 +Query 1/1: Action query time = 5.122 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9297 +t=186: Selected seed 195 with value = 0.9297 +Query 1/1: Action query time = 5.416 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9133 +t=202: Selected seed 195 with value = 0.9133 +Query 1/1: Action query time = 5.180 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9553 +t=218: Selected seed 195 with value = 0.9553 +Query 1/1: Action query time = 5.071 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9624 +t=234: Selected seed 195 with value = 0.9624 +Query 1/1: Action query time = 5.165 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9414 +t=250: Selected seed 195 with value = 0.9414 +Query 1/1: Action query time = 5.551 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9191 +t=266: Selected seed 195 with value = 0.9191 +Query 1/1: Action query time = 5.085 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9120 +t=282: Selected seed 195 with value = 0.9120 +Query 1/1: Action query time = 4.648 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9061 +t=298: Selected seed 195 with value = 0.9061 +Saved rollout MP4 at path ./rollouts/ft6100_t3_s07/2026_08_02-09_05_11--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_12_01--ft6100_t4_s03.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_12_01--ft6100_t4_s03.txt new file mode 100644 index 0000000000000000000000000000000000000000..9a95ce26de724bd535a50203faae4d15032b75da --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_12_01--ft6100_t4_s03.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t4_s03', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,19,35', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.195 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4644 +t=10: Selected seed 195 with value = 0.4644 +Query 1/1: Action query time = 5.284 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5262 +t=26: Selected seed 195 with value = 0.5262 +Query 1/1: Action query time = 4.653 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5964 +t=42: Selected seed 195 with value = 0.5964 +Query 1/1: Action query time = 4.854 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5902 +t=58: Selected seed 195 with value = 0.5902 +Query 1/1: Action query time = 4.944 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6144 +t=74: Selected seed 195 with value = 0.6144 +Query 1/1: Action query time = 5.045 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6536 +t=90: Selected seed 195 with value = 0.6536 +Query 1/1: Action query time = 5.484 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6752 +t=106: Selected seed 195 with value = 0.6752 +Query 1/1: Action query time = 5.301 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6390 +t=122: Selected seed 195 with value = 0.6390 +Query 1/1: Action query time = 5.317 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6785 +t=138: Selected seed 195 with value = 0.6785 +Query 1/1: Action query time = 4.726 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6324 +t=154: Selected seed 195 with value = 0.6324 +Query 1/1: Action query time = 4.803 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6486 +t=170: Selected seed 195 with value = 0.6486 +Query 1/1: Action query time = 5.009 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6149 +t=186: Selected seed 195 with value = 0.6149 +Query 1/1: Action query time = 5.392 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6176 +t=202: Selected seed 195 with value = 0.6176 +Query 1/1: Action query time = 5.255 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5824 +t=218: Selected seed 195 with value = 0.5824 +Query 1/1: Action query time = 5.174 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2955 +t=234: Selected seed 195 with value = 0.2955 +Query 1/1: Action query time = 5.415 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5262 +t=250: Selected seed 195 with value = 0.5262 +Query 1/1: Action query time = 5.204 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6167 +t=266: Selected seed 195 with value = 0.6167 +Query 1/1: Action query time = 4.990 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5484 +t=282: Selected seed 195 with value = 0.5484 +Query 1/1: Action query time = 4.182 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6458 +t=298: Selected seed 195 with value = 0.6458 +Saved rollout MP4 at path ./rollouts/ft6100_t4_s03/2026_08_02-09_12_01--with_future_img--episode=1--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.177 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4710 +t=10: Selected seed 195 with value = 0.4710 +Query 1/1: Action query time = 5.391 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5206 +t=26: Selected seed 195 with value = 0.5206 +Query 1/1: Action query time = 5.898 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5947 +t=42: Selected seed 195 with value = 0.5947 +Query 1/1: Action query time = 5.279 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7132 +t=58: Selected seed 195 with value = 0.7132 +Query 1/1: Action query time = 5.411 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8774 +t=74: Selected seed 195 with value = 0.8774 +Query 1/1: Action query time = 5.297 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9632 +t=90: Selected seed 195 with value = 0.9632 +Query 1/1: Action query time = 5.602 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=106: Selected seed 195 with value = 0.9955 +Query 1/1: Action query time = 5.187 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9878 +t=122: Selected seed 195 with value = 0.9878 +Query 1/1: Action query time = 4.757 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9736 +t=138: Selected seed 195 with value = 0.9736 +Query 1/1: Action query time = 4.725 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9616 +t=154: Selected seed 195 with value = 0.9616 +Query 1/1: Action query time = 4.096 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9441 +t=170: Selected seed 195 with value = 0.9441 +Query 1/1: Action query time = 6.160 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9653 +t=186: Selected seed 195 with value = 0.9653 +Query 1/1: Action query time = 5.817 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9562 +t=202: Selected seed 195 with value = 0.9562 +Query 1/1: Action query time = 5.293 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9278 +t=218: Selected seed 195 with value = 0.9278 +Query 1/1: Action query time = 4.792 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9059 +t=234: Selected seed 195 with value = 0.9059 +Query 1/1: Action query time = 4.928 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8916 +t=250: Selected seed 195 with value = 0.8916 +Query 1/1: Action query time = 5.003 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8888 +t=266: Selected seed 195 with value = 0.8888 +Query 1/1: Action query time = 5.335 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8942 +t=282: Selected seed 195 with value = 0.8942 +Query 1/1: Action query time = 4.118 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9088 +t=298: Selected seed 195 with value = 0.9088 +Saved rollout MP4 at path ./rollouts/ft6100_t4_s03/2026_08_02-09_12_01--with_future_img--episode=2--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.662 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4944 +t=10: Selected seed 195 with value = 0.4944 +Query 1/1: Action query time = 5.092 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5048 +t=26: Selected seed 195 with value = 0.5048 +Query 1/1: Action query time = 4.937 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5975 +t=42: Selected seed 195 with value = 0.5975 +Query 1/1: Action query time = 4.929 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7151 +t=58: Selected seed 195 with value = 0.7151 +Query 1/1: Action query time = 4.894 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8763 +t=74: Selected seed 195 with value = 0.8763 +Query 1/1: Action query time = 5.268 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9622 +t=90: Selected seed 195 with value = 0.9622 +Query 1/1: Action query time = 5.270 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=106: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 5.659 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9831 +t=122: Selected seed 195 with value = 0.9831 +Query 1/1: Action query time = 5.246 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9519 +t=138: Selected seed 195 with value = 0.9519 +Query 1/1: Action query time = 4.850 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9211 +t=154: Selected seed 195 with value = 0.9211 +Query 1/1: Action query time = 5.085 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9336 +t=170: Selected seed 195 with value = 0.9336 +Query 1/1: Action query time = 5.386 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9415 +t=186: Selected seed 195 with value = 0.9415 +Query 1/1: Action query time = 5.344 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9431 +t=202: Selected seed 195 with value = 0.9431 +Query 1/1: Action query time = 5.160 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9326 +t=218: Selected seed 195 with value = 0.9326 +Query 1/1: Action query time = 4.281 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9172 +t=234: Selected seed 195 with value = 0.9172 +Query 1/1: Action query time = 4.585 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9276 +t=250: Selected seed 195 with value = 0.9276 +Query 1/1: Action query time = 4.442 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9071 +t=266: Selected seed 195 with value = 0.9071 +Query 1/1: Action query time = 4.612 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9108 +t=282: Selected seed 195 with value = 0.9108 +Query 1/1: Action query time = 4.114 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9127 +t=298: Selected seed 195 with value = 0.9127 +Saved rollout MP4 at path ./rollouts/ft6100_t4_s03/2026_08_02-09_12_01--with_future_img--episode=3--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_12_01--ft6100_t4_s04.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_12_01--ft6100_t4_s04.txt new file mode 100644 index 0000000000000000000000000000000000000000..19fb204e7bc5f9080359a8099e42d9e86f36f502 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_12_01--ft6100_t4_s04.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t4_s04', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='4,20,36', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.380 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4703 +t=10: Selected seed 195 with value = 0.4703 +Query 1/1: Action query time = 4.966 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5434 +t=26: Selected seed 195 with value = 0.5434 +Query 1/1: Action query time = 5.931 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6020 +t=42: Selected seed 195 with value = 0.6020 +Query 1/1: Action query time = 5.013 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6139 +t=58: Selected seed 195 with value = 0.6139 +Query 1/1: Action query time = 4.656 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9380 +t=74: Selected seed 195 with value = 0.9380 +Query 1/1: Action query time = 4.674 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9709 +t=90: Selected seed 195 with value = 0.9709 +Query 1/1: Action query time = 5.098 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9533 +t=106: Selected seed 195 with value = 0.9533 +Query 1/1: Action query time = 5.351 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9659 +t=122: Selected seed 195 with value = 0.9659 +Query 1/1: Action query time = 5.270 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9538 +t=138: Selected seed 195 with value = 0.9538 +Query 1/1: Action query time = 5.157 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9535 +t=154: Selected seed 195 with value = 0.9535 +Query 1/1: Action query time = 5.215 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9456 +t=170: Selected seed 195 with value = 0.9456 +Query 1/1: Action query time = 5.179 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9074 +t=186: Selected seed 195 with value = 0.9074 +Query 1/1: Action query time = 5.470 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9199 +t=202: Selected seed 195 with value = 0.9199 +Query 1/1: Action query time = 5.397 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9142 +t=218: Selected seed 195 with value = 0.9142 +Query 1/1: Action query time = 4.972 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8749 +t=234: Selected seed 195 with value = 0.8749 +Query 1/1: Action query time = 5.142 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8290 +t=250: Selected seed 195 with value = 0.8290 +Query 1/1: Action query time = 5.435 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8266 +t=266: Selected seed 195 with value = 0.8266 +Query 1/1: Action query time = 4.819 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8986 +t=282: Selected seed 195 with value = 0.8986 +Query 1/1: Action query time = 4.141 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6100_t4_s04/2026_08_02-09_12_01--with_future_img--episode=1--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.290 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4877 +t=10: Selected seed 195 with value = 0.4877 +Query 1/1: Action query time = 5.031 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5176 +t=26: Selected seed 195 with value = 0.5176 +Query 1/1: Action query time = 5.328 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5990 +t=42: Selected seed 195 with value = 0.5990 +Query 1/1: Action query time = 5.304 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7329 +t=58: Selected seed 195 with value = 0.7329 +Query 1/1: Action query time = 5.374 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8993 +t=74: Selected seed 195 with value = 0.8993 +Query 1/1: Action query time = 5.307 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=90: Selected seed 195 with value = 0.9871 +Query 1/1: Action query time = 5.300 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=106: Selected seed 195 with value = 0.9955 +Query 1/1: Action query time = 5.379 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9881 +t=122: Selected seed 195 with value = 0.9881 +Query 1/1: Action query time = 5.592 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9789 +t=138: Selected seed 195 with value = 0.9789 +Query 1/1: Action query time = 6.049 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9674 +t=154: Selected seed 195 with value = 0.9674 +Query 1/1: Action query time = 5.339 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9447 +t=170: Selected seed 195 with value = 0.9447 +Query 1/1: Action query time = 4.656 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9132 +t=186: Selected seed 195 with value = 0.9132 +Query 1/1: Action query time = 4.531 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7894 +t=202: Selected seed 195 with value = 0.7894 +Query 1/1: Action query time = 5.188 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8753 +t=218: Selected seed 195 with value = 0.8753 +Query 1/1: Action query time = 5.811 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9350 +t=234: Selected seed 195 with value = 0.9350 +Query 1/1: Action query time = 4.270 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9780 +t=250: Selected seed 195 with value = 0.9780 +Query 1/1: Action query time = 3.998 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9564 +t=266: Selected seed 195 with value = 0.9564 +Query 1/1: Action query time = 4.867 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9439 +t=282: Selected seed 195 with value = 0.9439 +Query 1/1: Action query time = 4.510 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9287 +t=298: Selected seed 195 with value = 0.9287 +Saved rollout MP4 at path ./rollouts/ft6100_t4_s04/2026_08_02-09_12_01--with_future_img--episode=2--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.851 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4703 +t=10: Selected seed 195 with value = 0.4703 +Query 1/1: Action query time = 5.337 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5173 +t=26: Selected seed 195 with value = 0.5173 +Query 1/1: Action query time = 4.834 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5896 +t=42: Selected seed 195 with value = 0.5896 +Query 1/1: Action query time = 4.835 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5619 +t=58: Selected seed 195 with value = 0.5619 +Query 1/1: Action query time = 5.236 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6014 +t=74: Selected seed 195 with value = 0.6014 +Query 1/1: Action query time = 5.382 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6088 +t=90: Selected seed 195 with value = 0.6088 +Query 1/1: Action query time = 5.499 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5916 +t=106: Selected seed 195 with value = 0.5916 +Query 1/1: Action query time = 5.222 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6582 +t=122: Selected seed 195 with value = 0.6582 +Query 1/1: Action query time = 5.285 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8104 +t=138: Selected seed 195 with value = 0.8104 +Query 1/1: Action query time = 5.527 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6393 +t=154: Selected seed 195 with value = 0.6393 +Query 1/1: Action query time = 4.836 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8363 +t=170: Selected seed 195 with value = 0.8363 +Query 1/1: Action query time = 4.787 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6763 +t=186: Selected seed 195 with value = 0.6763 +Query 1/1: Action query time = 5.123 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6265 +t=202: Selected seed 195 with value = 0.6265 +Query 1/1: Action query time = 5.583 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6349 +t=218: Selected seed 195 with value = 0.6349 +Query 1/1: Action query time = 4.291 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6466 +t=234: Selected seed 195 with value = 0.6466 +Query 1/1: Action query time = 4.569 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6049 +t=250: Selected seed 195 with value = 0.6049 +Query 1/1: Action query time = 4.310 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6709 +t=266: Selected seed 195 with value = 0.6709 +Query 1/1: Action query time = 4.500 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5655 +t=282: Selected seed 195 with value = 0.5655 +Query 1/1: Action query time = 3.782 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5827 +t=298: Selected seed 195 with value = 0.5827 +Saved rollout MP4 at path ./rollouts/ft6100_t4_s04/2026_08_02-09_12_01--with_future_img--episode=3--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_12_03--ft6100_t4_s13.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_12_03--ft6100_t4_s13.txt new file mode 100644 index 0000000000000000000000000000000000000000..f5538dc9f338c6d744122940eff5434d88ae0e80 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_12_03--ft6100_t4_s13.txt @@ -0,0 +1,261 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t4_s13', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='13,29,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.298 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4806 +t=10: Selected seed 195 with value = 0.4806 +Query 1/1: Action query time = 5.394 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5294 +t=26: Selected seed 195 with value = 0.5294 +Query 1/1: Action query time = 4.906 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6121 +t=42: Selected seed 195 with value = 0.6121 +Query 1/1: Action query time = 4.398 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5915 +t=58: Selected seed 195 with value = 0.5915 +Query 1/1: Action query time = 4.668 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6372 +t=74: Selected seed 195 with value = 0.6372 +Query 1/1: Action query time = 5.370 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7130 +t=90: Selected seed 195 with value = 0.7130 +Query 1/1: Action query time = 5.276 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9646 +t=106: Selected seed 195 with value = 0.9646 +Query 1/1: Action query time = 5.062 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8961 +t=122: Selected seed 195 with value = 0.8961 +Query 1/1: Action query time = 4.776 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9714 +t=138: Selected seed 195 with value = 0.9714 +Query 1/1: Action query time = 4.810 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9587 +t=154: Selected seed 195 with value = 0.9587 +Query 1/1: Action query time = 5.380 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9706 +t=170: Selected seed 195 with value = 0.9706 +Query 1/1: Action query time = 5.569 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9693 +t=186: Selected seed 195 with value = 0.9693 +Query 1/1: Action query time = 5.294 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9645 +t=202: Selected seed 195 with value = 0.9645 +Query 1/1: Action query time = 5.000 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9596 +t=218: Selected seed 195 with value = 0.9596 +Query 1/1: Action query time = 5.005 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9362 +t=234: Selected seed 195 with value = 0.9362 +Query 1/1: Action query time = 5.202 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9335 +t=250: Selected seed 195 with value = 0.9335 +Query 1/1: Action query time = 5.474 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9090 +t=266: Selected seed 195 with value = 0.9090 +Query 1/1: Action query time = 4.819 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8595 +t=282: Selected seed 195 with value = 0.8595 +Query 1/1: Action query time = 1.656 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9109 +t=298: Selected seed 195 with value = 0.9109 +Saved rollout MP4 at path ./rollouts/ft6100_t4_s13/2026_08_02-09_12_03--with_future_img--episode=1--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 6.343 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4939 +t=10: Selected seed 195 with value = 0.4939 +Query 1/1: Action query time = 4.167 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5233 +t=26: Selected seed 195 with value = 0.5233 +Query 1/1: Action query time = 4.987 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5991 +t=42: Selected seed 195 with value = 0.5991 +Query 1/1: Action query time = 4.987 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5480 +t=58: Selected seed 195 with value = 0.5480 +Query 1/1: Action query time = 5.075 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5864 +t=74: Selected seed 195 with value = 0.5864 +Query 1/1: Action query time = 5.409 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5927 +t=90: Selected seed 195 with value = 0.5927 +Query 1/1: Action query time = 5.297 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7236 +t=106: Selected seed 195 with value = 0.7236 +Query 1/1: Action query time = 5.224 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7290 +t=122: Selected seed 195 with value = 0.7290 +Query 1/1: Action query time = 5.339 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8862 +t=138: Selected seed 195 with value = 0.8862 +Query 1/1: Action query time = 4.899 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6816 +t=154: Selected seed 195 with value = 0.6816 +Query 1/1: Action query time = 5.198 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6503 +t=170: Selected seed 195 with value = 0.6503 +Query 1/1: Action query time = 4.760 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6806 +t=186: Selected seed 195 with value = 0.6806 +Query 1/1: Action query time = 4.925 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6497 +t=202: Selected seed 195 with value = 0.6497 +Query 1/1: Action query time = 5.226 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6199 +t=218: Selected seed 195 with value = 0.6199 +Query 1/1: Action query time = 5.623 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6370 +t=234: Selected seed 195 with value = 0.6370 +Query 1/1: Action query time = 4.862 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6214 +t=250: Selected seed 195 with value = 0.6214 +Query 1/1: Action query time = 3.915 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6304 +t=266: Selected seed 195 with value = 0.6304 +Query 1/1: Action query time = 4.812 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6265 +t=282: Selected seed 195 with value = 0.6265 +Query 1/1: Action query time = 3.115 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6541 +t=298: Selected seed 195 with value = 0.6541 +Saved rollout MP4 at path ./rollouts/ft6100_t4_s13/2026_08_02-09_12_03--with_future_img--episode=2--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.713 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4812 +t=10: Selected seed 195 with value = 0.4812 +Query 1/1: Action query time = 5.525 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5179 +t=26: Selected seed 195 with value = 0.5179 +Query 1/1: Action query time = 5.039 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6141 +t=42: Selected seed 195 with value = 0.6141 +Query 1/1: Action query time = 4.494 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6902 +t=58: Selected seed 195 with value = 0.6902 +Query 1/1: Action query time = 4.330 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8760 +t=74: Selected seed 195 with value = 0.8760 +Query 1/1: Action query time = 2.955 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9116 +t=90: Selected seed 195 with value = 0.9116 +Query 1/1: Action query time = 3.548 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9974 +t=106: Selected seed 195 with value = 0.9974 +Query 1/1: Action query time = 4.173 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9875 +t=122: Selected seed 195 with value = 0.9875 +Query 1/1: Action query time = 4.191 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9681 +t=138: Selected seed 195 with value = 0.9681 +Query 1/1: Action query time = 3.637 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9546 +t=154: Selected seed 195 with value = 0.9546 +Query 1/1: Action query time = 3.653 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9468 +t=170: Selected seed 195 with value = 0.9468 +Query 1/1: Action query time = 4.124 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9414 +t=186: Selected seed 195 with value = 0.9414 +Query 1/1: Action query time = 4.277 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9414 +t=202: Selected seed 195 with value = 0.9414 +Query 1/1: Action query time = 3.458 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9467 +t=218: Selected seed 195 with value = 0.9467 +Query 1/1: Action query time = 2.987 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9451 +t=234: Selected seed 195 with value = 0.9451 +Query 1/1: Action query time = 2.401 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9428 +t=250: Selected seed 195 with value = 0.9428 +Query 1/1: Action query time = 3.226 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9416 +t=266: Selected seed 195 with value = 0.9416 +Query 1/1: Action query time = 3.742 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9406 +t=282: Selected seed 195 with value = 0.9406 +Query 1/1: Action query time = 1.962 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9436 +t=298: Selected seed 195 with value = 0.9436 +Saved rollout MP4 at path ./rollouts/ft6100_t4_s13/2026_08_02-09_12_03--with_future_img--episode=3--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 3 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_18_46--ft6100_t5_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_18_46--ft6100_t5_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..92818b63d7e87d7ddadce4eb74ac8e01aabd1224 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_18_46--ft6100_t5_s01.txt @@ -0,0 +1,296 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t5_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 6.424 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2752 +t=10: Selected seed 195 with value = 0.2752 +Query 1/1: Action query time = 5.199 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3540 +t=26: Selected seed 195 with value = 0.3540 +Query 1/1: Action query time = 5.421 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6835 +t=42: Selected seed 195 with value = 0.6835 +Query 1/1: Action query time = 5.193 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3590 +t=58: Selected seed 195 with value = 0.3590 +Query 1/1: Action query time = 5.437 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7567 +t=74: Selected seed 195 with value = 0.7567 +Query 1/1: Action query time = 5.095 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7105 +t=90: Selected seed 195 with value = 0.7105 +Query 1/1: Action query time = 5.031 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5387 +t=106: Selected seed 195 with value = 0.5387 +Query 1/1: Action query time = 5.086 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8056 +t=122: Selected seed 195 with value = 0.8056 +Query 1/1: Action query time = 4.022 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5025 +t=138: Selected seed 195 with value = 0.5025 +Query 1/1: Action query time = 4.542 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8071 +t=154: Selected seed 195 with value = 0.8071 +Query 1/1: Action query time = 5.029 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5440 +t=170: Selected seed 195 with value = 0.5440 +Query 1/1: Action query time = 4.747 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7432 +t=186: Selected seed 195 with value = 0.7432 +Query 1/1: Action query time = 4.731 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5078 +t=202: Selected seed 195 with value = 0.5078 +Query 1/1: Action query time = 4.183 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7215 +t=218: Selected seed 195 with value = 0.7215 +Query 1/1: Action query time = 4.852 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4837 +t=234: Selected seed 195 with value = 0.4837 +Query 1/1: Action query time = 5.268 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7690 +t=250: Selected seed 195 with value = 0.7690 +Query 1/1: Action query time = 5.432 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4628 +t=266: Selected seed 195 with value = 0.4628 +Query 1/1: Action query time = 3.635 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8200 +t=282: Selected seed 195 with value = 0.8200 +Query 1/1: Action query time = 2.899 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4164 +t=298: Selected seed 195 with value = 0.4164 +Saved rollout MP4 at path ./rollouts/ft6100_t5_s01/2026_08_02-09_18_46--with_future_img--episode=1--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 5.215 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2717 +t=10: Selected seed 195 with value = 0.2717 +Query 1/1: Action query time = 4.806 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3524 +t=26: Selected seed 195 with value = 0.3524 +Query 1/1: Action query time = 5.036 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6780 +t=42: Selected seed 195 with value = 0.6780 +Query 1/1: Action query time = 5.116 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4659 +t=58: Selected seed 195 with value = 0.4659 +Query 1/1: Action query time = 5.117 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7013 +t=74: Selected seed 195 with value = 0.7013 +Query 1/1: Action query time = 5.225 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9195 +t=90: Selected seed 195 with value = 0.9195 +Query 1/1: Action query time = 5.073 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9795 +t=106: Selected seed 195 with value = 0.9795 +Query 1/1: Action query time = 5.023 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9379 +t=122: Selected seed 195 with value = 0.9379 +Query 1/1: Action query time = 4.870 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9581 +t=138: Selected seed 195 with value = 0.9581 +Query 1/1: Action query time = 4.951 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9831 +t=154: Selected seed 195 with value = 0.9831 +Query 1/1: Action query time = 5.018 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=170: Selected seed 195 with value = 0.9971 +Query 1/1: Action query time = 4.891 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=186: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 4.531 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9905 +t=202: Selected seed 195 with value = 0.9905 +Query 1/1: Action query time = 4.896 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9754 +t=218: Selected seed 195 with value = 0.9754 +Query 1/1: Action query time = 5.212 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7749 +t=234: Selected seed 195 with value = 0.7749 +Query 1/1: Action query time = 5.401 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7231 +t=250: Selected seed 195 with value = 0.7231 +Query 1/1: Action query time = 4.856 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8648 +t=266: Selected seed 195 with value = 0.8648 +Query 1/1: Action query time = 3.012 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9275 +t=282: Selected seed 195 with value = 0.9275 +Query 1/1: Action query time = 2.010 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9449 +t=298: Selected seed 195 with value = 0.9449 +Saved rollout MP4 at path ./rollouts/ft6100_t5_s01/2026_08_02-09_18_46--with_future_img--episode=2--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 4.440 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2680 +t=10: Selected seed 195 with value = 0.2680 +Query 1/1: Action query time = 4.425 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3536 +t=26: Selected seed 195 with value = 0.3536 +Query 1/1: Action query time = 4.407 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6238 +t=42: Selected seed 195 with value = 0.6238 +Query 1/1: Action query time = 4.377 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3436 +t=58: Selected seed 195 with value = 0.3436 +Query 1/1: Action query time = 4.397 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7627 +t=74: Selected seed 195 with value = 0.7627 +Query 1/1: Action query time = 4.422 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7005 +t=90: Selected seed 195 with value = 0.7005 +Query 1/1: Action query time = 3.706 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5377 +t=106: Selected seed 195 with value = 0.5377 +Query 1/1: Action query time = 3.032 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7597 +t=122: Selected seed 195 with value = 0.7597 +Query 1/1: Action query time = 3.237 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4343 +t=138: Selected seed 195 with value = 0.4343 +Query 1/1: Action query time = 3.705 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8733 +t=154: Selected seed 195 with value = 0.8733 +Query 1/1: Action query time = 3.740 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4208 +t=170: Selected seed 195 with value = 0.4208 +Query 1/1: Action query time = 3.255 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8641 +t=186: Selected seed 195 with value = 0.8641 +Query 1/1: Action query time = 3.206 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8735 +t=202: Selected seed 195 with value = 0.8735 +Query 1/1: Action query time = 3.305 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7760 +t=218: Selected seed 195 with value = 0.7760 +Query 1/1: Action query time = 3.365 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4434 +t=234: Selected seed 195 with value = 0.4434 +Query 1/1: Action query time = 3.346 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7661 +t=250: Selected seed 195 with value = 0.7661 +Query 1/1: Action query time = 3.346 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4674 +t=266: Selected seed 195 with value = 0.4674 +Query 1/1: Action query time = 3.310 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7464 +t=282: Selected seed 195 with value = 0.7464 +Query 1/1: Action query time = 1.363 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4561 +t=298: Selected seed 195 with value = 0.4561 +Saved rollout MP4 at path ./rollouts/ft6100_t5_s01/2026_08_02-09_18_46--with_future_img--episode=3--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 1.259 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3298 +t=10: Selected seed 195 with value = 0.3298 +Query 1/1: Action query time = 1.256 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3916 +t=26: Selected seed 195 with value = 0.3916 +Query 1/1: Action query time = 1.215 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6319 +t=42: Selected seed 195 with value = 0.6319 +Query 1/1: Action query time = 1.193 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6849 +t=58: Selected seed 195 with value = 0.6849 +Query 1/1: Action query time = 0.967 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7336 +t=74: Selected seed 195 with value = 0.7336 +Query 1/1: Action query time = 0.979 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8457 +t=90: Selected seed 195 with value = 0.8457 +Query 1/1: Action query time = 0.960 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/ft6100_t5_s01/2026_08_02-09_18_46--with_future_img--episode=4--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 4 +# successes: 1 (25.0%) +Current task success rate: 0.25 +Current total success rate: 0.25 +Final results: +Total episodes: 4 +Total successes: 1 +Overall success rate: 0.2500 (25.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_18_46--ft6100_t5_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_18_46--ft6100_t5_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..a78550aa67fe8472170b265bc3f3a24f387d406a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_18_46--ft6100_t5_s02.txt @@ -0,0 +1,177 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t5_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 3.259 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3379 +t=10: Selected seed 195 with value = 0.3379 +Query 1/1: Action query time = 4.959 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3699 +t=26: Selected seed 195 with value = 0.3699 +Query 1/1: Action query time = 5.659 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6318 +t=42: Selected seed 195 with value = 0.6318 +Query 1/1: Action query time = 5.355 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3116 +t=58: Selected seed 195 with value = 0.3116 +Query 1/1: Action query time = 4.832 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4192 +t=74: Selected seed 195 with value = 0.4192 +Query 1/1: Action query time = 4.898 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6466 +t=90: Selected seed 195 with value = 0.6466 +Query 1/1: Action query time = 5.120 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8163 +t=106: Selected seed 195 with value = 0.8163 +Query 1/1: Action query time = 5.060 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6897 +t=122: Selected seed 195 with value = 0.6897 +Query 1/1: Action query time = 5.051 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7686 +t=138: Selected seed 195 with value = 0.7686 +Query 1/1: Action query time = 4.936 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8543 +t=154: Selected seed 195 with value = 0.8543 +Saved rollout MP4 at path ./rollouts/ft6100_t5_s02/2026_08_02-09_18_46--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.958 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3191 +t=10: Selected seed 195 with value = 0.3191 +Query 1/1: Action query time = 5.131 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3291 +t=26: Selected seed 195 with value = 0.3291 +Query 1/1: Action query time = 5.180 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5810 +t=42: Selected seed 195 with value = 0.5810 +Query 1/1: Action query time = 4.938 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6287 +t=58: Selected seed 195 with value = 0.6287 +Query 1/1: Action query time = 5.226 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6748 +t=74: Selected seed 195 with value = 0.6748 +Query 1/1: Action query time = 5.507 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7500 +t=90: Selected seed 195 with value = 0.7500 +Query 1/1: Action query time = 5.425 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8946 +t=106: Selected seed 195 with value = 0.8946 +Saved rollout MP4 at path ./rollouts/ft6100_t5_s02/2026_08_02-09_18_46--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 2.522 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2763 +t=10: Selected seed 195 with value = 0.2763 +Query 1/1: Action query time = 2.138 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3298 +t=26: Selected seed 195 with value = 0.3298 +Query 1/1: Action query time = 3.294 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6002 +t=42: Selected seed 195 with value = 0.6002 +Query 1/1: Action query time = 4.455 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5857 +t=58: Selected seed 195 with value = 0.5857 +Query 1/1: Action query time = 5.260 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7505 +t=74: Selected seed 195 with value = 0.7505 +Query 1/1: Action query time = 5.305 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8011 +t=90: Selected seed 195 with value = 0.8011 +Query 1/1: Action query time = 5.148 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5195 +t=106: Selected seed 195 with value = 0.5195 +Query 1/1: Action query time = 4.943 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9037 +t=122: Selected seed 195 with value = 0.9037 +Query 1/1: Action query time = 5.284 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7672 +t=138: Selected seed 195 with value = 0.7672 +Query 1/1: Action query time = 4.903 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4409 +t=154: Selected seed 195 with value = 0.4409 +Query 1/1: Action query time = 4.890 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9682 +t=170: Selected seed 195 with value = 0.9682 +Query 1/1: Action query time = 4.263 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9017 +t=186: Selected seed 195 with value = 0.9017 +Query 1/1: Action query time = 5.056 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6276 +t=202: Selected seed 195 with value = 0.6276 +Query 1/1: Action query time = 5.094 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4616 +t=218: Selected seed 195 with value = 0.4616 +Query 1/1: Action query time = 5.144 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6643 +t=234: Selected seed 195 with value = 0.6643 +Query 1/1: Action query time = 5.440 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4411 +t=250: Selected seed 195 with value = 0.4411 +Query 1/1: Action query time = 4.527 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7427 +t=266: Selected seed 195 with value = 0.7427 +Query 1/1: Action query time = 4.921 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4549 +t=282: Selected seed 195 with value = 0.4549 +Query 1/1: Action query time = 5.168 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7393 +t=298: Selected seed 195 with value = 0.7393 +Saved rollout MP4 at path ./rollouts/ft6100_t5_s02/2026_08_02-09_18_46--with_future_img--episode=3--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_41--ft6100_t6_s02.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_41--ft6100_t6_s02.txt new file mode 100644 index 0000000000000000000000000000000000000000..74b5ddc57385f4b269b33991a622461f59670be7 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_41--ft6100_t6_s02.txt @@ -0,0 +1,165 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t6_s02', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,18,34', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 3.362 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4322 +t=10: Selected seed 195 with value = 0.4322 +Query 1/1: Action query time = 5.810 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4456 +t=26: Selected seed 195 with value = 0.4456 +Query 1/1: Action query time = 4.307 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5524 +t=42: Selected seed 195 with value = 0.5524 +Query 1/1: Action query time = 4.646 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6619 +t=58: Selected seed 195 with value = 0.6619 +Query 1/1: Action query time = 5.161 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6171 +t=74: Selected seed 195 with value = 0.6171 +Query 1/1: Action query time = 5.494 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5848 +t=90: Selected seed 195 with value = 0.5848 +Query 1/1: Action query time = 4.966 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5553 +t=106: Selected seed 195 with value = 0.5553 +Query 1/1: Action query time = 4.768 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6068 +t=122: Selected seed 195 with value = 0.6068 +Query 1/1: Action query time = 5.307 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5997 +t=138: Selected seed 195 with value = 0.5997 +Query 1/1: Action query time = 5.195 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6447 +t=154: Selected seed 195 with value = 0.6447 +Query 1/1: Action query time = 5.097 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6350 +t=170: Selected seed 195 with value = 0.6350 +Query 1/1: Action query time = 4.726 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7765 +t=186: Selected seed 195 with value = 0.7765 +Query 1/1: Action query time = 5.365 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9130 +t=202: Selected seed 195 with value = 0.9130 +Query 1/1: Action query time = 4.086 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.780 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9773 +t=234: Selected seed 195 with value = 0.9773 +Query 1/1: Action query time = 6.029 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9803 +t=250: Selected seed 195 with value = 0.9803 +Query 1/1: Action query time = 5.586 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=266: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 4.554 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9967 +t=282: Selected seed 195 with value = 0.9967 +Query 1/1: Action query time = 4.126 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=298: Selected seed 195 with value = 0.9964 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s02/2026_08_02-09_24_41--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 3.239 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4661 +t=10: Selected seed 195 with value = 0.4661 +Query 1/1: Action query time = 4.297 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5126 +t=26: Selected seed 195 with value = 0.5126 +Query 1/1: Action query time = 4.870 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5865 +t=42: Selected seed 195 with value = 0.5865 +Query 1/1: Action query time = 5.090 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7279 +t=58: Selected seed 195 with value = 0.7279 +Query 1/1: Action query time = 3.395 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8476 +t=74: Selected seed 195 with value = 0.8476 +Query 1/1: Action query time = 3.488 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=90: Selected seed 195 with value = 0.9912 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s02/2026_08_02-09_24_41--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 3.229 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4528 +t=10: Selected seed 195 with value = 0.4528 +Query 1/1: Action query time = 4.047 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4678 +t=26: Selected seed 195 with value = 0.4678 +Query 1/1: Action query time = 3.876 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5744 +t=42: Selected seed 195 with value = 0.5744 +Query 1/1: Action query time = 4.495 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7303 +t=58: Selected seed 195 with value = 0.7303 +Query 1/1: Action query time = 2.739 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8457 +t=74: Selected seed 195 with value = 0.8457 +Query 1/1: Action query time = 4.634 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9828 +t=90: Selected seed 195 with value = 0.9828 +Query 1/1: Action query time = 3.979 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.443 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9853 +t=122: Selected seed 195 with value = 0.9853 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s02/2026_08_02-09_24_41--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_42--ft6100_t6_s07.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_42--ft6100_t6_s07.txt new file mode 100644 index 0000000000000000000000000000000000000000..0be8817494bcea6202ff516c4695c05499e1566b --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_42--ft6100_t6_s07.txt @@ -0,0 +1,209 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t6_s07', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,23,39', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 6.828 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4461 +t=10: Selected seed 195 with value = 0.4461 +Query 1/1: Action query time = 4.056 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4975 +t=26: Selected seed 195 with value = 0.4975 +Query 1/1: Action query time = 4.834 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5768 +t=42: Selected seed 195 with value = 0.5768 +Query 1/1: Action query time = 5.257 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7237 +t=58: Selected seed 195 with value = 0.7237 +Query 1/1: Action query time = 5.358 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8316 +t=74: Selected seed 195 with value = 0.8316 +Query 1/1: Action query time = 5.183 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9815 +t=90: Selected seed 195 with value = 0.9815 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s07/2026_08_02-09_24_42--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 5.374 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4461 +t=10: Selected seed 195 with value = 0.4461 +Query 1/1: Action query time = 5.360 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4507 +t=26: Selected seed 195 with value = 0.4507 +Query 1/1: Action query time = 5.199 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5646 +t=42: Selected seed 195 with value = 0.5646 +Query 1/1: Action query time = 4.894 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6693 +t=58: Selected seed 195 with value = 0.6693 +Query 1/1: Action query time = 4.761 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6223 +t=74: Selected seed 195 with value = 0.6223 +Query 1/1: Action query time = 4.178 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5888 +t=90: Selected seed 195 with value = 0.5888 +Query 1/1: Action query time = 5.511 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5575 +t=106: Selected seed 195 with value = 0.5575 +Query 1/1: Action query time = 4.792 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6140 +t=122: Selected seed 195 with value = 0.6140 +Query 1/1: Action query time = 4.542 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5988 +t=138: Selected seed 195 with value = 0.5988 +Query 1/1: Action query time = 5.266 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6281 +t=154: Selected seed 195 with value = 0.6281 +Query 1/1: Action query time = 5.268 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6048 +t=170: Selected seed 195 with value = 0.6048 +Query 1/1: Action query time = 2.694 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6473 +t=186: Selected seed 195 with value = 0.6473 +Query 1/1: Action query time = 1.424 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6095 +t=202: Selected seed 195 with value = 0.6095 +Query 1/1: Action query time = 1.452 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6988 +t=218: Selected seed 195 with value = 0.6988 +Query 1/1: Action query time = 2.925 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6101 +t=234: Selected seed 195 with value = 0.6101 +Query 1/1: Action query time = 5.255 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7444 +t=250: Selected seed 195 with value = 0.7444 +Query 1/1: Action query time = 4.725 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7136 +t=266: Selected seed 195 with value = 0.7136 +Query 1/1: Action query time = 3.895 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8734 +t=282: Selected seed 195 with value = 0.8734 +Query 1/1: Action query time = 4.430 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=298: Selected seed 195 with value = 0.9958 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s07/2026_08_02-09_24_42--with_future_img--episode=2--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 3.162 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4509 +t=10: Selected seed 195 with value = 0.4509 +Query 1/1: Action query time = 3.663 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4970 +t=26: Selected seed 195 with value = 0.4970 +Query 1/1: Action query time = 3.847 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5812 +t=42: Selected seed 195 with value = 0.5812 +Query 1/1: Action query time = 3.672 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6694 +t=58: Selected seed 195 with value = 0.6694 +Query 1/1: Action query time = 3.434 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6744 +t=74: Selected seed 195 with value = 0.6744 +Query 1/1: Action query time = 2.814 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7108 +t=90: Selected seed 195 with value = 0.7108 +Query 1/1: Action query time = 4.323 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6810 +t=106: Selected seed 195 with value = 0.6810 +Query 1/1: Action query time = 3.466 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9523 +t=122: Selected seed 195 with value = 0.9523 +Query 1/1: Action query time = 2.880 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9288 +t=138: Selected seed 195 with value = 0.9288 +Query 1/1: Action query time = 3.000 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=154: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 2.472 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=170: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 1.610 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=186: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 2.354 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9920 +t=202: Selected seed 195 with value = 0.9920 +Query 1/1: Action query time = 2.082 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9883 +t=218: Selected seed 195 with value = 0.9883 +Query 1/1: Action query time = 2.342 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9854 +t=234: Selected seed 195 with value = 0.9854 +Query 1/1: Action query time = 2.460 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9937 +t=250: Selected seed 195 with value = 0.9937 +Query 1/1: Action query time = 2.482 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=266: Selected seed 195 with value = 0.9924 +Query 1/1: Action query time = 2.532 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9839 +t=282: Selected seed 195 with value = 0.9839 +Query 1/1: Action query time = 2.576 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9897 +t=298: Selected seed 195 with value = 0.9897 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s07/2026_08_02-09_24_42--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) +Current task success rate: 0.3333333333333333 +Current total success rate: 0.3333333333333333 +Final results: +Total episodes: 3 +Total successes: 1 +Overall success rate: 0.3333 (33.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_42--ft6100_t6_s09.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_42--ft6100_t6_s09.txt new file mode 100644 index 0000000000000000000000000000000000000000..03ae4598b7fa3b5bfe04898c32081f2b2d3e5b84 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_42--ft6100_t6_s09.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t6_s09', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='9,25,41', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 6.150 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4633 +t=10: Selected seed 195 with value = 0.4633 +Query 1/1: Action query time = 4.229 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5161 +t=26: Selected seed 195 with value = 0.5161 +Query 1/1: Action query time = 4.749 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5768 +t=42: Selected seed 195 with value = 0.5768 +Query 1/1: Action query time = 4.947 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7404 +t=58: Selected seed 195 with value = 0.7404 +Query 1/1: Action query time = 4.886 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8296 +t=74: Selected seed 195 with value = 0.8296 +Query 1/1: Action query time = 4.093 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9690 +t=90: Selected seed 195 with value = 0.9690 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s09/2026_08_02-09_24_42--with_future_img--episode=1--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 4.981 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4493 +t=10: Selected seed 195 with value = 0.4493 +Query 1/1: Action query time = 5.017 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4939 +t=26: Selected seed 195 with value = 0.4939 +Query 1/1: Action query time = 4.992 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5734 +t=42: Selected seed 195 with value = 0.5734 +Query 1/1: Action query time = 5.153 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7304 +t=58: Selected seed 195 with value = 0.7304 +Query 1/1: Action query time = 5.316 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8438 +t=74: Selected seed 195 with value = 0.8438 +Query 1/1: Action query time = 4.664 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=90: Selected seed 195 with value = 0.9903 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s09/2026_08_02-09_24_42--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 4.652 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4433 +t=10: Selected seed 195 with value = 0.4433 +Query 1/1: Action query time = 5.454 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5005 +t=26: Selected seed 195 with value = 0.5005 +Query 1/1: Action query time = 5.743 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5703 +t=42: Selected seed 195 with value = 0.5703 +Query 1/1: Action query time = 5.659 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6847 +t=58: Selected seed 195 with value = 0.6847 +Query 1/1: Action query time = 5.114 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8401 +t=74: Selected seed 195 with value = 0.8401 +Query 1/1: Action query time = 2.131 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9578 +t=90: Selected seed 195 with value = 0.9578 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s09/2026_08_02-09_24_42--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_42--ft6100_t6_s14.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_42--ft6100_t6_s14.txt new file mode 100644 index 0000000000000000000000000000000000000000..2b5ca980912ea6d95aa313d7ffe95e6839736599 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-09_24_42--ft6100_t6_s14.txt @@ -0,0 +1,157 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_ft_from40k_2gpu_100step/checkpoints/iter_000000100/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='ft6100_t6_s14', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='14,30,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 4.451 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4517 +t=10: Selected seed 195 with value = 0.4517 +Query 1/1: Action query time = 5.054 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4877 +t=26: Selected seed 195 with value = 0.4877 +Query 1/1: Action query time = 5.278 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5684 +t=42: Selected seed 195 with value = 0.5684 +Query 1/1: Action query time = 5.652 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6292 +t=58: Selected seed 195 with value = 0.6292 +Query 1/1: Action query time = 5.137 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6888 +t=74: Selected seed 195 with value = 0.6888 +Query 1/1: Action query time = 3.647 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7270 +t=90: Selected seed 195 with value = 0.7270 +Query 1/1: Action query time = 3.860 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7373 +t=106: Selected seed 195 with value = 0.7373 +Query 1/1: Action query time = 5.233 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9878 +t=122: Selected seed 195 with value = 0.9878 +Query 1/1: Action query time = 5.148 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=138: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 5.157 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.446 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.778 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9996 +t=186: Selected seed 195 with value = 0.9996 +Query 1/1: Action query time = 4.084 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9869 +t=202: Selected seed 195 with value = 0.9869 +Query 1/1: Action query time = 3.992 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9701 +t=218: Selected seed 195 with value = 0.9701 +Query 1/1: Action query time = 4.722 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9539 +t=234: Selected seed 195 with value = 0.9539 +Query 1/1: Action query time = 5.839 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9168 +t=250: Selected seed 195 with value = 0.9168 +Query 1/1: Action query time = 5.269 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.981 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=282: Selected seed 195 with value = 0.9971 +Query 1/1: Action query time = 4.838 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=298: Selected seed 195 with value = 0.9922 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s14/2026_08_02-09_24_42--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 1.400 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4607 +t=10: Selected seed 195 with value = 0.4607 +Query 1/1: Action query time = 2.742 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5237 +t=26: Selected seed 195 with value = 0.5237 +Query 1/1: Action query time = 3.446 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5767 +t=42: Selected seed 195 with value = 0.5767 +Query 1/1: Action query time = 4.071 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7207 +t=58: Selected seed 195 with value = 0.7207 +Query 1/1: Action query time = 3.765 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8341 +t=74: Selected seed 195 with value = 0.8341 +Query 1/1: Action query time = 3.031 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9752 +t=90: Selected seed 195 with value = 0.9752 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s14/2026_08_02-09_24_42--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 2.947 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4415 +t=10: Selected seed 195 with value = 0.4415 +Query 1/1: Action query time = 3.462 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4625 +t=26: Selected seed 195 with value = 0.4625 +Query 1/1: Action query time = 3.840 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5603 +t=42: Selected seed 195 with value = 0.5603 +Query 1/1: Action query time = 3.576 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6602 +t=58: Selected seed 195 with value = 0.6602 +Query 1/1: Action query time = 3.019 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8344 +t=74: Selected seed 195 with value = 0.8344 +Query 1/1: Action query time = 3.848 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9527 +t=90: Selected seed 195 with value = 0.9527 +Saved rollout MP4 at path ./rollouts/ft6100_t6_s14/2026_08_02-09_24_42--with_future_img--episode=3--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 3 +Total successes: 2 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_32_32--base40kRE_t0_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_32_32--base40kRE_t0_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..990a0f170bff0d3220562056281b342fbec51030 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_32_32--base40kRE_t0_s06.txt @@ -0,0 +1,129 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t0_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.502 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2235 +t=10: Selected seed 195 with value = 0.2235 +Query 1/1: Action query time = 4.907 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2251 +t=26: Selected seed 195 with value = 0.2251 +Query 1/1: Action query time = 5.530 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2586 +t=42: Selected seed 195 with value = 0.2586 +Query 1/1: Action query time = 5.262 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3049 +t=58: Selected seed 195 with value = 0.3049 +Query 1/1: Action query time = 4.745 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5786 +t=74: Selected seed 195 with value = 0.5786 +Query 1/1: Action query time = 5.061 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5350 +t=90: Selected seed 195 with value = 0.5350 +Query 1/1: Action query time = 4.846 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7775 +t=106: Selected seed 195 with value = 0.7775 +Query 1/1: Action query time = 3.730 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9449 +t=122: Selected seed 195 with value = 0.9449 +Saved rollout MP4 at path ./rollouts/base40kRE_t0_s06/2026_08_02-14_32_32--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.144 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3300 +t=10: Selected seed 195 with value = 0.3300 +Query 1/1: Action query time = 5.153 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3894 +t=26: Selected seed 195 with value = 0.3894 +Query 1/1: Action query time = 5.091 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4478 +t=42: Selected seed 195 with value = 0.4478 +Query 1/1: Action query time = 5.166 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5268 +t=58: Selected seed 195 with value = 0.5268 +Query 1/1: Action query time = 5.126 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6354 +t=74: Selected seed 195 with value = 0.6354 +Query 1/1: Action query time = 5.214 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7488 +t=90: Selected seed 195 with value = 0.7488 +Query 1/1: Action query time = 4.846 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8646 +t=106: Selected seed 195 with value = 0.8646 +Query 1/1: Action query time = 3.150 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/base40kRE_t0_s06/2026_08_02-14_32_32--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.164 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3243 +t=10: Selected seed 195 with value = 0.3243 +Query 1/1: Action query time = 3.912 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3791 +t=26: Selected seed 195 with value = 0.3791 +Query 1/1: Action query time = 4.978 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4507 +t=42: Selected seed 195 with value = 0.4507 +Query 1/1: Action query time = 5.038 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5243 +t=58: Selected seed 195 with value = 0.5243 +Query 1/1: Action query time = 4.917 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6200 +t=74: Selected seed 195 with value = 0.6200 +Query 1/1: Action query time = 5.074 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7276 +t=90: Selected seed 195 with value = 0.7276 +Query 1/1: Action query time = 5.177 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8480 +t=106: Selected seed 195 with value = 0.8480 +Query 1/1: Action query time = 3.718 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9874 +t=122: Selected seed 195 with value = 0.9874 +Saved rollout MP4 at path ./rollouts/base40kRE_t0_s06/2026_08_02-14_32_32--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_32_32--base40kRE_t0_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_32_32--base40kRE_t0_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..05c4c29cea2b164d8acad3eb3a23d3aacaf75035 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_32_32--base40kRE_t0_s11.txt @@ -0,0 +1,133 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t0_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 1.794 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2957 +t=10: Selected seed 195 with value = 0.2957 +Query 1/1: Action query time = 2.465 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3066 +t=26: Selected seed 195 with value = 0.3066 +Query 1/1: Action query time = 5.353 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4054 +t=42: Selected seed 195 with value = 0.4054 +Query 1/1: Action query time = 4.779 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4732 +t=58: Selected seed 195 with value = 0.4732 +Query 1/1: Action query time = 4.771 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5714 +t=74: Selected seed 195 with value = 0.5714 +Query 1/1: Action query time = 5.421 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6983 +t=90: Selected seed 195 with value = 0.6983 +Query 1/1: Action query time = 5.428 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7994 +t=106: Selected seed 195 with value = 0.7994 +Query 1/1: Action query time = 5.450 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9182 +t=122: Selected seed 195 with value = 0.9182 +Query 1/1: Action query time = 4.829 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/base40kRE_t0_s11/2026_08_02-14_32_32--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.996 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3288 +t=10: Selected seed 195 with value = 0.3288 +Query 1/1: Action query time = 1.823 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3528 +t=26: Selected seed 195 with value = 0.3528 +Query 1/1: Action query time = 5.221 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4142 +t=42: Selected seed 195 with value = 0.4142 +Query 1/1: Action query time = 4.981 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5016 +t=58: Selected seed 195 with value = 0.5016 +Query 1/1: Action query time = 5.086 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5914 +t=74: Selected seed 195 with value = 0.5914 +Query 1/1: Action query time = 5.183 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6828 +t=90: Selected seed 195 with value = 0.6828 +Query 1/1: Action query time = 5.199 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8047 +t=106: Selected seed 195 with value = 0.8047 +Query 1/1: Action query time = 4.651 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9500 +t=122: Selected seed 195 with value = 0.9500 +Saved rollout MP4 at path ./rollouts/base40kRE_t0_s11/2026_08_02-14_32_32--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.089 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2664 +t=10: Selected seed 195 with value = 0.2664 +Query 1/1: Action query time = 4.803 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2502 +t=26: Selected seed 195 with value = 0.2502 +Query 1/1: Action query time = 4.776 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4364 +t=42: Selected seed 195 with value = 0.4364 +Query 1/1: Action query time = 3.308 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5188 +t=58: Selected seed 195 with value = 0.5188 +Query 1/1: Action query time = 4.849 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6192 +t=74: Selected seed 195 with value = 0.6192 +Query 1/1: Action query time = 4.921 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7683 +t=90: Selected seed 195 with value = 0.7683 +Query 1/1: Action query time = 5.448 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8444 +t=106: Selected seed 195 with value = 0.8444 +Query 1/1: Action query time = 5.552 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9856 +t=122: Selected seed 195 with value = 0.9856 +Saved rollout MP4 at path ./rollouts/base40kRE_t0_s11/2026_08_02-14_32_32--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_32_33--base40kRE_t0_s10.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_32_33--base40kRE_t0_s10.txt new file mode 100644 index 0000000000000000000000000000000000000000..9bcaa0f02ba25bc57f22e801708ef656c10961bb --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_32_33--base40kRE_t0_s10.txt @@ -0,0 +1,137 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t0_s10', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='10,26,42', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.680 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3045 +t=10: Selected seed 195 with value = 0.3045 +Query 1/1: Action query time = 4.665 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3112 +t=26: Selected seed 195 with value = 0.3112 +Query 1/1: Action query time = 5.214 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3670 +t=42: Selected seed 195 with value = 0.3670 +Query 1/1: Action query time = 5.294 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4293 +t=58: Selected seed 195 with value = 0.4293 +Query 1/1: Action query time = 5.176 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4893 +t=74: Selected seed 195 with value = 0.4893 +Query 1/1: Action query time = 4.903 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5831 +t=90: Selected seed 195 with value = 0.5831 +Query 1/1: Action query time = 4.983 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6903 +t=106: Selected seed 195 with value = 0.6903 +Query 1/1: Action query time = 3.896 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8104 +t=122: Selected seed 195 with value = 0.8104 +Query 1/1: Action query time = 3.903 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9580 +t=138: Selected seed 195 with value = 0.9580 +Saved rollout MP4 at path ./rollouts/base40kRE_t0_s10/2026_08_02-14_32_33--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.143 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2783 +t=10: Selected seed 195 with value = 0.2783 +Query 1/1: Action query time = 5.359 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3344 +t=26: Selected seed 195 with value = 0.3344 +Query 1/1: Action query time = 5.269 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3843 +t=42: Selected seed 195 with value = 0.3843 +Query 1/1: Action query time = 5.038 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4633 +t=58: Selected seed 195 with value = 0.4633 +Query 1/1: Action query time = 5.013 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5462 +t=74: Selected seed 195 with value = 0.5462 +Query 1/1: Action query time = 5.216 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6420 +t=90: Selected seed 195 with value = 0.6420 +Query 1/1: Action query time = 3.813 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7492 +t=106: Selected seed 195 with value = 0.7492 +Query 1/1: Action query time = 4.434 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8813 +t=122: Selected seed 195 with value = 0.8813 +Query 1/1: Action query time = 4.583 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/base40kRE_t0_s10/2026_08_02-14_32_33--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.587 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3436 +t=10: Selected seed 195 with value = 0.3436 +Query 1/1: Action query time = 5.173 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3740 +t=26: Selected seed 195 with value = 0.3740 +Query 1/1: Action query time = 5.215 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4432 +t=42: Selected seed 195 with value = 0.4432 +Query 1/1: Action query time = 4.987 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5238 +t=58: Selected seed 195 with value = 0.5238 +Query 1/1: Action query time = 4.361 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5713 +t=74: Selected seed 195 with value = 0.5713 +Query 1/1: Action query time = 2.864 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6535 +t=90: Selected seed 195 with value = 0.6535 +Query 1/1: Action query time = 3.548 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7799 +t=106: Selected seed 195 with value = 0.7799 +Query 1/1: Action query time = 2.533 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9206 +t=122: Selected seed 195 with value = 0.9206 +Saved rollout MP4 at path ./rollouts/base40kRE_t0_s10/2026_08_02-14_32_33--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_35_55--base40kRE_t1_s05.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_35_55--base40kRE_t1_s05.txt new file mode 100644 index 0000000000000000000000000000000000000000..73610d32e03ce4f85d7ef797115cdc1a1bc2ff88 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_35_55--base40kRE_t1_s05.txt @@ -0,0 +1,109 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t1_s05', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,21,37', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 2.858 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4206 +t=10: Selected seed 195 with value = 0.4206 +Query 1/1: Action query time = 4.831 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4957 +t=26: Selected seed 195 with value = 0.4957 +Query 1/1: Action query time = 4.705 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=42: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 4.502 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6842 +t=58: Selected seed 195 with value = 0.6842 +Query 1/1: Action query time = 4.630 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8145 +t=74: Selected seed 195 with value = 0.8145 +Query 1/1: Action query time = 5.178 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9566 +t=90: Selected seed 195 with value = 0.9566 +Saved rollout MP4 at path ./rollouts/base40kRE_t1_s05/2026_08_02-14_35_55--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 4.338 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3707 +t=10: Selected seed 195 with value = 0.3707 +Query 1/1: Action query time = 4.009 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4217 +t=26: Selected seed 195 with value = 0.4217 +Query 1/1: Action query time = 6.041 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5101 +t=42: Selected seed 195 with value = 0.5101 +Query 1/1: Action query time = 4.508 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6013 +t=58: Selected seed 195 with value = 0.6013 +Query 1/1: Action query time = 4.716 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6978 +t=74: Selected seed 195 with value = 0.6978 +Query 1/1: Action query time = 4.969 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8830 +t=90: Selected seed 195 with value = 0.8830 +Query 1/1: Action query time = 4.294 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9975 +t=106: Selected seed 195 with value = 0.9975 +Saved rollout MP4 at path ./rollouts/base40kRE_t1_s05/2026_08_02-14_35_55--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 3.231 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4519 +t=10: Selected seed 195 with value = 0.4519 +Query 1/1: Action query time = 4.848 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5234 +t=26: Selected seed 195 with value = 0.5234 +Query 1/1: Action query time = 4.619 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5991 +t=42: Selected seed 195 with value = 0.5991 +Query 1/1: Action query time = 4.997 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7090 +t=58: Selected seed 195 with value = 0.7090 +Query 1/1: Action query time = 4.904 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8372 +t=74: Selected seed 195 with value = 0.8372 +Query 1/1: Action query time = 4.251 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9663 +t=90: Selected seed 195 with value = 0.9663 +Saved rollout MP4 at path ./rollouts/base40kRE_t1_s05/2026_08_02-14_35_55--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_35_55--base40kRE_t1_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_35_55--base40kRE_t1_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..a2812088ed22e666170603a98ab2a3c0924a91d6 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_35_55--base40kRE_t1_s06.txt @@ -0,0 +1,109 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t1_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 4.719 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3566 +t=10: Selected seed 195 with value = 0.3566 +Query 1/1: Action query time = 4.383 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4237 +t=26: Selected seed 195 with value = 0.4237 +Query 1/1: Action query time = 4.954 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5047 +t=42: Selected seed 195 with value = 0.5047 +Query 1/1: Action query time = 5.101 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6456 +t=58: Selected seed 195 with value = 0.6456 +Query 1/1: Action query time = 5.247 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7660 +t=74: Selected seed 195 with value = 0.7660 +Query 1/1: Action query time = 3.835 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8842 +t=90: Selected seed 195 with value = 0.8842 +Query 1/1: Action query time = 4.487 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9916 +t=106: Selected seed 195 with value = 0.9916 +Saved rollout MP4 at path ./rollouts/base40kRE_t1_s06/2026_08_02-14_35_55--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 6.312 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3967 +t=10: Selected seed 195 with value = 0.3967 +Query 1/1: Action query time = 5.301 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5057 +t=26: Selected seed 195 with value = 0.5057 +Query 1/1: Action query time = 4.818 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6097 +t=42: Selected seed 195 with value = 0.6097 +Query 1/1: Action query time = 5.164 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6953 +t=58: Selected seed 195 with value = 0.6953 +Query 1/1: Action query time = 4.143 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7955 +t=74: Selected seed 195 with value = 0.7955 +Query 1/1: Action query time = 3.413 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9458 +t=90: Selected seed 195 with value = 0.9458 +Saved rollout MP4 at path ./rollouts/base40kRE_t1_s06/2026_08_02-14_35_55--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.288 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4170 +t=10: Selected seed 195 with value = 0.4170 +Query 1/1: Action query time = 5.141 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4756 +t=26: Selected seed 195 with value = 0.4756 +Query 1/1: Action query time = 3.935 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5911 +t=42: Selected seed 195 with value = 0.5911 +Query 1/1: Action query time = 4.213 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6858 +t=58: Selected seed 195 with value = 0.6858 +Query 1/1: Action query time = 3.972 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8137 +t=74: Selected seed 195 with value = 0.8137 +Query 1/1: Action query time = 2.528 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9478 +t=90: Selected seed 195 with value = 0.9478 +Saved rollout MP4 at path ./rollouts/base40kRE_t1_s06/2026_08_02-14_35_55--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_38_39--base40kRE_t2_s06.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_38_39--base40kRE_t2_s06.txt new file mode 100644 index 0000000000000000000000000000000000000000..056751af47c15b3c4a0426ffe2a738b6f60b087a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_38_39--base40kRE_t2_s06.txt @@ -0,0 +1,109 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t2_s06', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='6,22,38', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.164 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3094 +t=10: Selected seed 195 with value = 0.3094 +Query 1/1: Action query time = 2.336 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3780 +t=26: Selected seed 195 with value = 0.3780 +Query 1/1: Action query time = 4.861 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4663 +t=42: Selected seed 195 with value = 0.4663 +Query 1/1: Action query time = 4.739 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5871 +t=58: Selected seed 195 with value = 0.5871 +Query 1/1: Action query time = 5.250 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6923 +t=74: Selected seed 195 with value = 0.6923 +Query 1/1: Action query time = 4.370 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8408 +t=90: Selected seed 195 with value = 0.8408 +Query 1/1: Action query time = 4.827 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9734 +t=106: Selected seed 195 with value = 0.9734 +Saved rollout MP4 at path ./rollouts/base40kRE_t2_s06/2026_08_02-14_38_39--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.978 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4340 +t=10: Selected seed 195 with value = 0.4340 +Query 1/1: Action query time = 4.539 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4939 +t=26: Selected seed 195 with value = 0.4939 +Query 1/1: Action query time = 4.585 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5601 +t=42: Selected seed 195 with value = 0.5601 +Query 1/1: Action query time = 4.851 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6848 +t=58: Selected seed 195 with value = 0.6848 +Query 1/1: Action query time = 4.303 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8018 +t=74: Selected seed 195 with value = 0.8018 +Query 1/1: Action query time = 3.865 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9451 +t=90: Selected seed 195 with value = 0.9451 +Saved rollout MP4 at path ./rollouts/base40kRE_t2_s06/2026_08_02-14_38_39--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.162 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4345 +t=10: Selected seed 195 with value = 0.4345 +Query 1/1: Action query time = 4.936 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4826 +t=26: Selected seed 195 with value = 0.4826 +Query 1/1: Action query time = 5.315 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6173 +t=42: Selected seed 195 with value = 0.6173 +Query 1/1: Action query time = 4.853 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7263 +t=58: Selected seed 195 with value = 0.7263 +Query 1/1: Action query time = 3.410 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8378 +t=74: Selected seed 195 with value = 0.8378 +Query 1/1: Action query time = 4.842 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9742 +t=90: Selected seed 195 with value = 0.9742 +Saved rollout MP4 at path ./rollouts/base40kRE_t2_s06/2026_08_02-14_38_39--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_38_40--base40kRE_t2_s12.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_38_40--base40kRE_t2_s12.txt new file mode 100644 index 0000000000000000000000000000000000000000..9830015ddd448bcac1ed2f09081d59597302afc0 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_38_40--base40kRE_t2_s12.txt @@ -0,0 +1,121 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t2_s12', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='12,28,44', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.422 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4656 +t=10: Selected seed 195 with value = 0.4656 +Query 1/1: Action query time = 5.148 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5461 +t=26: Selected seed 195 with value = 0.5461 +Query 1/1: Action query time = 5.191 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6258 +t=42: Selected seed 195 with value = 0.6258 +Query 1/1: Action query time = 4.117 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7313 +t=58: Selected seed 195 with value = 0.7313 +Query 1/1: Action query time = 4.360 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8638 +t=74: Selected seed 195 with value = 0.8638 +Query 1/1: Action query time = 1.782 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/base40kRE_t2_s12/2026_08_02-14_38_40--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.934 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3309 +t=10: Selected seed 195 with value = 0.3309 +Query 1/1: Action query time = 4.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3394 +t=26: Selected seed 195 with value = 0.3394 +Query 1/1: Action query time = 5.289 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4064 +t=42: Selected seed 195 with value = 0.4064 +Query 1/1: Action query time = 5.270 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4865 +t=58: Selected seed 195 with value = 0.4865 +Query 1/1: Action query time = 4.272 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5236 +t=74: Selected seed 195 with value = 0.5236 +Query 1/1: Action query time = 3.159 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5862 +t=90: Selected seed 195 with value = 0.5862 +Query 1/1: Action query time = 3.583 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7055 +t=106: Selected seed 195 with value = 0.7055 +Query 1/1: Action query time = 5.858 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8435 +t=122: Selected seed 195 with value = 0.8435 +Query 1/1: Action query time = 5.250 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=138: Selected seed 195 with value = 0.9912 +Saved rollout MP4 at path ./rollouts/base40kRE_t2_s12/2026_08_02-14_38_40--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.247 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3133 +t=10: Selected seed 195 with value = 0.3133 +Query 1/1: Action query time = 1.976 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3640 +t=26: Selected seed 195 with value = 0.3640 +Query 1/1: Action query time = 3.242 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4682 +t=42: Selected seed 195 with value = 0.4682 +Query 1/1: Action query time = 1.641 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5579 +t=58: Selected seed 195 with value = 0.5579 +Query 1/1: Action query time = 1.613 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6344 +t=74: Selected seed 195 with value = 0.6344 +Query 1/1: Action query time = 1.459 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7575 +t=90: Selected seed 195 with value = 0.7575 +Query 1/1: Action query time = 1.871 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8976 +t=106: Selected seed 195 with value = 0.8976 +Saved rollout MP4 at path ./rollouts/base40kRE_t2_s12/2026_08_02-14_38_40--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_41_33--base40kRE_t3_s01.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_41_33--base40kRE_t3_s01.txt new file mode 100644 index 0000000000000000000000000000000000000000..5e6e3bcf2f1cd7e2829739d06b5a6f604117391e --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_41_33--base40kRE_t3_s01.txt @@ -0,0 +1,220 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t3_s01', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,17,33,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 7.261 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1819 +t=10: Selected seed 195 with value = 0.1819 +Query 1/1: Action query time = 5.671 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2239 +t=26: Selected seed 195 with value = 0.2239 +Query 1/1: Action query time = 5.107 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2412 +t=42: Selected seed 195 with value = 0.2412 +Query 1/1: Action query time = 4.986 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2817 +t=58: Selected seed 195 with value = 0.2817 +Query 1/1: Action query time = 5.080 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3718 +t=74: Selected seed 195 with value = 0.3718 +Query 1/1: Action query time = 4.783 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4426 +t=90: Selected seed 195 with value = 0.4426 +Query 1/1: Action query time = 4.623 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5224 +t=106: Selected seed 195 with value = 0.5224 +Query 1/1: Action query time = 5.107 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5892 +t=122: Selected seed 195 with value = 0.5892 +Query 1/1: Action query time = 5.333 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7088 +t=138: Selected seed 195 with value = 0.7088 +Query 1/1: Action query time = 4.795 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8715 +t=154: Selected seed 195 with value = 0.8715 +Query 1/1: Action query time = 2.744 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/base40kRE_t3_s01/2026_08_02-14_41_33--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 3.595 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1807 +t=10: Selected seed 195 with value = 0.1807 +Query 1/1: Action query time = 4.716 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2061 +t=26: Selected seed 195 with value = 0.2061 +Query 1/1: Action query time = 4.783 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2130 +t=42: Selected seed 195 with value = 0.2130 +Query 1/1: Action query time = 5.092 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3023 +t=58: Selected seed 195 with value = 0.3023 +Query 1/1: Action query time = 4.428 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3447 +t=74: Selected seed 195 with value = 0.3447 +Query 1/1: Action query time = 4.484 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4150 +t=90: Selected seed 195 with value = 0.4150 +Query 1/1: Action query time = 3.959 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4899 +t=106: Selected seed 195 with value = 0.4899 +Query 1/1: Action query time = 2.168 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5821 +t=122: Selected seed 195 with value = 0.5821 +Query 1/1: Action query time = 4.617 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6633 +t=138: Selected seed 195 with value = 0.6633 +Query 1/1: Action query time = 5.262 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7781 +t=154: Selected seed 195 with value = 0.7781 +Query 1/1: Action query time = 5.156 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9582 +t=170: Selected seed 195 with value = 0.9582 +Saved rollout MP4 at path ./rollouts/base40kRE_t3_s01/2026_08_02-14_41_33--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.580 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1585 +t=10: Selected seed 195 with value = 0.1585 +Query 1/1: Action query time = 3.804 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1609 +t=26: Selected seed 195 with value = 0.1609 +Query 1/1: Action query time = 4.862 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2330 +t=42: Selected seed 195 with value = 0.2330 +Query 1/1: Action query time = 5.229 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2269 +t=58: Selected seed 195 with value = 0.2269 +Query 1/1: Action query time = 5.136 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3225 +t=74: Selected seed 195 with value = 0.3225 +Query 1/1: Action query time = 4.986 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3967 +t=90: Selected seed 195 with value = 0.3967 +Query 1/1: Action query time = 4.532 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4580 +t=106: Selected seed 195 with value = 0.4580 +Query 1/1: Action query time = 4.348 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5526 +t=122: Selected seed 195 with value = 0.5526 +Query 1/1: Action query time = 3.189 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4896 +t=138: Selected seed 195 with value = 0.4896 +Query 1/1: Action query time = 4.917 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7105 +t=154: Selected seed 195 with value = 0.7105 +Query 1/1: Action query time = 5.405 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8420 +t=170: Selected seed 195 with value = 0.8420 +Query 1/1: Action query time = 4.525 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9884 +t=186: Selected seed 195 with value = 0.9884 +Saved rollout MP4 at path ./rollouts/base40kRE_t3_s01/2026_08_02-14_41_33--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 4.214 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1942 +t=10: Selected seed 195 with value = 0.1942 +Query 1/1: Action query time = 3.350 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2273 +t=26: Selected seed 195 with value = 0.2273 +Query 1/1: Action query time = 1.980 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2830 +t=42: Selected seed 195 with value = 0.2830 +Query 1/1: Action query time = 2.089 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3192 +t=58: Selected seed 195 with value = 0.3192 +Query 1/1: Action query time = 2.333 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3904 +t=74: Selected seed 195 with value = 0.3904 +Query 1/1: Action query time = 2.634 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4550 +t=90: Selected seed 195 with value = 0.4550 +Query 1/1: Action query time = 2.997 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5171 +t=106: Selected seed 195 with value = 0.5171 +Query 1/1: Action query time = 2.268 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6111 +t=122: Selected seed 195 with value = 0.6111 +Query 1/1: Action query time = 2.136 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6882 +t=138: Selected seed 195 with value = 0.6882 +Query 1/1: Action query time = 2.004 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8097 +t=154: Selected seed 195 with value = 0.8097 +Query 1/1: Action query time = 1.568 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9365 +t=170: Selected seed 195 with value = 0.9365 +Saved rollout MP4 at path ./rollouts/base40kRE_t3_s01/2026_08_02-14_41_33--with_future_img--episode=4--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 4 +Total successes: 4 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_41_33--base40kRE_t3_s04.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_41_33--base40kRE_t3_s04.txt new file mode 100644 index 0000000000000000000000000000000000000000..c054c18fbf9774868b30081b8fdf7988f7e4625e --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_41_33--base40kRE_t3_s04.txt @@ -0,0 +1,185 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t3_s04', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='4,20,36', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 3.914 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1881 +t=10: Selected seed 195 with value = 0.1881 +Query 1/1: Action query time = 5.493 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2140 +t=26: Selected seed 195 with value = 0.2140 +Query 1/1: Action query time = 4.939 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2525 +t=42: Selected seed 195 with value = 0.2525 +Query 1/1: Action query time = 4.558 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2882 +t=58: Selected seed 195 with value = 0.2882 +Query 1/1: Action query time = 4.682 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3442 +t=74: Selected seed 195 with value = 0.3442 +Query 1/1: Action query time = 5.122 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4093 +t=90: Selected seed 195 with value = 0.4093 +Query 1/1: Action query time = 4.813 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5094 +t=106: Selected seed 195 with value = 0.5094 +Query 1/1: Action query time = 4.814 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4676 +t=122: Selected seed 195 with value = 0.4676 +Query 1/1: Action query time = 5.009 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6248 +t=138: Selected seed 195 with value = 0.6248 +Query 1/1: Action query time = 5.166 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8059 +t=154: Selected seed 195 with value = 0.8059 +Query 1/1: Action query time = 5.212 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3932 +t=170: Selected seed 195 with value = 0.3932 +Saved rollout MP4 at path ./rollouts/base40kRE_t3_s04/2026_08_02-14_41_33--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.063 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1956 +t=10: Selected seed 195 with value = 0.1956 +Query 1/1: Action query time = 3.831 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2192 +t=26: Selected seed 195 with value = 0.2192 +Query 1/1: Action query time = 4.613 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2265 +t=42: Selected seed 195 with value = 0.2265 +Query 1/1: Action query time = 4.592 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2212 +t=58: Selected seed 195 with value = 0.2212 +Query 1/1: Action query time = 5.352 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2521 +t=74: Selected seed 195 with value = 0.2521 +Query 1/1: Action query time = 4.821 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2944 +t=90: Selected seed 195 with value = 0.2944 +Query 1/1: Action query time = 4.035 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3522 +t=106: Selected seed 195 with value = 0.3522 +Query 1/1: Action query time = 3.829 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4130 +t=122: Selected seed 195 with value = 0.4130 +Query 1/1: Action query time = 3.825 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4839 +t=138: Selected seed 195 with value = 0.4839 +Query 1/1: Action query time = 4.386 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4002 +t=154: Selected seed 195 with value = 0.4002 +Query 1/1: Action query time = 4.594 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6419 +t=170: Selected seed 195 with value = 0.6419 +Query 1/1: Action query time = 5.141 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7395 +t=186: Selected seed 195 with value = 0.7395 +Query 1/1: Action query time = 3.862 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9020 +t=202: Selected seed 195 with value = 0.9020 +Query 1/1: Action query time = 4.809 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/base40kRE_t3_s04/2026_08_02-14_41_33--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.762 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1440 +t=10: Selected seed 195 with value = 0.1440 +Query 1/1: Action query time = 5.167 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1460 +t=26: Selected seed 195 with value = 0.1460 +Query 1/1: Action query time = 5.330 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1837 +t=42: Selected seed 195 with value = 0.1837 +Query 1/1: Action query time = 4.943 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2513 +t=58: Selected seed 195 with value = 0.2513 +Query 1/1: Action query time = 4.388 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2822 +t=74: Selected seed 195 with value = 0.2822 +Query 1/1: Action query time = 4.145 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2884 +t=90: Selected seed 195 with value = 0.2884 +Query 1/1: Action query time = 3.452 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3535 +t=106: Selected seed 195 with value = 0.3535 +Query 1/1: Action query time = 5.031 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4226 +t=122: Selected seed 195 with value = 0.4226 +Query 1/1: Action query time = 5.672 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4832 +t=138: Selected seed 195 with value = 0.4832 +Query 1/1: Action query time = 4.685 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6107 +t=154: Selected seed 195 with value = 0.6107 +Query 1/1: Action query time = 3.711 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7226 +t=170: Selected seed 195 with value = 0.7226 +Query 1/1: Action query time = 3.118 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8415 +t=186: Selected seed 195 with value = 0.8415 +Query 1/1: Action query time = 3.310 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/base40kRE_t3_s04/2026_08_02-14_41_33--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_42--base40kRE_t4_s00.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_42--base40kRE_t4_s00.txt new file mode 100644 index 0000000000000000000000000000000000000000..bf078541f16738cf00f1d012bf09a591377e721e --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_42--base40kRE_t4_s00.txt @@ -0,0 +1,128 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t4_s00', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,16,32,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 1.734 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4186 +t=10: Selected seed 195 with value = 0.4186 +Query 1/1: Action query time = 2.306 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5032 +t=26: Selected seed 195 with value = 0.5032 +Query 1/1: Action query time = 3.991 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5914 +t=42: Selected seed 195 with value = 0.5914 +Query 1/1: Action query time = 4.737 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6890 +t=58: Selected seed 195 with value = 0.6890 +Query 1/1: Action query time = 6.368 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8138 +t=74: Selected seed 195 with value = 0.8138 +Query 1/1: Action query time = 4.757 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9455 +t=90: Selected seed 195 with value = 0.9455 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s00/2026_08_02-14_46_42--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.216 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4702 +t=10: Selected seed 195 with value = 0.4702 +Query 1/1: Action query time = 4.291 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5569 +t=26: Selected seed 195 with value = 0.5569 +Query 1/1: Action query time = 2.977 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6643 +t=42: Selected seed 195 with value = 0.6643 +Query 1/1: Action query time = 5.330 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7558 +t=58: Selected seed 195 with value = 0.7558 +Query 1/1: Action query time = 4.815 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8708 +t=74: Selected seed 195 with value = 0.8708 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s00/2026_08_02-14_46_42--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.377 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4706 +t=10: Selected seed 195 with value = 0.4706 +Query 1/1: Action query time = 3.915 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5507 +t=26: Selected seed 195 with value = 0.5507 +Query 1/1: Action query time = 4.787 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6529 +t=42: Selected seed 195 with value = 0.6529 +Query 1/1: Action query time = 4.146 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7416 +t=58: Selected seed 195 with value = 0.7416 +Query 1/1: Action query time = 3.864 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8673 +t=74: Selected seed 195 with value = 0.8673 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s00/2026_08_02-14_46_42--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 5.797 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4333 +t=10: Selected seed 195 with value = 0.4333 +Query 1/1: Action query time = 4.804 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5396 +t=26: Selected seed 195 with value = 0.5396 +Query 1/1: Action query time = 3.568 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5967 +t=42: Selected seed 195 with value = 0.5967 +Query 1/1: Action query time = 2.651 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7017 +t=58: Selected seed 195 with value = 0.7017 +Query 1/1: Action query time = 2.165 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8504 +t=74: Selected seed 195 with value = 0.8504 +Query 1/1: Action query time = 1.708 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9683 +t=90: Selected seed 195 with value = 0.9683 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s00/2026_08_02-14_46_42--with_future_img--episode=4--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 4 +Total successes: 4 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_43--base40kRE_t4_s08.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_43--base40kRE_t4_s08.txt new file mode 100644 index 0000000000000000000000000000000000000000..3bdabd0465d58104059531b97bfb786aa8ad3ae1 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_43--base40kRE_t4_s08.txt @@ -0,0 +1,101 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t4_s08', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='8,24,40', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.700 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4367 +t=10: Selected seed 195 with value = 0.4367 +Query 1/1: Action query time = 4.712 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5303 +t=26: Selected seed 195 with value = 0.5303 +Query 1/1: Action query time = 5.886 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5557 +t=42: Selected seed 195 with value = 0.5557 +Query 1/1: Action query time = 3.732 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6569 +t=58: Selected seed 195 with value = 0.6569 +Query 1/1: Action query time = 5.173 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8602 +t=74: Selected seed 195 with value = 0.8602 +Query 1/1: Action query time = 4.822 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s08/2026_08_02-14_46_43--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.030 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4260 +t=10: Selected seed 195 with value = 0.4260 +Query 1/1: Action query time = 5.086 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5321 +t=26: Selected seed 195 with value = 0.5321 +Query 1/1: Action query time = 4.661 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6044 +t=42: Selected seed 195 with value = 0.6044 +Query 1/1: Action query time = 4.878 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7590 +t=58: Selected seed 195 with value = 0.7590 +Query 1/1: Action query time = 5.645 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8946 +t=74: Selected seed 195 with value = 0.8946 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s08/2026_08_02-14_46_43--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.860 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4403 +t=10: Selected seed 195 with value = 0.4403 +Query 1/1: Action query time = 3.318 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5427 +t=26: Selected seed 195 with value = 0.5427 +Query 1/1: Action query time = 4.499 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6359 +t=42: Selected seed 195 with value = 0.6359 +Query 1/1: Action query time = 4.342 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7393 +t=58: Selected seed 195 with value = 0.7393 +Query 1/1: Action query time = 4.831 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8864 +t=74: Selected seed 195 with value = 0.8864 +Query 1/1: Action query time = 4.208 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9868 +t=90: Selected seed 195 with value = 0.9868 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s08/2026_08_02-14_46_43--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_43--base40kRE_t4_s10.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_43--base40kRE_t4_s10.txt new file mode 100644 index 0000000000000000000000000000000000000000..8273410f931df2e222a34a77ec19af018e72d892 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_43--base40kRE_t4_s10.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t4_s10', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='10,26,42', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.873 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4369 +t=10: Selected seed 195 with value = 0.4369 +Query 1/1: Action query time = 5.524 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5311 +t=26: Selected seed 195 with value = 0.5311 +Query 1/1: Action query time = 5.328 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6113 +t=42: Selected seed 195 with value = 0.6113 +Query 1/1: Action query time = 4.784 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7174 +t=58: Selected seed 195 with value = 0.7174 +Query 1/1: Action query time = 5.536 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8339 +t=74: Selected seed 195 with value = 0.8339 +Query 1/1: Action query time = 4.320 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9876 +t=90: Selected seed 195 with value = 0.9876 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s10/2026_08_02-14_46_43--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 1.820 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4243 +t=10: Selected seed 195 with value = 0.4243 +Query 1/1: Action query time = 4.617 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5143 +t=26: Selected seed 195 with value = 0.5143 +Query 1/1: Action query time = 4.752 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5980 +t=42: Selected seed 195 with value = 0.5980 +Query 1/1: Action query time = 4.207 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6875 +t=58: Selected seed 195 with value = 0.6875 +Query 1/1: Action query time = 4.199 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8231 +t=74: Selected seed 195 with value = 0.8231 +Query 1/1: Action query time = 5.065 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9807 +t=90: Selected seed 195 with value = 0.9807 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s10/2026_08_02-14_46_43--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.755 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4287 +t=10: Selected seed 195 with value = 0.4287 +Query 1/1: Action query time = 3.357 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5098 +t=26: Selected seed 195 with value = 0.5098 +Query 1/1: Action query time = 4.904 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5865 +t=42: Selected seed 195 with value = 0.5865 +Query 1/1: Action query time = 4.617 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6924 +t=58: Selected seed 195 with value = 0.6924 +Query 1/1: Action query time = 5.071 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8182 +t=74: Selected seed 195 with value = 0.8182 +Query 1/1: Action query time = 4.230 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9583 +t=90: Selected seed 195 with value = 0.9583 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s10/2026_08_02-14_46_43--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_44--base40kRE_t4_s11.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_44--base40kRE_t4_s11.txt new file mode 100644 index 0000000000000000000000000000000000000000..bd53e1f28d8eecadab6f30a40ac3560f10cfbb62 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_46_44--base40kRE_t4_s11.txt @@ -0,0 +1,105 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t4_s11', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='11,27,43', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.106 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4450 +t=10: Selected seed 195 with value = 0.4450 +Query 1/1: Action query time = 5.512 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5429 +t=26: Selected seed 195 with value = 0.5429 +Query 1/1: Action query time = 5.318 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6072 +t=42: Selected seed 195 with value = 0.6072 +Query 1/1: Action query time = 4.292 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7010 +t=58: Selected seed 195 with value = 0.7010 +Query 1/1: Action query time = 5.453 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8382 +t=74: Selected seed 195 with value = 0.8382 +Query 1/1: Action query time = 4.868 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9829 +t=90: Selected seed 195 with value = 0.9829 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s11/2026_08_02-14_46_44--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 2.106 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4631 +t=10: Selected seed 195 with value = 0.4631 +Query 1/1: Action query time = 5.148 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5257 +t=26: Selected seed 195 with value = 0.5257 +Query 1/1: Action query time = 4.993 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6178 +t=42: Selected seed 195 with value = 0.6178 +Query 1/1: Action query time = 5.476 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7207 +t=58: Selected seed 195 with value = 0.7207 +Query 1/1: Action query time = 5.315 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8419 +t=74: Selected seed 195 with value = 0.8419 +Query 1/1: Action query time = 2.833 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9949 +t=90: Selected seed 195 with value = 0.9949 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s11/2026_08_02-14_46_44--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.023 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4616 +t=10: Selected seed 195 with value = 0.4616 +Query 1/1: Action query time = 3.809 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5327 +t=26: Selected seed 195 with value = 0.5327 +Query 1/1: Action query time = 4.911 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6260 +t=42: Selected seed 195 with value = 0.6260 +Query 1/1: Action query time = 5.236 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7336 +t=58: Selected seed 195 with value = 0.7336 +Query 1/1: Action query time = 4.439 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8857 +t=74: Selected seed 195 with value = 0.8857 +Query 1/1: Action query time = 3.503 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/base40kRE_t4_s11/2026_08_02-14_46_44--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_49_43--base40kRE_t5_s05.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_49_43--base40kRE_t5_s05.txt new file mode 100644 index 0000000000000000000000000000000000000000..40d6219e2448881e9d5f60a1431e2bf4e83bc065 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_49_43--base40kRE_t5_s05.txt @@ -0,0 +1,141 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t5_s05', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,21,37', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 2.864 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2802 +t=10: Selected seed 195 with value = 0.2802 +Query 1/1: Action query time = 4.462 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3137 +t=26: Selected seed 195 with value = 0.3137 +Query 1/1: Action query time = 4.536 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3828 +t=42: Selected seed 195 with value = 0.3828 +Query 1/1: Action query time = 5.133 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4701 +t=58: Selected seed 195 with value = 0.4701 +Query 1/1: Action query time = 4.980 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2714 +t=74: Selected seed 195 with value = 0.2714 +Query 1/1: Action query time = 4.785 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3559 +t=90: Selected seed 195 with value = 0.3559 +Query 1/1: Action query time = 3.762 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4603 +t=106: Selected seed 195 with value = 0.4603 +Query 1/1: Action query time = 4.937 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8186 +t=122: Selected seed 195 with value = 0.8186 +Query 1/1: Action query time = 4.724 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9773 +t=138: Selected seed 195 with value = 0.9773 +Saved rollout MP4 at path ./rollouts/base40kRE_t5_s05/2026_08_02-14_49_43--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.421 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2598 +t=10: Selected seed 195 with value = 0.2598 +Query 1/1: Action query time = 4.970 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2730 +t=26: Selected seed 195 with value = 0.2730 +Query 1/1: Action query time = 4.100 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2828 +t=42: Selected seed 195 with value = 0.2828 +Query 1/1: Action query time = 5.402 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4594 +t=58: Selected seed 195 with value = 0.4594 +Query 1/1: Action query time = 4.845 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3938 +t=74: Selected seed 195 with value = 0.3938 +Query 1/1: Action query time = 4.668 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6144 +t=90: Selected seed 195 with value = 0.6144 +Query 1/1: Action query time = 4.839 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7602 +t=106: Selected seed 195 with value = 0.7602 +Query 1/1: Action query time = 4.164 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8811 +t=122: Selected seed 195 with value = 0.8811 +Saved rollout MP4 at path ./rollouts/base40kRE_t5_s05/2026_08_02-14_49_43--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 3.542 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2399 +t=10: Selected seed 195 with value = 0.2399 +Query 1/1: Action query time = 5.331 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1856 +t=26: Selected seed 195 with value = 0.1856 +Query 1/1: Action query time = 4.674 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2166 +t=42: Selected seed 195 with value = 0.2166 +Query 1/1: Action query time = 2.760 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4166 +t=58: Selected seed 195 with value = 0.4166 +Query 1/1: Action query time = 4.925 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3363 +t=74: Selected seed 195 with value = 0.3363 +Query 1/1: Action query time = 4.543 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3489 +t=90: Selected seed 195 with value = 0.3489 +Query 1/1: Action query time = 4.674 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4237 +t=106: Selected seed 195 with value = 0.4237 +Query 1/1: Action query time = 4.195 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7284 +t=122: Selected seed 195 with value = 0.7284 +Query 1/1: Action query time = 4.940 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5707 +t=138: Selected seed 195 with value = 0.5707 +Query 1/1: Action query time = 5.168 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=154: Selected seed 195 with value = 0.9871 +Saved rollout MP4 at path ./rollouts/base40kRE_t5_s05/2026_08_02-14_49_43--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_49_44--base40kRE_t5_s12.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_49_44--base40kRE_t5_s12.txt new file mode 100644 index 0000000000000000000000000000000000000000..118a7426f2086df3bdb8d6b50e3705cc635e48e9 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_02-14_49_44--base40kRE_t5_s12.txt @@ -0,0 +1,137 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/checkpoints/imaginaire4-output/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_base_stage_3gpu_40k_save4k_20260726/checkpoints/iter_000040000/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='base40kRE_t5_s12', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='12,28,44', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 5.606 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3005 +t=10: Selected seed 195 with value = 0.3005 +Query 1/1: Action query time = 5.596 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2638 +t=26: Selected seed 195 with value = 0.2638 +Query 1/1: Action query time = 5.262 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3291 +t=42: Selected seed 195 with value = 0.3291 +Query 1/1: Action query time = 4.958 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3811 +t=58: Selected seed 195 with value = 0.3811 +Query 1/1: Action query time = 4.377 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5166 +t=74: Selected seed 195 with value = 0.5166 +Query 1/1: Action query time = 5.032 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6457 +t=90: Selected seed 195 with value = 0.6457 +Query 1/1: Action query time = 4.169 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7324 +t=106: Selected seed 195 with value = 0.7324 +Query 1/1: Action query time = 4.026 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8699 +t=122: Selected seed 195 with value = 0.8699 +Query 1/1: Action query time = 2.547 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9796 +t=138: Selected seed 195 with value = 0.9796 +Saved rollout MP4 at path ./rollouts/base40kRE_t5_s12/2026_08_02-14_49_44--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 5.029 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2422 +t=10: Selected seed 195 with value = 0.2422 +Query 1/1: Action query time = 5.404 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2826 +t=26: Selected seed 195 with value = 0.2826 +Query 1/1: Action query time = 3.805 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3421 +t=42: Selected seed 195 with value = 0.3421 +Query 1/1: Action query time = 4.478 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4511 +t=58: Selected seed 195 with value = 0.4511 +Query 1/1: Action query time = 4.127 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4565 +t=74: Selected seed 195 with value = 0.4565 +Query 1/1: Action query time = 5.023 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5816 +t=90: Selected seed 195 with value = 0.5816 +Query 1/1: Action query time = 5.125 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6561 +t=106: Selected seed 195 with value = 0.6561 +Query 1/1: Action query time = 4.108 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7641 +t=122: Selected seed 195 with value = 0.7641 +Query 1/1: Action query time = 2.959 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9453 +t=138: Selected seed 195 with value = 0.9453 +Saved rollout MP4 at path ./rollouts/base40kRE_t5_s12/2026_08_02-14_49_44--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 4.459 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2886 +t=10: Selected seed 195 with value = 0.2886 +Query 1/1: Action query time = 5.113 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2119 +t=26: Selected seed 195 with value = 0.2119 +Query 1/1: Action query time = 4.857 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3736 +t=42: Selected seed 195 with value = 0.3736 +Query 1/1: Action query time = 4.096 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4469 +t=58: Selected seed 195 with value = 0.4469 +Query 1/1: Action query time = 5.171 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5711 +t=74: Selected seed 195 with value = 0.5711 +Query 1/1: Action query time = 4.887 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6781 +t=90: Selected seed 195 with value = 0.6781 +Query 1/1: Action query time = 3.466 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7887 +t=106: Selected seed 195 with value = 0.7887 +Query 1/1: Action query time = 1.694 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9290 +t=122: Selected seed 195 with value = 0.9290 +Saved rollout MP4 at path ./rollouts/base40kRE_t5_s12/2026_08_02-14_49_44--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 3 +Total successes: 3 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-00_48_01--t7demochan800_t5.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-00_48_01--t7demochan800_t5.txt new file mode 100644 index 0000000000000000000000000000000000000000..ff7d35fc1a357e49f8e417a823c38b6d7bc35cc8 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-00_48_01--t7demochan800_t5.txt @@ -0,0 +1,682 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_demochan800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=10, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7demochan800_t5', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 4.121 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4325 +t=10: Selected seed 195 with value = 0.4325 +Query 1/1: Action query time = 2.974 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5208 +t=26: Selected seed 195 with value = 0.5208 +Query 1/1: Action query time = 3.839 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6092 +t=42: Selected seed 195 with value = 0.6092 +Query 1/1: Action query time = 3.677 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6708 +t=58: Selected seed 195 with value = 0.6708 +Query 1/1: Action query time = 3.621 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7325 +t=74: Selected seed 195 with value = 0.7325 +Query 1/1: Action query time = 2.378 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8582 +t=90: Selected seed 195 with value = 0.8582 +Query 1/1: Action query time = 3.462 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9310 +t=106: Selected seed 195 with value = 0.9310 +Query 1/1: Action query time = 2.847 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9817 +t=122: Selected seed 195 with value = 0.9817 +Query 1/1: Action query time = 2.935 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=138: Selected seed 195 with value = 0.9989 +Query 1/1: Action query time = 2.675 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.457 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7demochan800_t5/2026_08_03-00_48_01--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 2.786 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4534 +t=10: Selected seed 195 with value = 0.4534 +Query 1/1: Action query time = 3.059 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5322 +t=26: Selected seed 195 with value = 0.5322 +Query 1/1: Action query time = 2.931 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6315 +t=42: Selected seed 195 with value = 0.6315 +Query 1/1: Action query time = 3.213 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8168 +t=58: Selected seed 195 with value = 0.8168 +Query 1/1: Action query time = 2.407 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9253 +t=74: Selected seed 195 with value = 0.9253 +Query 1/1: Action query time = 2.216 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9741 +t=90: Selected seed 195 with value = 0.9741 +Query 1/1: Action query time = 1.093 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.898 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.831 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.587 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.666 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.189 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.758 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7demochan800_t5/2026_08_03-00_48_01--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 2.301 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4543 +t=10: Selected seed 195 with value = 0.4543 +Query 1/1: Action query time = 2.381 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5020 +t=26: Selected seed 195 with value = 0.5020 +Query 1/1: Action query time = 2.180 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5817 +t=42: Selected seed 195 with value = 0.5817 +Query 1/1: Action query time = 2.128 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6954 +t=58: Selected seed 195 with value = 0.6954 +Query 1/1: Action query time = 2.931 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8675 +t=74: Selected seed 195 with value = 0.8675 +Query 1/1: Action query time = 2.903 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9485 +t=90: Selected seed 195 with value = 0.9485 +Query 1/1: Action query time = 2.983 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9801 +t=106: Selected seed 195 with value = 0.9801 +Query 1/1: Action query time = 1.949 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9949 +t=122: Selected seed 195 with value = 0.9949 +Query 1/1: Action query time = 3.237 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.767 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.812 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.824 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7demochan800_t5/2026_08_03-00_48_01--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 2.322 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4426 +t=10: Selected seed 195 with value = 0.4426 +Query 1/1: Action query time = 1.987 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5407 +t=26: Selected seed 195 with value = 0.5407 +Query 1/1: Action query time = 2.883 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6308 +t=42: Selected seed 195 with value = 0.6308 +Query 1/1: Action query time = 3.168 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7169 +t=58: Selected seed 195 with value = 0.7169 +Query 1/1: Action query time = 2.502 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7691 +t=74: Selected seed 195 with value = 0.7691 +Query 1/1: Action query time = 2.545 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8073 +t=90: Selected seed 195 with value = 0.8073 +Query 1/1: Action query time = 3.249 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8862 +t=106: Selected seed 195 with value = 0.8862 +Query 1/1: Action query time = 3.499 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9539 +t=122: Selected seed 195 with value = 0.9539 +Query 1/1: Action query time = 3.413 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9619 +t=138: Selected seed 195 with value = 0.9619 +Query 1/1: Action query time = 2.995 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9719 +t=154: Selected seed 195 with value = 0.9719 +Query 1/1: Action query time = 2.739 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9711 +t=170: Selected seed 195 with value = 0.9711 +Query 1/1: Action query time = 2.731 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9747 +t=186: Selected seed 195 with value = 0.9747 +Query 1/1: Action query time = 3.303 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=202: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 2.678 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9905 +t=218: Selected seed 195 with value = 0.9905 +Query 1/1: Action query time = 2.945 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9414 +t=234: Selected seed 195 with value = 0.9414 +Query 1/1: Action query time = 2.863 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9267 +t=250: Selected seed 195 with value = 0.9267 +Query 1/1: Action query time = 3.197 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8924 +t=266: Selected seed 195 with value = 0.8924 +Query 1/1: Action query time = 2.583 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9433 +t=282: Selected seed 195 with value = 0.9433 +Query 1/1: Action query time = 2.026 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9637 +t=298: Selected seed 195 with value = 0.9637 +Saved rollout MP4 at path ./rollouts/t7demochan800_t5/2026_08_03-00_48_01--with_future_img--episode=4--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 4 +# successes: 3 (75.0%) + +Task: push the plate to the front of the stove +Starting episode 5... +Query 1/1: Action query time = 3.429 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4341 +t=10: Selected seed 195 with value = 0.4341 +Query 1/1: Action query time = 3.556 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5034 +t=26: Selected seed 195 with value = 0.5034 +Query 1/1: Action query time = 2.815 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5682 +t=42: Selected seed 195 with value = 0.5682 +Query 1/1: Action query time = 2.021 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6535 +t=58: Selected seed 195 with value = 0.6535 +Query 1/1: Action query time = 3.490 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8341 +t=74: Selected seed 195 with value = 0.8341 +Query 1/1: Action query time = 1.974 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9435 +t=90: Selected seed 195 with value = 0.9435 +Query 1/1: Action query time = 1.655 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=106: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 2.261 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.523 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.486 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.288 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.793 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.221 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.865 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7demochan800_t5/2026_08_03-00_48_01--with_future_img--episode=5--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 5 +# successes: 4 (80.0%) + +Task: push the plate to the front of the stove +Starting episode 6... +Query 1/1: Action query time = 1.471 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4515 +t=10: Selected seed 195 with value = 0.4515 +Query 1/1: Action query time = 2.259 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5198 +t=26: Selected seed 195 with value = 0.5198 +Query 1/1: Action query time = 2.629 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6088 +t=42: Selected seed 195 with value = 0.6088 +Query 1/1: Action query time = 1.876 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7596 +t=58: Selected seed 195 with value = 0.7596 +Query 1/1: Action query time = 2.375 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8626 +t=74: Selected seed 195 with value = 0.8626 +Query 1/1: Action query time = 2.305 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9476 +t=90: Selected seed 195 with value = 0.9476 +Query 1/1: Action query time = 2.805 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=106: Selected seed 195 with value = 0.9924 +Query 1/1: Action query time = 3.143 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.774 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.332 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.540 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.433 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9849 +t=186: Selected seed 195 with value = 0.9849 +Query 1/1: Action query time = 1.675 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9780 +t=202: Selected seed 195 with value = 0.9780 +Query 1/1: Action query time = 3.060 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=218: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 2.557 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=234: Selected seed 195 with value = 0.9926 +Query 1/1: Action query time = 2.768 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=250: Selected seed 195 with value = 0.9926 +Query 1/1: Action query time = 2.884 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=266: Selected seed 195 with value = 0.9946 +Saved rollout MP4 at path ./rollouts/t7demochan800_t5/2026_08_03-00_48_01--with_future_img--episode=6--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 6 +# successes: 5 (83.3%) + +Task: push the plate to the front of the stove +Starting episode 7... +Query 1/1: Action query time = 1.799 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4288 +t=10: Selected seed 195 with value = 0.4288 +Query 1/1: Action query time = 2.398 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5191 +t=26: Selected seed 195 with value = 0.5191 +Query 1/1: Action query time = 1.600 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5987 +t=42: Selected seed 195 with value = 0.5987 +Query 1/1: Action query time = 2.320 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7081 +t=58: Selected seed 195 with value = 0.7081 +Query 1/1: Action query time = 2.621 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8220 +t=74: Selected seed 195 with value = 0.8220 +Query 1/1: Action query time = 1.640 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9319 +t=90: Selected seed 195 with value = 0.9319 +Query 1/1: Action query time = 2.564 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=106: Selected seed 195 with value = 0.9909 +Query 1/1: Action query time = 2.288 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.714 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9793 +t=138: Selected seed 195 with value = 0.9793 +Query 1/1: Action query time = 2.237 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9828 +t=154: Selected seed 195 with value = 0.9828 +Query 1/1: Action query time = 2.879 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9764 +t=170: Selected seed 195 with value = 0.9764 +Query 1/1: Action query time = 1.524 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9896 +t=186: Selected seed 195 with value = 0.9896 +Query 1/1: Action query time = 2.220 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=202: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 1.681 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.394 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.951 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.682 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.428 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.659 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7demochan800_t5/2026_08_03-00_48_01--with_future_img--episode=7--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 7 +# successes: 5 (71.4%) + +Task: push the plate to the front of the stove +Starting episode 8... +Query 1/1: Action query time = 3.115 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4448 +t=10: Selected seed 195 with value = 0.4448 +Query 1/1: Action query time = 2.350 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5023 +t=26: Selected seed 195 with value = 0.5023 +Query 1/1: Action query time = 1.673 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5674 +t=42: Selected seed 195 with value = 0.5674 +Query 1/1: Action query time = 2.010 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6543 +t=58: Selected seed 195 with value = 0.6543 +Query 1/1: Action query time = 2.074 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8089 +t=74: Selected seed 195 with value = 0.8089 +Query 1/1: Action query time = 2.875 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7642 +t=90: Selected seed 195 with value = 0.7642 +Query 1/1: Action query time = 1.813 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8882 +t=106: Selected seed 195 with value = 0.8882 +Query 1/1: Action query time = 2.169 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9710 +t=122: Selected seed 195 with value = 0.9710 +Query 1/1: Action query time = 2.883 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9849 +t=138: Selected seed 195 with value = 0.9849 +Query 1/1: Action query time = 1.997 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9665 +t=154: Selected seed 195 with value = 0.9665 +Query 1/1: Action query time = 2.324 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=170: Selected seed 195 with value = 0.9955 +Query 1/1: Action query time = 2.883 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.192 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.407 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7demochan800_t5/2026_08_03-00_48_01--with_future_img--episode=8--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 8 +# successes: 6 (75.0%) + +Task: push the plate to the front of the stove +Starting episode 9... +Query 1/1: Action query time = 1.224 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4780 +t=10: Selected seed 195 with value = 0.4780 +Query 1/1: Action query time = 1.463 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5256 +t=26: Selected seed 195 with value = 0.5256 +Query 1/1: Action query time = 2.225 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5430 +t=42: Selected seed 195 with value = 0.5430 +Query 1/1: Action query time = 2.425 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7055 +t=58: Selected seed 195 with value = 0.7055 +Query 1/1: Action query time = 1.430 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8519 +t=74: Selected seed 195 with value = 0.8519 +Query 1/1: Action query time = 2.728 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9368 +t=90: Selected seed 195 with value = 0.9368 +Query 1/1: Action query time = 1.849 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9935 +t=106: Selected seed 195 with value = 0.9935 +Query 1/1: Action query time = 1.878 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.492 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.622 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=154: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 2.429 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9895 +t=170: Selected seed 195 with value = 0.9895 +Query 1/1: Action query time = 1.642 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=186: Selected seed 195 with value = 0.9955 +Query 1/1: Action query time = 2.696 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=202: Selected seed 195 with value = 0.9997 +Query 1/1: Action query time = 2.592 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.879 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.048 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.058 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.387 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.456 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7demochan800_t5/2026_08_03-00_48_01--with_future_img--episode=9--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 9 +# successes: 6 (66.7%) + +Task: push the plate to the front of the stove +Starting episode 10... +Query 1/1: Action query time = 1.582 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4406 +t=10: Selected seed 195 with value = 0.4406 +Query 1/1: Action query time = 1.386 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5138 +t=26: Selected seed 195 with value = 0.5138 +Query 1/1: Action query time = 1.441 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6048 +t=42: Selected seed 195 with value = 0.6048 +Query 1/1: Action query time = 1.743 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6574 +t=58: Selected seed 195 with value = 0.6574 +Query 1/1: Action query time = 1.084 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7911 +t=74: Selected seed 195 with value = 0.7911 +Query 1/1: Action query time = 1.391 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8876 +t=90: Selected seed 195 with value = 0.8876 +Query 1/1: Action query time = 1.155 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9632 +t=106: Selected seed 195 with value = 0.9632 +Query 1/1: Action query time = 1.032 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9802 +t=122: Selected seed 195 with value = 0.9802 +Query 1/1: Action query time = 1.192 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.295 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.781 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.018 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7demochan800_t5/2026_08_03-00_48_01--with_future_img--episode=10--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 10 +# successes: 7 (70.0%) +Current task success rate: 0.7 +Current total success rate: 0.7 +Final results: +Total episodes: 10 +Total successes: 7 +Overall success rate: 0.7000 (70.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-00_48_01--t7realcl800_t7.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-00_48_01--t7realcl800_t7.txt new file mode 100644 index 0000000000000000000000000000000000000000..fc1d07ba7decf887e081446e62720802cbcd8e2f --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-00_48_01--t7realcl800_t7.txt @@ -0,0 +1,302 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=10, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7realcl800_t7', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 2.476 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4774 +t=10: Selected seed 195 with value = 0.4774 +Query 1/1: Action query time = 3.458 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5468 +t=26: Selected seed 195 with value = 0.5468 +Query 1/1: Action query time = 3.111 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6590 +t=42: Selected seed 195 with value = 0.6590 +Query 1/1: Action query time = 3.417 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7363 +t=58: Selected seed 195 with value = 0.7363 +Query 1/1: Action query time = 3.419 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8993 +t=74: Selected seed 195 with value = 0.8993 +Query 1/1: Action query time = 2.709 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9919 +t=90: Selected seed 195 with value = 0.9919 +Saved rollout MP4 at path ./rollouts/t7realcl800_t7/2026_08_03-00_48_01--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 2.541 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4671 +t=10: Selected seed 195 with value = 0.4671 +Query 1/1: Action query time = 2.609 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5364 +t=26: Selected seed 195 with value = 0.5364 +Query 1/1: Action query time = 3.641 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6683 +t=42: Selected seed 195 with value = 0.6683 +Query 1/1: Action query time = 2.479 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8064 +t=58: Selected seed 195 with value = 0.8064 +Query 1/1: Action query time = 2.561 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8158 +t=74: Selected seed 195 with value = 0.8158 +Query 1/1: Action query time = 2.550 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9651 +t=90: Selected seed 195 with value = 0.9651 +Saved rollout MP4 at path ./rollouts/t7realcl800_t7/2026_08_03-00_48_01--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 2.391 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4928 +t=10: Selected seed 195 with value = 0.4928 +Query 1/1: Action query time = 2.538 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5681 +t=26: Selected seed 195 with value = 0.5681 +Query 1/1: Action query time = 2.632 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6499 +t=42: Selected seed 195 with value = 0.6499 +Query 1/1: Action query time = 2.321 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7906 +t=58: Selected seed 195 with value = 0.7906 +Query 1/1: Action query time = 2.472 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9718 +t=74: Selected seed 195 with value = 0.9718 +Saved rollout MP4 at path ./rollouts/t7realcl800_t7/2026_08_03-00_48_01--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 2.800 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4502 +t=10: Selected seed 195 with value = 0.4502 +Query 1/1: Action query time = 2.878 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5710 +t=26: Selected seed 195 with value = 0.5710 +Query 1/1: Action query time = 2.779 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6403 +t=42: Selected seed 195 with value = 0.6403 +Query 1/1: Action query time = 3.085 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8020 +t=58: Selected seed 195 with value = 0.8020 +Query 1/1: Action query time = 3.732 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9676 +t=74: Selected seed 195 with value = 0.9676 +Saved rollout MP4 at path ./rollouts/t7realcl800_t7/2026_08_03-00_48_01--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 2.864 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4766 +t=10: Selected seed 195 with value = 0.4766 +Query 1/1: Action query time = 3.000 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5641 +t=26: Selected seed 195 with value = 0.5641 +Query 1/1: Action query time = 2.468 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6698 +t=42: Selected seed 195 with value = 0.6698 +Query 1/1: Action query time = 3.329 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7805 +t=58: Selected seed 195 with value = 0.7805 +Query 1/1: Action query time = 2.107 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8955 +t=74: Selected seed 195 with value = 0.8955 +Saved rollout MP4 at path ./rollouts/t7realcl800_t7/2026_08_03-00_48_01--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 3.145 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4520 +t=10: Selected seed 195 with value = 0.4520 +Query 1/1: Action query time = 2.480 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5425 +t=26: Selected seed 195 with value = 0.5425 +Query 1/1: Action query time = 1.966 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6441 +t=42: Selected seed 195 with value = 0.6441 +Query 1/1: Action query time = 2.918 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7135 +t=58: Selected seed 195 with value = 0.7135 +Query 1/1: Action query time = 2.150 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8282 +t=74: Selected seed 195 with value = 0.8282 +Query 1/1: Action query time = 2.295 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9634 +t=90: Selected seed 195 with value = 0.9634 +Saved rollout MP4 at path ./rollouts/t7realcl800_t7/2026_08_03-00_48_01--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: turn on the stove +Starting episode 7... +Query 1/1: Action query time = 2.743 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4830 +t=10: Selected seed 195 with value = 0.4830 +Query 1/1: Action query time = 2.214 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5540 +t=26: Selected seed 195 with value = 0.5540 +Query 1/1: Action query time = 2.704 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6623 +t=42: Selected seed 195 with value = 0.6623 +Query 1/1: Action query time = 3.198 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7501 +t=58: Selected seed 195 with value = 0.7501 +Query 1/1: Action query time = 3.308 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8655 +t=74: Selected seed 195 with value = 0.8655 +Query 1/1: Action query time = 2.925 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7realcl800_t7/2026_08_03-00_48_01--with_future_img--episode=7--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: turn on the stove +Starting episode 8... +Query 1/1: Action query time = 3.530 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4880 +t=10: Selected seed 195 with value = 0.4880 +Query 1/1: Action query time = 2.505 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5605 +t=26: Selected seed 195 with value = 0.5605 +Query 1/1: Action query time = 3.417 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6762 +t=42: Selected seed 195 with value = 0.6762 +Query 1/1: Action query time = 3.473 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7899 +t=58: Selected seed 195 with value = 0.7899 +Query 1/1: Action query time = 3.302 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9021 +t=74: Selected seed 195 with value = 0.9021 +Query 1/1: Action query time = 1.837 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=90: Selected seed 195 with value = 0.9997 +Saved rollout MP4 at path ./rollouts/t7realcl800_t7/2026_08_03-00_48_01--with_future_img--episode=8--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: turn on the stove +Starting episode 9... +Query 1/1: Action query time = 2.558 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5022 +t=10: Selected seed 195 with value = 0.5022 +Query 1/1: Action query time = 2.658 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6349 +t=26: Selected seed 195 with value = 0.6349 +Query 1/1: Action query time = 3.432 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7262 +t=42: Selected seed 195 with value = 0.7262 +Query 1/1: Action query time = 3.097 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8156 +t=58: Selected seed 195 with value = 0.8156 +Query 1/1: Action query time = 3.511 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9336 +t=74: Selected seed 195 with value = 0.9336 +Saved rollout MP4 at path ./rollouts/t7realcl800_t7/2026_08_03-00_48_01--with_future_img--episode=9--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: turn on the stove +Starting episode 10... +Query 1/1: Action query time = 2.092 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5144 +t=10: Selected seed 195 with value = 0.5144 +Query 1/1: Action query time = 3.011 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5798 +t=26: Selected seed 195 with value = 0.5798 +Query 1/1: Action query time = 3.394 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6743 +t=42: Selected seed 195 with value = 0.6743 +Query 1/1: Action query time = 3.119 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7817 +t=58: Selected seed 195 with value = 0.7817 +Query 1/1: Action query time = 3.641 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9252 +t=74: Selected seed 195 with value = 0.9252 +Saved rollout MP4 at path ./rollouts/t7realcl800_t7/2026_08_03-00_48_01--with_future_img--episode=10--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 10 +Total successes: 10 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_08--t7rc800x50_t0_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_08--t7rc800x50_t0_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..fe26ed3e2af78a4aaebffb57b037340a4d3ae662 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_08--t7rc800x50_t0_s1.txt @@ -0,0 +1,171 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t0_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,6,11,16,21,26,31,36,41,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.273 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4445 +t=10: Selected seed 195 with value = 0.4445 +Query 1/1: Action query time = 2.411 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5234 +t=26: Selected seed 195 with value = 0.5234 +Query 1/1: Action query time = 3.399 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5723 +t=42: Selected seed 195 with value = 0.5723 +Query 1/1: Action query time = 4.953 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6795 +t=58: Selected seed 195 with value = 0.6795 +Query 1/1: Action query time = 5.126 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8216 +t=74: Selected seed 195 with value = 0.8216 +Query 1/1: Action query time = 4.210 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9429 +t=90: Selected seed 195 with value = 0.9429 +Query 1/1: Action query time = 7.340 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8600 +t=106: Selected seed 195 with value = 0.8600 +Query 1/1: Action query time = 5.412 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9079 +t=122: Selected seed 195 with value = 0.9079 +Query 1/1: Action query time = 7.271 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9138 +t=138: Selected seed 195 with value = 0.9138 +Query 1/1: Action query time = 8.818 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9064 +t=154: Selected seed 195 with value = 0.9064 +Query 1/1: Action query time = 7.729 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8686 +t=170: Selected seed 195 with value = 0.8686 +Query 1/1: Action query time = 7.149 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8768 +t=186: Selected seed 195 with value = 0.8768 +Query 1/1: Action query time = 9.065 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8760 +t=202: Selected seed 195 with value = 0.8760 +Query 1/1: Action query time = 7.659 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8658 +t=218: Selected seed 195 with value = 0.8658 +Query 1/1: Action query time = 5.202 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8659 +t=234: Selected seed 195 with value = 0.8659 +Query 1/1: Action query time = 4.538 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8679 +t=250: Selected seed 195 with value = 0.8679 +Query 1/1: Action query time = 7.051 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8699 +t=266: Selected seed 195 with value = 0.8699 +Query 1/1: Action query time = 7.232 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8642 +t=282: Selected seed 195 with value = 0.8642 +Query 1/1: Action query time = 6.116 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8663 +t=298: Selected seed 195 with value = 0.8663 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_09_08--with_future_img--episode=1--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 6.201 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4261 +t=10: Selected seed 195 with value = 0.4261 +Query 1/1: Action query time = 6.369 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4917 +t=26: Selected seed 195 with value = 0.4917 +Query 1/1: Action query time = 5.097 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5467 +t=42: Selected seed 195 with value = 0.5467 +Query 1/1: Action query time = 4.945 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6266 +t=58: Selected seed 195 with value = 0.6266 +Query 1/1: Action query time = 6.163 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7410 +t=74: Selected seed 195 with value = 0.7410 +Query 1/1: Action query time = 8.486 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9209 +t=90: Selected seed 195 with value = 0.9209 +Query 1/1: Action query time = 7.824 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9641 +t=106: Selected seed 195 with value = 0.9641 +Query 1/1: Action query time = 5.421 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9800 +t=122: Selected seed 195 with value = 0.9800 +Query 1/1: Action query time = 5.377 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9816 +t=138: Selected seed 195 with value = 0.9816 +Query 1/1: Action query time = 5.456 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_09_08--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.099 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4449 +t=10: Selected seed 195 with value = 0.4449 +Query 1/1: Action query time = 7.268 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4954 +t=26: Selected seed 195 with value = 0.4954 +Query 1/1: Action query time = 7.217 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5615 +t=42: Selected seed 195 with value = 0.5615 +Query 1/1: Action query time = 8.585 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6489 +t=58: Selected seed 195 with value = 0.6489 +Query 1/1: Action query time = 6.928 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7902 +t=74: Selected seed 195 with value = 0.7902 +Query 1/1: Action query time = 7.136 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9380 +t=90: Selected seed 195 with value = 0.9380 +Query 1/1: Action query time = 5.685 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9375 +t=106: Selected seed 195 with value = 0.9375 +Query 1/1: Action query time = 6.872 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9552 +t=122: Selected seed 195 with value = 0.9552 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_09--t7rc800x50_t0_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_09--t7rc800x50_t0_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..ab4f8985eb8752f16deb80dbcd911b76b2b0b422 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_09--t7rc800x50_t0_s0.txt @@ -0,0 +1,163 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t0_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,5,10,15,20,25,30,35,40,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.515 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4419 +t=10: Selected seed 195 with value = 0.4419 +Query 1/1: Action query time = 3.462 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5074 +t=26: Selected seed 195 with value = 0.5074 +Query 1/1: Action query time = 4.459 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5852 +t=42: Selected seed 195 with value = 0.5852 +Query 1/1: Action query time = 4.815 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6858 +t=58: Selected seed 195 with value = 0.6858 +Query 1/1: Action query time = 5.439 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8447 +t=74: Selected seed 195 with value = 0.8447 +Query 1/1: Action query time = 6.497 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9325 +t=90: Selected seed 195 with value = 0.9325 +Query 1/1: Action query time = 7.476 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9708 +t=106: Selected seed 195 with value = 0.9708 +Query 1/1: Action query time = 9.270 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 9.157 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s0/2026_08_03-01_09_09--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.428 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4523 +t=10: Selected seed 195 with value = 0.4523 +Query 1/1: Action query time = 9.227 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5182 +t=26: Selected seed 195 with value = 0.5182 +Query 1/1: Action query time = 6.958 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5816 +t=42: Selected seed 195 with value = 0.5816 +Query 1/1: Action query time = 6.947 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6851 +t=58: Selected seed 195 with value = 0.6851 +Query 1/1: Action query time = 9.145 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8358 +t=74: Selected seed 195 with value = 0.8358 +Query 1/1: Action query time = 6.329 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9426 +t=90: Selected seed 195 with value = 0.9426 +Query 1/1: Action query time = 6.653 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8893 +t=106: Selected seed 195 with value = 0.8893 +Query 1/1: Action query time = 6.047 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8797 +t=122: Selected seed 195 with value = 0.8797 +Query 1/1: Action query time = 7.143 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9011 +t=138: Selected seed 195 with value = 0.9011 +Query 1/1: Action query time = 5.205 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9261 +t=154: Selected seed 195 with value = 0.9261 +Query 1/1: Action query time = 6.208 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9393 +t=170: Selected seed 195 with value = 0.9393 +Query 1/1: Action query time = 6.058 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9490 +t=186: Selected seed 195 with value = 0.9490 +Query 1/1: Action query time = 4.861 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9526 +t=202: Selected seed 195 with value = 0.9526 +Query 1/1: Action query time = 4.068 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9561 +t=218: Selected seed 195 with value = 0.9561 +Query 1/1: Action query time = 5.358 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9604 +t=234: Selected seed 195 with value = 0.9604 +Query 1/1: Action query time = 7.925 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9655 +t=250: Selected seed 195 with value = 0.9655 +Query 1/1: Action query time = 7.694 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9704 +t=266: Selected seed 195 with value = 0.9704 +Query 1/1: Action query time = 7.159 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9727 +t=282: Selected seed 195 with value = 0.9727 +Query 1/1: Action query time = 7.629 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9755 +t=298: Selected seed 195 with value = 0.9755 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s0/2026_08_03-01_09_09--with_future_img--episode=2--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 7.569 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4422 +t=10: Selected seed 195 with value = 0.4422 +Query 1/1: Action query time = 6.978 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5072 +t=26: Selected seed 195 with value = 0.5072 +Query 1/1: Action query time = 8.118 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5815 +t=42: Selected seed 195 with value = 0.5815 +Query 1/1: Action query time = 4.985 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6585 +t=58: Selected seed 195 with value = 0.6585 +Query 1/1: Action query time = 7.901 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7797 +t=74: Selected seed 195 with value = 0.7797 +Query 1/1: Action query time = 5.246 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9797 +t=90: Selected seed 195 with value = 0.9797 +Query 1/1: Action query time = 6.088 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8036 +t=106: Selected seed 195 with value = 0.8036 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_10--t7rc800x50_t0_s2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_10--t7rc800x50_t0_s2.txt new file mode 100644 index 0000000000000000000000000000000000000000..56ff10fc7921cef063c1a80945a573026a9c8883 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_10--t7rc800x50_t0_s2.txt @@ -0,0 +1,156 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t0_s2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,7,12,17,22,27,32,37,42,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 4.811 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4574 +t=10: Selected seed 195 with value = 0.4574 +Query 1/1: Action query time = 3.897 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5186 +t=26: Selected seed 195 with value = 0.5186 +Query 1/1: Action query time = 6.171 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5885 +t=42: Selected seed 195 with value = 0.5885 +Query 1/1: Action query time = 5.796 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6913 +t=58: Selected seed 195 with value = 0.6913 +Query 1/1: Action query time = 6.836 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8264 +t=74: Selected seed 195 with value = 0.8264 +Query 1/1: Action query time = 6.875 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9206 +t=90: Selected seed 195 with value = 0.9206 +Query 1/1: Action query time = 7.241 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8169 +t=106: Selected seed 195 with value = 0.8169 +Query 1/1: Action query time = 7.110 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8948 +t=122: Selected seed 195 with value = 0.8948 +Query 1/1: Action query time = 6.499 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8854 +t=138: Selected seed 195 with value = 0.8854 +Query 1/1: Action query time = 6.854 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8750 +t=154: Selected seed 195 with value = 0.8750 +Query 1/1: Action query time = 6.797 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8785 +t=170: Selected seed 195 with value = 0.8785 +Query 1/1: Action query time = 6.214 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8754 +t=186: Selected seed 195 with value = 0.8754 +Query 1/1: Action query time = 7.379 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8656 +t=202: Selected seed 195 with value = 0.8656 +Query 1/1: Action query time = 6.002 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8659 +t=218: Selected seed 195 with value = 0.8659 +Query 1/1: Action query time = 8.193 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8663 +t=234: Selected seed 195 with value = 0.8663 +Query 1/1: Action query time = 5.970 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8620 +t=250: Selected seed 195 with value = 0.8620 +Query 1/1: Action query time = 6.514 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8570 +t=266: Selected seed 195 with value = 0.8570 +Query 1/1: Action query time = 7.693 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8466 +t=282: Selected seed 195 with value = 0.8466 +Query 1/1: Action query time = 5.582 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8376 +t=298: Selected seed 195 with value = 0.8376 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s2/2026_08_03-01_09_10--with_future_img--episode=1--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 6.457 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4385 +t=10: Selected seed 195 with value = 0.4385 +Query 1/1: Action query time = 6.292 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4958 +t=26: Selected seed 195 with value = 0.4958 +Query 1/1: Action query time = 6.782 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5648 +t=42: Selected seed 195 with value = 0.5648 +Query 1/1: Action query time = 7.609 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6565 +t=58: Selected seed 195 with value = 0.6565 +Query 1/1: Action query time = 6.267 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7783 +t=74: Selected seed 195 with value = 0.7783 +Query 1/1: Action query time = 7.130 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9125 +t=90: Selected seed 195 with value = 0.9125 +Query 1/1: Action query time = 7.207 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8184 +t=106: Selected seed 195 with value = 0.8184 +Query 1/1: Action query time = 5.088 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9023 +t=122: Selected seed 195 with value = 0.9023 +Query 1/1: Action query time = 5.150 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8950 +t=138: Selected seed 195 with value = 0.8950 +Query 1/1: Action query time = 5.324 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8816 +t=154: Selected seed 195 with value = 0.8816 +Query 1/1: Action query time = 6.926 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8789 +t=170: Selected seed 195 with value = 0.8789 +Query 1/1: Action query time = 5.955 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8720 +t=186: Selected seed 195 with value = 0.8720 +Query 1/1: Action query time = 6.031 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8585 +t=202: Selected seed 195 with value = 0.8585 +Query 1/1: Action query time = 6.404 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8588 +t=218: Selected seed 195 with value = 0.8588 +Query 1/1: Action query time = 8.061 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8610 +t=234: Selected seed 195 with value = 0.8610 +Query 1/1: Action query time = 7.211 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8655 +t=250: Selected seed 195 with value = 0.8655 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_11--t7dc800x50_t2_s2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_11--t7dc800x50_t2_s2.txt new file mode 100644 index 0000000000000000000000000000000000000000..5961b29d28d49213f21a061eecb42caef239bc06 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_11--t7dc800x50_t2_s2.txt @@ -0,0 +1,152 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_demochan800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7dc800x50_t2_s2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,7,12,17,22,27,32,37,42,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 8.336 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5860 +t=10: Selected seed 195 with value = 0.5860 +Query 1/1: Action query time = 7.207 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7335 +t=26: Selected seed 195 with value = 0.7335 +Query 1/1: Action query time = 6.949 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7511 +t=42: Selected seed 195 with value = 0.7511 +Query 1/1: Action query time = 5.706 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8867 +t=58: Selected seed 195 with value = 0.8867 +Query 1/1: Action query time = 8.186 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9055 +t=74: Selected seed 195 with value = 0.9055 +Query 1/1: Action query time = 8.450 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8899 +t=90: Selected seed 195 with value = 0.8899 +Query 1/1: Action query time = 6.185 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8845 +t=106: Selected seed 195 with value = 0.8845 +Query 1/1: Action query time = 7.303 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9008 +t=122: Selected seed 195 with value = 0.9008 +Query 1/1: Action query time = 7.022 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9043 +t=138: Selected seed 195 with value = 0.9043 +Query 1/1: Action query time = 6.434 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9080 +t=154: Selected seed 195 with value = 0.9080 +Query 1/1: Action query time = 7.864 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8978 +t=170: Selected seed 195 with value = 0.8978 +Query 1/1: Action query time = 8.701 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8743 +t=186: Selected seed 195 with value = 0.8743 +Query 1/1: Action query time = 9.712 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8479 +t=202: Selected seed 195 with value = 0.8479 +Query 1/1: Action query time = 6.210 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8397 +t=218: Selected seed 195 with value = 0.8397 +Query 1/1: Action query time = 8.086 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8470 +t=234: Selected seed 195 with value = 0.8470 +Query 1/1: Action query time = 6.415 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8805 +t=250: Selected seed 195 with value = 0.8805 +Query 1/1: Action query time = 6.303 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9111 +t=266: Selected seed 195 with value = 0.9111 +Query 1/1: Action query time = 4.651 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9122 +t=282: Selected seed 195 with value = 0.9122 +Query 1/1: Action query time = 4.697 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8627 +t=298: Selected seed 195 with value = 0.8627 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s2/2026_08_03-01_09_11--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 6.901 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5769 +t=10: Selected seed 195 with value = 0.5769 +Query 1/1: Action query time = 7.089 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7187 +t=26: Selected seed 195 with value = 0.7187 +Query 1/1: Action query time = 7.262 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8931 +t=42: Selected seed 195 with value = 0.8931 +Query 1/1: Action query time = 7.302 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9255 +t=58: Selected seed 195 with value = 0.9255 +Query 1/1: Action query time = 5.978 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8881 +t=74: Selected seed 195 with value = 0.8881 +Query 1/1: Action query time = 5.183 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8830 +t=90: Selected seed 195 with value = 0.8830 +Query 1/1: Action query time = 5.267 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9067 +t=106: Selected seed 195 with value = 0.9067 +Query 1/1: Action query time = 5.286 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8921 +t=122: Selected seed 195 with value = 0.8921 +Query 1/1: Action query time = 6.592 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8976 +t=138: Selected seed 195 with value = 0.8976 +Query 1/1: Action query time = 5.178 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9136 +t=154: Selected seed 195 with value = 0.9136 +Query 1/1: Action query time = 6.553 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8868 +t=170: Selected seed 195 with value = 0.8868 +Query 1/1: Action query time = 6.613 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9098 +t=186: Selected seed 195 with value = 0.9098 +Query 1/1: Action query time = 5.183 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9122 +t=202: Selected seed 195 with value = 0.9122 +Query 1/1: Action query time = 5.874 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8236 +t=218: Selected seed 195 with value = 0.8236 +Query 1/1: Action query time = 4.360 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8027 +t=234: Selected seed 195 with value = 0.8027 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_11--t7rc800x50_t2_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_11--t7rc800x50_t2_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..b63cbe6339ea3ab5cc84ddfcc7e462143e0b524c --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_11--t7rc800x50_t2_s0.txt @@ -0,0 +1,168 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t2_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,5,10,15,20,25,30,35,40,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.953 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5962 +t=10: Selected seed 195 with value = 0.5962 +Query 1/1: Action query time = 2.234 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6981 +t=26: Selected seed 195 with value = 0.6981 +Query 1/1: Action query time = 2.485 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8009 +t=42: Selected seed 195 with value = 0.8009 +Query 1/1: Action query time = 4.447 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8637 +t=58: Selected seed 195 with value = 0.8637 +Query 1/1: Action query time = 6.446 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=74: Selected seed 195 with value = 0.9955 +Query 1/1: Action query time = 7.094 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.988 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 8.345 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 9.854 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.433 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.214 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.280 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 8.315 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.767 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.350 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.487 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 8.179 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 8.026 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.024 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s0/2026_08_03-01_09_11--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.904 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5874 +t=10: Selected seed 195 with value = 0.5874 +Query 1/1: Action query time = 2.845 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6144 +t=26: Selected seed 195 with value = 0.6144 +Query 1/1: Action query time = 3.226 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7125 +t=42: Selected seed 195 with value = 0.7125 +Query 1/1: Action query time = 4.179 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8694 +t=58: Selected seed 195 with value = 0.8694 +Query 1/1: Action query time = 4.196 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7615 +t=74: Selected seed 195 with value = 0.7615 +Query 1/1: Action query time = 4.334 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7806 +t=90: Selected seed 195 with value = 0.7806 +Query 1/1: Action query time = 6.917 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7881 +t=106: Selected seed 195 with value = 0.7881 +Query 1/1: Action query time = 7.806 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9129 +t=122: Selected seed 195 with value = 0.9129 +Query 1/1: Action query time = 5.851 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9506 +t=138: Selected seed 195 with value = 0.9506 +Query 1/1: Action query time = 7.177 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9354 +t=154: Selected seed 195 with value = 0.9354 +Query 1/1: Action query time = 7.164 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9320 +t=170: Selected seed 195 with value = 0.9320 +Query 1/1: Action query time = 6.496 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9347 +t=186: Selected seed 195 with value = 0.9347 +Query 1/1: Action query time = 7.697 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8835 +t=202: Selected seed 195 with value = 0.8835 +Query 1/1: Action query time = 7.587 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7826 +t=218: Selected seed 195 with value = 0.7826 +Query 1/1: Action query time = 6.375 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8583 +t=234: Selected seed 195 with value = 0.8583 +Query 1/1: Action query time = 6.463 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9059 +t=250: Selected seed 195 with value = 0.9059 +Query 1/1: Action query time = 6.789 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8795 +t=266: Selected seed 195 with value = 0.8795 +Query 1/1: Action query time = 5.965 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8683 +t=282: Selected seed 195 with value = 0.8683 +Query 1/1: Action query time = 6.346 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8564 +t=298: Selected seed 195 with value = 0.8564 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_11--t7rc800x50_t2_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_11--t7rc800x50_t2_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..4215b81194c5764bdfd062c1cafe7a6376c5dd2b --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_11--t7rc800x50_t2_s1.txt @@ -0,0 +1,148 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t2_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,6,11,16,21,26,31,36,41,46', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 10.930 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5554 +t=10: Selected seed 195 with value = 0.5554 +Query 1/1: Action query time = 7.238 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6360 +t=26: Selected seed 195 with value = 0.6360 +Query 1/1: Action query time = 7.096 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7377 +t=42: Selected seed 195 with value = 0.7377 +Query 1/1: Action query time = 6.018 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8774 +t=58: Selected seed 195 with value = 0.8774 +Query 1/1: Action query time = 8.406 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9819 +t=74: Selected seed 195 with value = 0.9819 +Query 1/1: Action query time = 7.433 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.597 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.687 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.408 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.843 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.438 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.107 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 8.514 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.675 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.950 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.992 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.408 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.210 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.416 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s1/2026_08_03-01_09_11--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 7.305 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5593 +t=10: Selected seed 195 with value = 0.5593 +Query 1/1: Action query time = 7.328 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6819 +t=26: Selected seed 195 with value = 0.6819 +Query 1/1: Action query time = 7.493 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7576 +t=42: Selected seed 195 with value = 0.7576 +Query 1/1: Action query time = 6.529 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8711 +t=58: Selected seed 195 with value = 0.8711 +Query 1/1: Action query time = 8.345 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9672 +t=74: Selected seed 195 with value = 0.9672 +Query 1/1: Action query time = 8.184 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.066 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.186 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9310 +t=122: Selected seed 195 with value = 0.9310 +Query 1/1: Action query time = 7.504 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9934 +t=138: Selected seed 195 with value = 0.9934 +Query 1/1: Action query time = 8.376 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.316 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 9.373 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 8.278 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.677 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_12--t7dc800x50_t0_s2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_12--t7dc800x50_t0_s2.txt new file mode 100644 index 0000000000000000000000000000000000000000..6bc9ccf3c860722ee7924a370e3556dd02b13195 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_12--t7dc800x50_t0_s2.txt @@ -0,0 +1,132 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_demochan800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7dc800x50_t0_s2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,7,12,17,22,27,32,37,42,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 10.722 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4610 +t=10: Selected seed 195 with value = 0.4610 +Query 1/1: Action query time = 9.374 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5109 +t=26: Selected seed 195 with value = 0.5109 +Query 1/1: Action query time = 7.873 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5770 +t=42: Selected seed 195 with value = 0.5770 +Query 1/1: Action query time = 8.770 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6845 +t=58: Selected seed 195 with value = 0.6845 +Query 1/1: Action query time = 9.812 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8748 +t=74: Selected seed 195 with value = 0.8748 +Query 1/1: Action query time = 7.188 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 9.530 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8971 +t=106: Selected seed 195 with value = 0.8971 +Query 1/1: Action query time = 9.149 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8495 +t=122: Selected seed 195 with value = 0.8495 +Query 1/1: Action query time = 7.169 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8163 +t=138: Selected seed 195 with value = 0.8163 +Query 1/1: Action query time = 6.952 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8045 +t=154: Selected seed 195 with value = 0.8045 +Query 1/1: Action query time = 7.096 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8105 +t=170: Selected seed 195 with value = 0.8105 +Query 1/1: Action query time = 9.022 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8136 +t=186: Selected seed 195 with value = 0.8136 +Query 1/1: Action query time = 8.061 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8154 +t=202: Selected seed 195 with value = 0.8154 +Query 1/1: Action query time = 7.019 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8187 +t=218: Selected seed 195 with value = 0.8187 +Query 1/1: Action query time = 6.111 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8141 +t=234: Selected seed 195 with value = 0.8141 +Query 1/1: Action query time = 7.030 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8129 +t=250: Selected seed 195 with value = 0.8129 +Query 1/1: Action query time = 6.374 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8420 +t=266: Selected seed 195 with value = 0.8420 +Query 1/1: Action query time = 6.488 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8486 +t=282: Selected seed 195 with value = 0.8486 +Query 1/1: Action query time = 3.902 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8495 +t=298: Selected seed 195 with value = 0.8495 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t0_s2/2026_08_03-01_09_12--with_future_img--episode=1--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 6.910 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4384 +t=10: Selected seed 195 with value = 0.4384 +Query 1/1: Action query time = 8.134 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4973 +t=26: Selected seed 195 with value = 0.4973 +Query 1/1: Action query time = 7.696 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5661 +t=42: Selected seed 195 with value = 0.5661 +Query 1/1: Action query time = 9.223 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6400 +t=58: Selected seed 195 with value = 0.6400 +Query 1/1: Action query time = 8.193 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8020 +t=74: Selected seed 195 with value = 0.8020 +Query 1/1: Action query time = 8.260 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.572 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.199 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8953 +t=122: Selected seed 195 with value = 0.8953 +Query 1/1: Action query time = 8.821 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8853 +t=138: Selected seed 195 with value = 0.8853 +Query 1/1: Action query time = 9.076 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8829 +t=154: Selected seed 195 with value = 0.8829 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_12--t7dc800x50_t3_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_12--t7dc800x50_t3_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..9c24e7c021c1c4e24abfa3967c1d9f0e7865e8f0 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_12--t7dc800x50_t3_s3.txt @@ -0,0 +1,156 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_demochan800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7dc800x50_t3_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,8,13,18,23,28,33,38,43,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 8.988 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4711 +t=10: Selected seed 195 with value = 0.4711 +Query 1/1: Action query time = 6.245 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6095 +t=26: Selected seed 195 with value = 0.6095 +Query 1/1: Action query time = 4.330 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8133 +t=42: Selected seed 195 with value = 0.8133 +Query 1/1: Action query time = 5.829 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9347 +t=58: Selected seed 195 with value = 0.9347 +Query 1/1: Action query time = 6.514 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9126 +t=74: Selected seed 195 with value = 0.9126 +Query 1/1: Action query time = 6.456 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9237 +t=90: Selected seed 195 with value = 0.9237 +Query 1/1: Action query time = 5.965 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8788 +t=106: Selected seed 195 with value = 0.8788 +Query 1/1: Action query time = 6.465 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8798 +t=122: Selected seed 195 with value = 0.8798 +Query 1/1: Action query time = 5.598 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8868 +t=138: Selected seed 195 with value = 0.8868 +Query 1/1: Action query time = 6.106 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8935 +t=154: Selected seed 195 with value = 0.8935 +Query 1/1: Action query time = 6.152 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8979 +t=170: Selected seed 195 with value = 0.8979 +Query 1/1: Action query time = 5.849 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9267 +t=186: Selected seed 195 with value = 0.9267 +Query 1/1: Action query time = 8.821 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8659 +t=202: Selected seed 195 with value = 0.8659 +Query 1/1: Action query time = 9.604 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9206 +t=218: Selected seed 195 with value = 0.9206 +Query 1/1: Action query time = 6.801 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8549 +t=234: Selected seed 195 with value = 0.8549 +Query 1/1: Action query time = 8.016 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8620 +t=250: Selected seed 195 with value = 0.8620 +Query 1/1: Action query time = 8.523 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8443 +t=266: Selected seed 195 with value = 0.8443 +Query 1/1: Action query time = 5.127 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8487 +t=282: Selected seed 195 with value = 0.8487 +Query 1/1: Action query time = 5.173 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8285 +t=298: Selected seed 195 with value = 0.8285 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t3_s3/2026_08_03-01_09_12--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.602 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4660 +t=10: Selected seed 195 with value = 0.4660 +Query 1/1: Action query time = 4.737 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7309 +t=26: Selected seed 195 with value = 0.7309 +Query 1/1: Action query time = 7.580 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7999 +t=42: Selected seed 195 with value = 0.7999 +Query 1/1: Action query time = 6.900 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9213 +t=58: Selected seed 195 with value = 0.9213 +Query 1/1: Action query time = 7.215 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8963 +t=74: Selected seed 195 with value = 0.8963 +Query 1/1: Action query time = 6.888 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8892 +t=90: Selected seed 195 with value = 0.8892 +Query 1/1: Action query time = 6.550 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9130 +t=106: Selected seed 195 with value = 0.9130 +Query 1/1: Action query time = 8.560 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8895 +t=122: Selected seed 195 with value = 0.8895 +Query 1/1: Action query time = 8.916 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8998 +t=138: Selected seed 195 with value = 0.8998 +Query 1/1: Action query time = 6.119 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8726 +t=154: Selected seed 195 with value = 0.8726 +Query 1/1: Action query time = 5.917 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8872 +t=170: Selected seed 195 with value = 0.8872 +Query 1/1: Action query time = 7.882 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8548 +t=186: Selected seed 195 with value = 0.8548 +Query 1/1: Action query time = 4.453 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8789 +t=202: Selected seed 195 with value = 0.8789 +Query 1/1: Action query time = 5.985 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7926 +t=218: Selected seed 195 with value = 0.7926 +Query 1/1: Action query time = 6.284 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8615 +t=234: Selected seed 195 with value = 0.8615 +Query 1/1: Action query time = 7.348 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8797 +t=250: Selected seed 195 with value = 0.8797 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_13--t7dc800x50_t2_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_13--t7dc800x50_t2_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..e1f7cd635005cf7ecfc629c439b36e15ab0bca66 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_13--t7dc800x50_t2_s0.txt @@ -0,0 +1,128 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_demochan800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7dc800x50_t2_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,5,10,15,20,25,30,35,40,45', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 8.599 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=10: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 7.819 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6835 +t=26: Selected seed 195 with value = 0.6835 +Query 1/1: Action query time = 8.498 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7952 +t=42: Selected seed 195 with value = 0.7952 +Query 1/1: Action query time = 9.104 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7994 +t=58: Selected seed 195 with value = 0.7994 +Query 1/1: Action query time = 6.225 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9282 +t=74: Selected seed 195 with value = 0.9282 +Query 1/1: Action query time = 12.189 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9666 +t=90: Selected seed 195 with value = 0.9666 +Query 1/1: Action query time = 8.806 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9392 +t=106: Selected seed 195 with value = 0.9392 +Query 1/1: Action query time = 9.600 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8933 +t=122: Selected seed 195 with value = 0.8933 +Query 1/1: Action query time = 10.259 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9194 +t=138: Selected seed 195 with value = 0.9194 +Query 1/1: Action query time = 8.621 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9090 +t=154: Selected seed 195 with value = 0.9090 +Query 1/1: Action query time = 8.548 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9098 +t=170: Selected seed 195 with value = 0.9098 +Query 1/1: Action query time = 8.887 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9077 +t=186: Selected seed 195 with value = 0.9077 +Query 1/1: Action query time = 7.928 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9102 +t=202: Selected seed 195 with value = 0.9102 +Query 1/1: Action query time = 8.774 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9018 +t=218: Selected seed 195 with value = 0.9018 +Query 1/1: Action query time = 7.393 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9039 +t=234: Selected seed 195 with value = 0.9039 +Query 1/1: Action query time = 5.636 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9173 +t=250: Selected seed 195 with value = 0.9173 +Query 1/1: Action query time = 6.898 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9072 +t=266: Selected seed 195 with value = 0.9072 +Query 1/1: Action query time = 6.773 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9011 +t=282: Selected seed 195 with value = 0.9011 +Query 1/1: Action query time = 5.505 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9038 +t=298: Selected seed 195 with value = 0.9038 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s0/2026_08_03-01_09_13--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 7.726 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5849 +t=10: Selected seed 195 with value = 0.5849 +Query 1/1: Action query time = 7.541 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6259 +t=26: Selected seed 195 with value = 0.6259 +Query 1/1: Action query time = 8.475 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7223 +t=42: Selected seed 195 with value = 0.7223 +Query 1/1: Action query time = 6.153 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8330 +t=58: Selected seed 195 with value = 0.8330 +Query 1/1: Action query time = 8.633 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8531 +t=74: Selected seed 195 with value = 0.8531 +Query 1/1: Action query time = 11.814 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8842 +t=90: Selected seed 195 with value = 0.8842 +Query 1/1: Action query time = 9.362 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8778 +t=106: Selected seed 195 with value = 0.8778 +Query 1/1: Action query time = 5.613 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8963 +t=122: Selected seed 195 with value = 0.8963 +Query 1/1: Action query time = 5.998 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9046 +t=138: Selected seed 195 with value = 0.9046 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_14--t7dc800x50_t2_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_14--t7dc800x50_t2_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..f52fa0baf11d5a892a15e3f8547740364e54500c --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_09_14--t7dc800x50_t2_s3.txt @@ -0,0 +1,132 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_demochan800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7dc800x50_t2_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,8,13,18,23,28,33,38,43,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.093 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5691 +t=10: Selected seed 195 with value = 0.5691 +Query 1/1: Action query time = 6.314 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6818 +t=26: Selected seed 195 with value = 0.6818 +Query 1/1: Action query time = 7.579 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7292 +t=42: Selected seed 195 with value = 0.7292 +Query 1/1: Action query time = 8.746 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8848 +t=58: Selected seed 195 with value = 0.8848 +Query 1/1: Action query time = 9.798 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8784 +t=74: Selected seed 195 with value = 0.8784 +Query 1/1: Action query time = 7.974 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8831 +t=90: Selected seed 195 with value = 0.8831 +Query 1/1: Action query time = 8.184 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9009 +t=106: Selected seed 195 with value = 0.9009 +Query 1/1: Action query time = 10.804 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9163 +t=122: Selected seed 195 with value = 0.9163 +Query 1/1: Action query time = 9.065 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9335 +t=138: Selected seed 195 with value = 0.9335 +Query 1/1: Action query time = 9.526 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9400 +t=154: Selected seed 195 with value = 0.9400 +Query 1/1: Action query time = 9.940 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9321 +t=170: Selected seed 195 with value = 0.9321 +Query 1/1: Action query time = 10.468 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9341 +t=186: Selected seed 195 with value = 0.9341 +Query 1/1: Action query time = 7.849 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9433 +t=202: Selected seed 195 with value = 0.9433 +Query 1/1: Action query time = 7.526 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9457 +t=218: Selected seed 195 with value = 0.9457 +Query 1/1: Action query time = 7.825 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8874 +t=234: Selected seed 195 with value = 0.8874 +Query 1/1: Action query time = 6.736 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8077 +t=250: Selected seed 195 with value = 0.8077 +Query 1/1: Action query time = 6.082 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8012 +t=266: Selected seed 195 with value = 0.8012 +Query 1/1: Action query time = 6.380 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7980 +t=282: Selected seed 195 with value = 0.7980 +Query 1/1: Action query time = 6.340 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7913 +t=298: Selected seed 195 with value = 0.7913 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_09_14--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.261 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5648 +t=10: Selected seed 195 with value = 0.5648 +Query 1/1: Action query time = 4.903 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6404 +t=26: Selected seed 195 with value = 0.6404 +Query 1/1: Action query time = 6.580 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7124 +t=42: Selected seed 195 with value = 0.7124 +Query 1/1: Action query time = 7.839 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8842 +t=58: Selected seed 195 with value = 0.8842 +Query 1/1: Action query time = 9.128 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8676 +t=74: Selected seed 195 with value = 0.8676 +Query 1/1: Action query time = 11.682 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8920 +t=90: Selected seed 195 with value = 0.8920 +Query 1/1: Action query time = 8.503 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9057 +t=106: Selected seed 195 with value = 0.9057 +Query 1/1: Action query time = 8.048 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9012 +t=122: Selected seed 195 with value = 0.9012 +Query 1/1: Action query time = 6.906 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9234 +t=138: Selected seed 195 with value = 0.9234 +Query 1/1: Action query time = 8.272 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9317 +t=154: Selected seed 195 with value = 0.9317 diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_45--t7rc800x50_t0_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_45--t7rc800x50_t0_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..d1b70874e6cf2e8db53b87d02864a02343b61a42 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_45--t7rc800x50_t0_s1.txt @@ -0,0 +1,947 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t0_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.021 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4445 +t=10: Selected seed 195 with value = 0.4445 +Query 1/1: Action query time = 2.410 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5234 +t=26: Selected seed 195 with value = 0.5234 +Query 1/1: Action query time = 3.946 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5723 +t=42: Selected seed 195 with value = 0.5723 +Query 1/1: Action query time = 5.359 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6795 +t=58: Selected seed 195 with value = 0.6795 +Query 1/1: Action query time = 7.452 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8216 +t=74: Selected seed 195 with value = 0.8216 +Query 1/1: Action query time = 5.201 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9429 +t=90: Selected seed 195 with value = 0.9429 +Query 1/1: Action query time = 5.401 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8600 +t=106: Selected seed 195 with value = 0.8600 +Query 1/1: Action query time = 6.436 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9079 +t=122: Selected seed 195 with value = 0.9079 +Query 1/1: Action query time = 6.664 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9138 +t=138: Selected seed 195 with value = 0.9138 +Query 1/1: Action query time = 5.867 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9064 +t=154: Selected seed 195 with value = 0.9064 +Query 1/1: Action query time = 6.097 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8686 +t=170: Selected seed 195 with value = 0.8686 +Query 1/1: Action query time = 4.848 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8768 +t=186: Selected seed 195 with value = 0.8768 +Query 1/1: Action query time = 6.521 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8760 +t=202: Selected seed 195 with value = 0.8760 +Query 1/1: Action query time = 5.901 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8658 +t=218: Selected seed 195 with value = 0.8658 +Query 1/1: Action query time = 5.249 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8659 +t=234: Selected seed 195 with value = 0.8659 +Query 1/1: Action query time = 6.498 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8679 +t=250: Selected seed 195 with value = 0.8679 +Query 1/1: Action query time = 6.278 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8699 +t=266: Selected seed 195 with value = 0.8699 +Query 1/1: Action query time = 4.635 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8642 +t=282: Selected seed 195 with value = 0.8642 +Query 1/1: Action query time = 5.020 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8663 +t=298: Selected seed 195 with value = 0.8663 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=1--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.739 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4523 +t=10: Selected seed 195 with value = 0.4523 +Query 1/1: Action query time = 5.176 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5182 +t=26: Selected seed 195 with value = 0.5182 +Query 1/1: Action query time = 5.287 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5819 +t=42: Selected seed 195 with value = 0.5819 +Query 1/1: Action query time = 5.866 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6844 +t=58: Selected seed 195 with value = 0.6844 +Query 1/1: Action query time = 5.956 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8370 +t=74: Selected seed 195 with value = 0.8370 +Query 1/1: Action query time = 5.268 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9436 +t=90: Selected seed 195 with value = 0.9436 +Query 1/1: Action query time = 6.516 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8874 +t=106: Selected seed 195 with value = 0.8874 +Query 1/1: Action query time = 4.582 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8817 +t=122: Selected seed 195 with value = 0.8817 +Query 1/1: Action query time = 5.697 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8877 +t=138: Selected seed 195 with value = 0.8877 +Query 1/1: Action query time = 5.601 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8862 +t=154: Selected seed 195 with value = 0.8862 +Query 1/1: Action query time = 6.276 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8869 +t=170: Selected seed 195 with value = 0.8869 +Query 1/1: Action query time = 5.062 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8935 +t=186: Selected seed 195 with value = 0.8935 +Query 1/1: Action query time = 4.928 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8948 +t=202: Selected seed 195 with value = 0.8948 +Query 1/1: Action query time = 6.180 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8982 +t=218: Selected seed 195 with value = 0.8982 +Query 1/1: Action query time = 6.580 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9027 +t=234: Selected seed 195 with value = 0.9027 +Query 1/1: Action query time = 5.579 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9033 +t=250: Selected seed 195 with value = 0.9033 +Query 1/1: Action query time = 5.841 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9057 +t=266: Selected seed 195 with value = 0.9057 +Query 1/1: Action query time = 6.900 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9083 +t=282: Selected seed 195 with value = 0.9083 +Query 1/1: Action query time = 5.500 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9087 +t=298: Selected seed 195 with value = 0.9087 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=2--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.346 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4677 +t=10: Selected seed 195 with value = 0.4677 +Query 1/1: Action query time = 4.842 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5373 +t=26: Selected seed 195 with value = 0.5373 +Query 1/1: Action query time = 4.938 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5951 +t=42: Selected seed 195 with value = 0.5951 +Query 1/1: Action query time = 5.290 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7341 +t=58: Selected seed 195 with value = 0.7341 +Query 1/1: Action query time = 4.255 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8615 +t=74: Selected seed 195 with value = 0.8615 +Query 1/1: Action query time = 5.350 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9120 +t=90: Selected seed 195 with value = 0.9120 +Query 1/1: Action query time = 5.811 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8701 +t=106: Selected seed 195 with value = 0.8701 +Query 1/1: Action query time = 5.603 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9399 +t=122: Selected seed 195 with value = 0.9399 +Query 1/1: Action query time = 5.531 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9671 +t=138: Selected seed 195 with value = 0.9671 +Query 1/1: Action query time = 6.606 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9329 +t=154: Selected seed 195 with value = 0.9329 +Query 1/1: Action query time = 5.143 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8713 +t=170: Selected seed 195 with value = 0.8713 +Query 1/1: Action query time = 3.897 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8743 +t=186: Selected seed 195 with value = 0.8743 +Query 1/1: Action query time = 5.399 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8760 +t=202: Selected seed 195 with value = 0.8760 +Query 1/1: Action query time = 4.996 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8663 +t=218: Selected seed 195 with value = 0.8663 +Query 1/1: Action query time = 4.683 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8699 +t=234: Selected seed 195 with value = 0.8699 +Query 1/1: Action query time = 3.940 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8665 +t=250: Selected seed 195 with value = 0.8665 +Query 1/1: Action query time = 4.733 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8676 +t=266: Selected seed 195 with value = 0.8676 +Query 1/1: Action query time = 4.943 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8652 +t=282: Selected seed 195 with value = 0.8652 +Query 1/1: Action query time = 3.806 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8631 +t=298: Selected seed 195 with value = 0.8631 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=3--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 4... +Query 1/1: Action query time = 5.414 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4451 +t=10: Selected seed 195 with value = 0.4451 +Query 1/1: Action query time = 5.722 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4975 +t=26: Selected seed 195 with value = 0.4975 +Query 1/1: Action query time = 3.456 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5693 +t=42: Selected seed 195 with value = 0.5693 +Query 1/1: Action query time = 6.813 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6770 +t=58: Selected seed 195 with value = 0.6770 +Query 1/1: Action query time = 5.105 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7971 +t=74: Selected seed 195 with value = 0.7971 +Query 1/1: Action query time = 3.859 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9042 +t=90: Selected seed 195 with value = 0.9042 +Query 1/1: Action query time = 4.747 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8079 +t=106: Selected seed 195 with value = 0.8079 +Query 1/1: Action query time = 4.956 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8804 +t=122: Selected seed 195 with value = 0.8804 +Query 1/1: Action query time = 4.267 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8783 +t=138: Selected seed 195 with value = 0.8783 +Query 1/1: Action query time = 6.384 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8772 +t=154: Selected seed 195 with value = 0.8772 +Query 1/1: Action query time = 5.981 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8722 +t=170: Selected seed 195 with value = 0.8722 +Query 1/1: Action query time = 6.033 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8770 +t=186: Selected seed 195 with value = 0.8770 +Query 1/1: Action query time = 4.077 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8709 +t=202: Selected seed 195 with value = 0.8709 +Query 1/1: Action query time = 4.114 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8659 +t=218: Selected seed 195 with value = 0.8659 +Query 1/1: Action query time = 5.210 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8578 +t=234: Selected seed 195 with value = 0.8578 +Query 1/1: Action query time = 3.695 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8583 +t=250: Selected seed 195 with value = 0.8583 +Query 1/1: Action query time = 3.796 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8656 +t=266: Selected seed 195 with value = 0.8656 +Query 1/1: Action query time = 5.043 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8544 +t=282: Selected seed 195 with value = 0.8544 +Query 1/1: Action query time = 4.424 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8549 +t=298: Selected seed 195 with value = 0.8549 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=4--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 5... +Query 1/1: Action query time = 4.698 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4421 +t=10: Selected seed 195 with value = 0.4421 +Query 1/1: Action query time = 4.992 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5177 +t=26: Selected seed 195 with value = 0.5177 +Query 1/1: Action query time = 5.207 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6006 +t=42: Selected seed 195 with value = 0.6006 +Query 1/1: Action query time = 5.365 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7126 +t=58: Selected seed 195 with value = 0.7126 +Query 1/1: Action query time = 5.199 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8654 +t=74: Selected seed 195 with value = 0.8654 +Query 1/1: Action query time = 3.646 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9167 +t=90: Selected seed 195 with value = 0.9167 +Query 1/1: Action query time = 5.556 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9128 +t=106: Selected seed 195 with value = 0.9128 +Query 1/1: Action query time = 4.684 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8723 +t=122: Selected seed 195 with value = 0.8723 +Query 1/1: Action query time = 4.440 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9285 +t=138: Selected seed 195 with value = 0.9285 +Query 1/1: Action query time = 5.219 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9715 +t=154: Selected seed 195 with value = 0.9715 +Query 1/1: Action query time = 4.067 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8658 +t=170: Selected seed 195 with value = 0.8658 +Query 1/1: Action query time = 4.253 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8577 +t=186: Selected seed 195 with value = 0.8577 +Query 1/1: Action query time = 5.938 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8789 +t=202: Selected seed 195 with value = 0.8789 +Query 1/1: Action query time = 6.243 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8739 +t=218: Selected seed 195 with value = 0.8739 +Query 1/1: Action query time = 5.244 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8546 +t=234: Selected seed 195 with value = 0.8546 +Query 1/1: Action query time = 5.590 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8637 +t=250: Selected seed 195 with value = 0.8637 +Query 1/1: Action query time = 3.720 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8605 +t=266: Selected seed 195 with value = 0.8605 +Query 1/1: Action query time = 4.961 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8663 +t=282: Selected seed 195 with value = 0.8663 +Query 1/1: Action query time = 3.799 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8626 +t=298: Selected seed 195 with value = 0.8626 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=5--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 5 +# successes: 0 (0.0%) + +Task: open the middle drawer of the cabinet +Starting episode 6... +Query 1/1: Action query time = 4.237 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4428 +t=10: Selected seed 195 with value = 0.4428 +Query 1/1: Action query time = 4.671 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4834 +t=26: Selected seed 195 with value = 0.4834 +Query 1/1: Action query time = 5.845 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5641 +t=42: Selected seed 195 with value = 0.5641 +Query 1/1: Action query time = 4.946 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6672 +t=58: Selected seed 195 with value = 0.6672 +Query 1/1: Action query time = 4.811 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8208 +t=74: Selected seed 195 with value = 0.8208 +Query 1/1: Action query time = 4.693 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9051 +t=90: Selected seed 195 with value = 0.9051 +Query 1/1: Action query time = 5.754 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9614 +t=106: Selected seed 195 with value = 0.9614 +Query 1/1: Action query time = 4.869 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9856 +t=122: Selected seed 195 with value = 0.9856 +Query 1/1: Action query time = 4.617 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=6--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 6 +# successes: 1 (16.7%) + +Task: open the middle drawer of the cabinet +Starting episode 7... +Query 1/1: Action query time = 4.635 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4406 +t=10: Selected seed 195 with value = 0.4406 +Query 1/1: Action query time = 5.563 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4872 +t=26: Selected seed 195 with value = 0.4872 +Query 1/1: Action query time = 4.740 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5599 +t=42: Selected seed 195 with value = 0.5599 +Query 1/1: Action query time = 6.328 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6387 +t=58: Selected seed 195 with value = 0.6387 +Query 1/1: Action query time = 5.769 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7576 +t=74: Selected seed 195 with value = 0.7576 +Query 1/1: Action query time = 6.466 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8940 +t=90: Selected seed 195 with value = 0.8940 +Query 1/1: Action query time = 5.079 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8375 +t=106: Selected seed 195 with value = 0.8375 +Query 1/1: Action query time = 4.173 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=122: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 3.320 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8884 +t=138: Selected seed 195 with value = 0.8884 +Query 1/1: Action query time = 4.742 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=154: Selected seed 195 with value = 0.9912 +Query 1/1: Action query time = 3.870 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.706 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=7--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 7 +# successes: 2 (28.6%) + +Task: open the middle drawer of the cabinet +Starting episode 8... +Query 1/1: Action query time = 4.511 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4382 +t=10: Selected seed 195 with value = 0.4382 +Query 1/1: Action query time = 5.228 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4972 +t=26: Selected seed 195 with value = 0.4972 +Query 1/1: Action query time = 4.819 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5458 +t=42: Selected seed 195 with value = 0.5458 +Query 1/1: Action query time = 6.780 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6662 +t=58: Selected seed 195 with value = 0.6662 +Query 1/1: Action query time = 6.519 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8293 +t=74: Selected seed 195 with value = 0.8293 +Query 1/1: Action query time = 6.801 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9531 +t=90: Selected seed 195 with value = 0.9531 +Query 1/1: Action query time = 4.163 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.869 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=8--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 8 +# successes: 3 (37.5%) + +Task: open the middle drawer of the cabinet +Starting episode 9... +Query 1/1: Action query time = 6.134 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4387 +t=10: Selected seed 195 with value = 0.4387 +Query 1/1: Action query time = 6.429 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4876 +t=26: Selected seed 195 with value = 0.4876 +Query 1/1: Action query time = 5.203 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5664 +t=42: Selected seed 195 with value = 0.5664 +Query 1/1: Action query time = 3.383 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6439 +t=58: Selected seed 195 with value = 0.6439 +Query 1/1: Action query time = 5.098 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7522 +t=74: Selected seed 195 with value = 0.7522 +Query 1/1: Action query time = 4.826 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9603 +t=90: Selected seed 195 with value = 0.9603 +Query 1/1: Action query time = 4.149 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9666 +t=106: Selected seed 195 with value = 0.9666 +Query 1/1: Action query time = 3.052 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9485 +t=122: Selected seed 195 with value = 0.9485 +Query 1/1: Action query time = 4.566 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.956 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.074 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=9--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 9 +# successes: 4 (44.4%) + +Task: open the middle drawer of the cabinet +Starting episode 10... +Query 1/1: Action query time = 5.721 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4447 +t=10: Selected seed 195 with value = 0.4447 +Query 1/1: Action query time = 5.002 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5124 +t=26: Selected seed 195 with value = 0.5124 +Query 1/1: Action query time = 4.140 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5549 +t=42: Selected seed 195 with value = 0.5549 +Query 1/1: Action query time = 6.599 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6709 +t=58: Selected seed 195 with value = 0.6709 +Query 1/1: Action query time = 5.991 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8079 +t=74: Selected seed 195 with value = 0.8079 +Query 1/1: Action query time = 5.899 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9168 +t=90: Selected seed 195 with value = 0.9168 +Query 1/1: Action query time = 4.882 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9610 +t=106: Selected seed 195 with value = 0.9610 +Query 1/1: Action query time = 4.770 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7662 +t=122: Selected seed 195 with value = 0.7662 +Query 1/1: Action query time = 4.507 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7732 +t=138: Selected seed 195 with value = 0.7732 +Query 1/1: Action query time = 4.637 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7848 +t=154: Selected seed 195 with value = 0.7848 +Query 1/1: Action query time = 5.311 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8063 +t=170: Selected seed 195 with value = 0.8063 +Query 1/1: Action query time = 5.362 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8464 +t=186: Selected seed 195 with value = 0.8464 +Query 1/1: Action query time = 5.949 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9022 +t=202: Selected seed 195 with value = 0.9022 +Query 1/1: Action query time = 4.938 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8544 +t=218: Selected seed 195 with value = 0.8544 +Query 1/1: Action query time = 4.846 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8874 +t=234: Selected seed 195 with value = 0.8874 +Query 1/1: Action query time = 4.362 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8862 +t=250: Selected seed 195 with value = 0.8862 +Query 1/1: Action query time = 4.374 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8669 +t=266: Selected seed 195 with value = 0.8669 +Query 1/1: Action query time = 4.454 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9421 +t=282: Selected seed 195 with value = 0.9421 +Query 1/1: Action query time = 5.105 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9289 +t=298: Selected seed 195 with value = 0.9289 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=10--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 10 +# successes: 4 (40.0%) + +Task: open the middle drawer of the cabinet +Starting episode 11... +Query 1/1: Action query time = 5.335 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4416 +t=10: Selected seed 195 with value = 0.4416 +Query 1/1: Action query time = 6.150 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5004 +t=26: Selected seed 195 with value = 0.5004 +Query 1/1: Action query time = 4.578 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5791 +t=42: Selected seed 195 with value = 0.5791 +Query 1/1: Action query time = 5.456 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6556 +t=58: Selected seed 195 with value = 0.6556 +Query 1/1: Action query time = 6.205 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7673 +t=74: Selected seed 195 with value = 0.7673 +Query 1/1: Action query time = 2.524 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9058 +t=90: Selected seed 195 with value = 0.9058 +Query 1/1: Action query time = 5.918 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8725 +t=106: Selected seed 195 with value = 0.8725 +Query 1/1: Action query time = 5.224 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8593 +t=122: Selected seed 195 with value = 0.8593 +Query 1/1: Action query time = 5.433 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9329 +t=138: Selected seed 195 with value = 0.9329 +Query 1/1: Action query time = 5.308 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9405 +t=154: Selected seed 195 with value = 0.9405 +Query 1/1: Action query time = 4.476 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9477 +t=170: Selected seed 195 with value = 0.9477 +Query 1/1: Action query time = 4.250 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9699 +t=186: Selected seed 195 with value = 0.9699 +Query 1/1: Action query time = 4.108 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9888 +t=202: Selected seed 195 with value = 0.9888 +Query 1/1: Action query time = 4.000 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=218: Selected seed 195 with value = 0.9912 +Query 1/1: Action query time = 3.656 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9910 +t=234: Selected seed 195 with value = 0.9910 +Query 1/1: Action query time = 4.037 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=250: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 3.645 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=266: Selected seed 195 with value = 0.9914 +Query 1/1: Action query time = 3.780 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9916 +t=282: Selected seed 195 with value = 0.9916 +Query 1/1: Action query time = 4.882 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=298: Selected seed 195 with value = 0.9917 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=11--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 11 +# successes: 4 (36.4%) + +Task: open the middle drawer of the cabinet +Starting episode 12... +Query 1/1: Action query time = 4.462 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4405 +t=10: Selected seed 195 with value = 0.4405 +Query 1/1: Action query time = 4.468 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4856 +t=26: Selected seed 195 with value = 0.4856 +Query 1/1: Action query time = 6.082 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5583 +t=42: Selected seed 195 with value = 0.5583 +Query 1/1: Action query time = 4.562 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6357 +t=58: Selected seed 195 with value = 0.6357 +Query 1/1: Action query time = 5.377 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7492 +t=74: Selected seed 195 with value = 0.7492 +Query 1/1: Action query time = 4.677 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9145 +t=90: Selected seed 195 with value = 0.9145 +Query 1/1: Action query time = 4.649 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9380 +t=106: Selected seed 195 with value = 0.9380 +Query 1/1: Action query time = 5.832 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9127 +t=122: Selected seed 195 with value = 0.9127 +Query 1/1: Action query time = 6.482 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9322 +t=138: Selected seed 195 with value = 0.9322 +Query 1/1: Action query time = 4.548 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9527 +t=154: Selected seed 195 with value = 0.9527 +Query 1/1: Action query time = 4.140 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9427 +t=170: Selected seed 195 with value = 0.9427 +Query 1/1: Action query time = 4.872 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9505 +t=186: Selected seed 195 with value = 0.9505 +Query 1/1: Action query time = 3.547 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9688 +t=202: Selected seed 195 with value = 0.9688 +Query 1/1: Action query time = 4.491 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9345 +t=218: Selected seed 195 with value = 0.9345 +Query 1/1: Action query time = 3.824 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9273 +t=234: Selected seed 195 with value = 0.9273 +Query 1/1: Action query time = 3.988 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9262 +t=250: Selected seed 195 with value = 0.9262 +Query 1/1: Action query time = 3.738 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9279 +t=266: Selected seed 195 with value = 0.9279 +Query 1/1: Action query time = 3.145 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9285 +t=282: Selected seed 195 with value = 0.9285 +Query 1/1: Action query time = 3.684 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9301 +t=298: Selected seed 195 with value = 0.9301 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=12--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 12 +# successes: 4 (33.3%) + +Task: open the middle drawer of the cabinet +Starting episode 13... +Query 1/1: Action query time = 5.103 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4498 +t=10: Selected seed 195 with value = 0.4498 +Query 1/1: Action query time = 4.336 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4987 +t=26: Selected seed 195 with value = 0.4987 +Query 1/1: Action query time = 4.003 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5743 +t=42: Selected seed 195 with value = 0.5743 +Query 1/1: Action query time = 5.098 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6585 +t=58: Selected seed 195 with value = 0.6585 +Query 1/1: Action query time = 4.845 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7927 +t=74: Selected seed 195 with value = 0.7927 +Query 1/1: Action query time = 3.275 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9678 +t=90: Selected seed 195 with value = 0.9678 +Query 1/1: Action query time = 4.850 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8804 +t=106: Selected seed 195 with value = 0.8804 +Query 1/1: Action query time = 4.144 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9383 +t=122: Selected seed 195 with value = 0.9383 +Query 1/1: Action query time = 4.345 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9480 +t=138: Selected seed 195 with value = 0.9480 +Query 1/1: Action query time = 4.790 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9485 +t=154: Selected seed 195 with value = 0.9485 +Query 1/1: Action query time = 3.707 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9460 +t=170: Selected seed 195 with value = 0.9460 +Query 1/1: Action query time = 3.784 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9491 +t=186: Selected seed 195 with value = 0.9491 +Query 1/1: Action query time = 3.786 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9508 +t=202: Selected seed 195 with value = 0.9508 +Query 1/1: Action query time = 4.156 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9459 +t=218: Selected seed 195 with value = 0.9459 +Query 1/1: Action query time = 3.482 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9393 +t=234: Selected seed 195 with value = 0.9393 +Query 1/1: Action query time = 3.142 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9325 +t=250: Selected seed 195 with value = 0.9325 +Query 1/1: Action query time = 1.995 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9202 +t=266: Selected seed 195 with value = 0.9202 +Query 1/1: Action query time = 1.839 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9054 +t=282: Selected seed 195 with value = 0.9054 +Query 1/1: Action query time = 3.383 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8756 +t=298: Selected seed 195 with value = 0.8756 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t0_s1/2026_08_03-01_16_45--with_future_img--episode=13--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 13 +# successes: 4 (30.8%) +Current task success rate: 0.3076923076923077 +Current total success rate: 0.3076923076923077 +Final results: +Total episodes: 13 +Total successes: 4 +Overall success rate: 0.3077 (30.8%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_46--t7dc800x50_t1_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_46--t7dc800x50_t1_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..a8344d874a8ff4a6e65ffacb09d1f4d4acc7c105 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_46--t7dc800x50_t1_s1.txt @@ -0,0 +1,883 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_demochan800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7dc800x50_t1_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 4.450 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5181 +t=10: Selected seed 195 with value = 0.5181 +Query 1/1: Action query time = 4.061 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6237 +t=26: Selected seed 195 with value = 0.6237 +Query 1/1: Action query time = 5.145 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7248 +t=42: Selected seed 195 with value = 0.7248 +Query 1/1: Action query time = 3.913 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9675 +t=58: Selected seed 195 with value = 0.9675 +Query 1/1: Action query time = 7.876 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9544 +t=74: Selected seed 195 with value = 0.9544 +Query 1/1: Action query time = 5.262 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9810 +t=90: Selected seed 195 with value = 0.9810 +Query 1/1: Action query time = 5.533 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9852 +t=106: Selected seed 195 with value = 0.9852 +Query 1/1: Action query time = 5.811 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9791 +t=122: Selected seed 195 with value = 0.9791 +Query 1/1: Action query time = 5.112 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9676 +t=138: Selected seed 195 with value = 0.9676 +Query 1/1: Action query time = 6.520 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9465 +t=154: Selected seed 195 with value = 0.9465 +Query 1/1: Action query time = 7.221 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9853 +t=170: Selected seed 195 with value = 0.9853 +Query 1/1: Action query time = 7.093 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9657 +t=186: Selected seed 195 with value = 0.9657 +Query 1/1: Action query time = 5.695 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9634 +t=202: Selected seed 195 with value = 0.9634 +Query 1/1: Action query time = 4.877 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9548 +t=218: Selected seed 195 with value = 0.9548 +Query 1/1: Action query time = 6.343 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9474 +t=234: Selected seed 195 with value = 0.9474 +Query 1/1: Action query time = 5.870 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9446 +t=250: Selected seed 195 with value = 0.9446 +Query 1/1: Action query time = 6.129 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9447 +t=266: Selected seed 195 with value = 0.9447 +Query 1/1: Action query time = 4.279 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9470 +t=282: Selected seed 195 with value = 0.9470 +Query 1/1: Action query time = 5.583 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9480 +t=298: Selected seed 195 with value = 0.9480 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 3.341 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5084 +t=10: Selected seed 195 with value = 0.5084 +Query 1/1: Action query time = 4.563 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5809 +t=26: Selected seed 195 with value = 0.5809 +Query 1/1: Action query time = 4.993 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6741 +t=42: Selected seed 195 with value = 0.6741 +Query 1/1: Action query time = 6.268 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8648 +t=58: Selected seed 195 with value = 0.8648 +Query 1/1: Action query time = 7.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9245 +t=74: Selected seed 195 with value = 0.9245 +Query 1/1: Action query time = 4.102 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=90: Selected seed 195 with value = 0.9961 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 6.035 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5200 +t=10: Selected seed 195 with value = 0.5200 +Query 1/1: Action query time = 6.804 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5757 +t=26: Selected seed 195 with value = 0.5757 +Query 1/1: Action query time = 4.796 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6900 +t=42: Selected seed 195 with value = 0.6900 +Query 1/1: Action query time = 6.818 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8149 +t=58: Selected seed 195 with value = 0.8149 +Query 1/1: Action query time = 5.296 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9679 +t=74: Selected seed 195 with value = 0.9679 +Query 1/1: Action query time = 7.157 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9837 +t=90: Selected seed 195 with value = 0.9837 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 4.604 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5160 +t=10: Selected seed 195 with value = 0.5160 +Query 1/1: Action query time = 6.243 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6126 +t=26: Selected seed 195 with value = 0.6126 +Query 1/1: Action query time = 6.796 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7082 +t=42: Selected seed 195 with value = 0.7082 +Query 1/1: Action query time = 3.930 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9550 +t=58: Selected seed 195 with value = 0.9550 +Query 1/1: Action query time = 4.489 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9578 +t=74: Selected seed 195 with value = 0.9578 +Query 1/1: Action query time = 3.842 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.158 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=106: Selected seed 195 with value = 0.9983 +Query 1/1: Action query time = 3.372 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9252 +t=122: Selected seed 195 with value = 0.9252 +Query 1/1: Action query time = 3.862 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=138: Selected seed 195 with value = 0.9871 +Query 1/1: Action query time = 4.065 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9477 +t=154: Selected seed 195 with value = 0.9477 +Query 1/1: Action query time = 4.781 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9441 +t=170: Selected seed 195 with value = 0.9441 +Query 1/1: Action query time = 4.314 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9452 +t=186: Selected seed 195 with value = 0.9452 +Query 1/1: Action query time = 6.777 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9480 +t=202: Selected seed 195 with value = 0.9480 +Query 1/1: Action query time = 5.294 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9592 +t=218: Selected seed 195 with value = 0.9592 +Query 1/1: Action query time = 5.450 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9639 +t=234: Selected seed 195 with value = 0.9639 +Query 1/1: Action query time = 6.463 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9689 +t=250: Selected seed 195 with value = 0.9689 +Query 1/1: Action query time = 5.192 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9623 +t=266: Selected seed 195 with value = 0.9623 +Query 1/1: Action query time = 6.645 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9584 +t=282: Selected seed 195 with value = 0.9584 +Query 1/1: Action query time = 5.875 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9574 +t=298: Selected seed 195 with value = 0.9574 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 4 +# successes: 2 (50.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 6.621 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5174 +t=10: Selected seed 195 with value = 0.5174 +Query 1/1: Action query time = 4.756 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6354 +t=26: Selected seed 195 with value = 0.6354 +Query 1/1: Action query time = 5.222 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7856 +t=42: Selected seed 195 with value = 0.7856 +Query 1/1: Action query time = 5.502 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9891 +t=58: Selected seed 195 with value = 0.9891 +Query 1/1: Action query time = 6.501 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=74: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 4.202 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=90: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 4.635 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=106: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 4.596 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=122: Selected seed 195 with value = 0.9968 +Query 1/1: Action query time = 4.345 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=138: Selected seed 195 with value = 0.9946 +Query 1/1: Action query time = 5.775 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=154: Selected seed 195 with value = 0.9886 +Query 1/1: Action query time = 6.567 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9855 +t=170: Selected seed 195 with value = 0.9855 +Query 1/1: Action query time = 7.061 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9885 +t=186: Selected seed 195 with value = 0.9885 +Query 1/1: Action query time = 6.102 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9923 +t=202: Selected seed 195 with value = 0.9923 +Query 1/1: Action query time = 4.113 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9943 +t=218: Selected seed 195 with value = 0.9943 +Query 1/1: Action query time = 5.792 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=234: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 5.559 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.902 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=266: Selected seed 195 with value = 0.9955 +Query 1/1: Action query time = 5.063 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.720 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9905 +t=298: Selected seed 195 with value = 0.9905 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=5--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 5 +# successes: 2 (40.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 6.740 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5095 +t=10: Selected seed 195 with value = 0.5095 +Query 1/1: Action query time = 6.244 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5964 +t=26: Selected seed 195 with value = 0.5964 +Query 1/1: Action query time = 5.049 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6756 +t=42: Selected seed 195 with value = 0.6756 +Query 1/1: Action query time = 5.261 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8763 +t=58: Selected seed 195 with value = 0.8763 +Query 1/1: Action query time = 4.515 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9538 +t=74: Selected seed 195 with value = 0.9538 +Query 1/1: Action query time = 5.697 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=90: Selected seed 195 with value = 0.9960 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=6--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 3 (50.0%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 5.398 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5248 +t=10: Selected seed 195 with value = 0.5248 +Query 1/1: Action query time = 5.924 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6071 +t=26: Selected seed 195 with value = 0.6071 +Query 1/1: Action query time = 6.109 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7285 +t=42: Selected seed 195 with value = 0.7285 +Query 1/1: Action query time = 4.784 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8981 +t=58: Selected seed 195 with value = 0.8981 +Query 1/1: Action query time = 6.656 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=74: Selected seed 195 with value = 0.9909 +Query 1/1: Action query time = 5.202 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9595 +t=90: Selected seed 195 with value = 0.9595 +Query 1/1: Action query time = 5.631 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9617 +t=106: Selected seed 195 with value = 0.9617 +Query 1/1: Action query time = 4.138 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9650 +t=122: Selected seed 195 with value = 0.9650 +Query 1/1: Action query time = 6.429 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9671 +t=138: Selected seed 195 with value = 0.9671 +Query 1/1: Action query time = 5.698 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9539 +t=154: Selected seed 195 with value = 0.9539 +Query 1/1: Action query time = 5.858 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=170: Selected seed 195 with value = 0.9871 +Query 1/1: Action query time = 5.196 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9807 +t=186: Selected seed 195 with value = 0.9807 +Query 1/1: Action query time = 6.266 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9508 +t=202: Selected seed 195 with value = 0.9508 +Query 1/1: Action query time = 6.059 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9105 +t=218: Selected seed 195 with value = 0.9105 +Query 1/1: Action query time = 6.688 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9322 +t=234: Selected seed 195 with value = 0.9322 +Query 1/1: Action query time = 5.458 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9252 +t=250: Selected seed 195 with value = 0.9252 +Query 1/1: Action query time = 5.366 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9269 +t=266: Selected seed 195 with value = 0.9269 +Query 1/1: Action query time = 5.214 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9383 +t=282: Selected seed 195 with value = 0.9383 +Query 1/1: Action query time = 5.020 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9533 +t=298: Selected seed 195 with value = 0.9533 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=7--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 7 +# successes: 3 (42.9%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 3.758 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5594 +t=10: Selected seed 195 with value = 0.5594 +Query 1/1: Action query time = 4.589 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6163 +t=26: Selected seed 195 with value = 0.6163 +Query 1/1: Action query time = 5.064 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6691 +t=42: Selected seed 195 with value = 0.6691 +Query 1/1: Action query time = 5.067 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8358 +t=58: Selected seed 195 with value = 0.8358 +Query 1/1: Action query time = 5.849 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8909 +t=74: Selected seed 195 with value = 0.8909 +Query 1/1: Action query time = 4.386 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9583 +t=90: Selected seed 195 with value = 0.9583 +Query 1/1: Action query time = 5.854 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9458 +t=106: Selected seed 195 with value = 0.9458 +Query 1/1: Action query time = 5.849 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9443 +t=122: Selected seed 195 with value = 0.9443 +Query 1/1: Action query time = 7.833 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9257 +t=138: Selected seed 195 with value = 0.9257 +Query 1/1: Action query time = 5.625 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9020 +t=154: Selected seed 195 with value = 0.9020 +Query 1/1: Action query time = 4.539 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8878 +t=170: Selected seed 195 with value = 0.8878 +Query 1/1: Action query time = 6.025 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8950 +t=186: Selected seed 195 with value = 0.8950 +Query 1/1: Action query time = 5.586 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8889 +t=202: Selected seed 195 with value = 0.8889 +Query 1/1: Action query time = 6.157 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8907 +t=218: Selected seed 195 with value = 0.8907 +Query 1/1: Action query time = 6.615 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8903 +t=234: Selected seed 195 with value = 0.8903 +Query 1/1: Action query time = 5.291 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9025 +t=250: Selected seed 195 with value = 0.9025 +Query 1/1: Action query time = 5.542 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9075 +t=266: Selected seed 195 with value = 0.9075 +Query 1/1: Action query time = 5.185 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8842 +t=282: Selected seed 195 with value = 0.8842 +Query 1/1: Action query time = 3.768 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8395 +t=298: Selected seed 195 with value = 0.8395 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=8--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 8 +# successes: 3 (37.5%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 5.145 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5080 +t=10: Selected seed 195 with value = 0.5080 +Query 1/1: Action query time = 6.076 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5579 +t=26: Selected seed 195 with value = 0.5579 +Query 1/1: Action query time = 7.324 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6847 +t=42: Selected seed 195 with value = 0.6847 +Query 1/1: Action query time = 4.213 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8028 +t=58: Selected seed 195 with value = 0.8028 +Query 1/1: Action query time = 6.381 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9026 +t=74: Selected seed 195 with value = 0.9026 +Query 1/1: Action query time = 6.213 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9128 +t=90: Selected seed 195 with value = 0.9128 +Query 1/1: Action query time = 5.112 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9237 +t=106: Selected seed 195 with value = 0.9237 +Query 1/1: Action query time = 5.541 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9875 +t=122: Selected seed 195 with value = 0.9875 +Query 1/1: Action query time = 6.254 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=138: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 6.702 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9907 +t=154: Selected seed 195 with value = 0.9907 +Query 1/1: Action query time = 4.577 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9274 +t=170: Selected seed 195 with value = 0.9274 +Query 1/1: Action query time = 6.345 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9053 +t=186: Selected seed 195 with value = 0.9053 +Query 1/1: Action query time = 6.515 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9473 +t=202: Selected seed 195 with value = 0.9473 +Query 1/1: Action query time = 5.591 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9408 +t=218: Selected seed 195 with value = 0.9408 +Query 1/1: Action query time = 6.970 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9428 +t=234: Selected seed 195 with value = 0.9428 +Query 1/1: Action query time = 6.389 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9451 +t=250: Selected seed 195 with value = 0.9451 +Query 1/1: Action query time = 5.568 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9519 +t=266: Selected seed 195 with value = 0.9519 +Query 1/1: Action query time = 4.693 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9511 +t=282: Selected seed 195 with value = 0.9511 +Query 1/1: Action query time = 7.123 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9451 +t=298: Selected seed 195 with value = 0.9451 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 9 +# successes: 3 (33.3%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 1.601 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5273 +t=10: Selected seed 195 with value = 0.5273 +Query 1/1: Action query time = 5.686 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6147 +t=26: Selected seed 195 with value = 0.6147 +Query 1/1: Action query time = 4.636 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7242 +t=42: Selected seed 195 with value = 0.7242 +Query 1/1: Action query time = 5.934 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8417 +t=58: Selected seed 195 with value = 0.8417 +Query 1/1: Action query time = 6.739 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9933 +t=74: Selected seed 195 with value = 0.9933 +Query 1/1: Action query time = 5.915 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9853 +t=90: Selected seed 195 with value = 0.9853 +Query 1/1: Action query time = 6.346 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9778 +t=106: Selected seed 195 with value = 0.9778 +Query 1/1: Action query time = 5.881 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9860 +t=122: Selected seed 195 with value = 0.9860 +Query 1/1: Action query time = 5.340 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9891 +t=138: Selected seed 195 with value = 0.9891 +Query 1/1: Action query time = 5.326 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9617 +t=154: Selected seed 195 with value = 0.9617 +Query 1/1: Action query time = 6.113 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9646 +t=170: Selected seed 195 with value = 0.9646 +Query 1/1: Action query time = 6.593 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9419 +t=186: Selected seed 195 with value = 0.9419 +Query 1/1: Action query time = 6.689 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9658 +t=202: Selected seed 195 with value = 0.9658 +Query 1/1: Action query time = 7.091 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9585 +t=218: Selected seed 195 with value = 0.9585 +Query 1/1: Action query time = 4.556 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9560 +t=234: Selected seed 195 with value = 0.9560 +Query 1/1: Action query time = 4.650 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9494 +t=250: Selected seed 195 with value = 0.9494 +Query 1/1: Action query time = 4.606 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9439 +t=266: Selected seed 195 with value = 0.9439 +Query 1/1: Action query time = 5.148 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9430 +t=282: Selected seed 195 with value = 0.9430 +Query 1/1: Action query time = 4.700 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9456 +t=298: Selected seed 195 with value = 0.9456 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=10--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 10 +# successes: 3 (30.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 4.332 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5681 +t=10: Selected seed 195 with value = 0.5681 +Query 1/1: Action query time = 3.262 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6689 +t=26: Selected seed 195 with value = 0.6689 +Query 1/1: Action query time = 4.321 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7758 +t=42: Selected seed 195 with value = 0.7758 +Query 1/1: Action query time = 6.675 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9592 +t=58: Selected seed 195 with value = 0.9592 +Query 1/1: Action query time = 7.360 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9884 +t=74: Selected seed 195 with value = 0.9884 +Query 1/1: Action query time = 8.072 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9608 +t=90: Selected seed 195 with value = 0.9608 +Query 1/1: Action query time = 5.575 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9649 +t=106: Selected seed 195 with value = 0.9649 +Query 1/1: Action query time = 7.872 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9813 +t=122: Selected seed 195 with value = 0.9813 +Query 1/1: Action query time = 7.567 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9939 +t=138: Selected seed 195 with value = 0.9939 +Query 1/1: Action query time = 7.102 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=154: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 7.036 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9943 +t=170: Selected seed 195 with value = 0.9943 +Query 1/1: Action query time = 5.692 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9934 +t=186: Selected seed 195 with value = 0.9934 +Query 1/1: Action query time = 4.960 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=202: Selected seed 195 with value = 0.9912 +Query 1/1: Action query time = 6.408 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9887 +t=218: Selected seed 195 with value = 0.9887 +Query 1/1: Action query time = 4.770 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9875 +t=234: Selected seed 195 with value = 0.9875 +Query 1/1: Action query time = 5.607 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9848 +t=250: Selected seed 195 with value = 0.9848 +Query 1/1: Action query time = 4.900 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9814 +t=266: Selected seed 195 with value = 0.9814 +Query 1/1: Action query time = 4.960 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9797 +t=282: Selected seed 195 with value = 0.9797 +Query 1/1: Action query time = 4.937 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9784 +t=298: Selected seed 195 with value = 0.9784 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=11--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 11 +# successes: 3 (27.3%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 3.714 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4949 +t=10: Selected seed 195 with value = 0.4949 +Query 1/1: Action query time = 4.494 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5863 +t=26: Selected seed 195 with value = 0.5863 +Query 1/1: Action query time = 4.792 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6849 +t=42: Selected seed 195 with value = 0.6849 +Query 1/1: Action query time = 6.444 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8472 +t=58: Selected seed 195 with value = 0.8472 +Query 1/1: Action query time = 4.620 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9471 +t=74: Selected seed 195 with value = 0.9471 +Query 1/1: Action query time = 5.783 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9939 +t=90: Selected seed 195 with value = 0.9939 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=12--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 4 (33.3%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 4.805 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5154 +t=10: Selected seed 195 with value = 0.5154 +Query 1/1: Action query time = 4.616 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5930 +t=26: Selected seed 195 with value = 0.5930 +Query 1/1: Action query time = 4.860 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6917 +t=42: Selected seed 195 with value = 0.6917 +Query 1/1: Action query time = 4.340 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8626 +t=58: Selected seed 195 with value = 0.8626 +Query 1/1: Action query time = 4.209 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9775 +t=74: Selected seed 195 with value = 0.9775 +Query 1/1: Action query time = 5.074 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9547 +t=90: Selected seed 195 with value = 0.9547 +Query 1/1: Action query time = 4.532 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9604 +t=106: Selected seed 195 with value = 0.9604 +Query 1/1: Action query time = 3.842 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9550 +t=122: Selected seed 195 with value = 0.9550 +Query 1/1: Action query time = 4.045 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9539 +t=138: Selected seed 195 with value = 0.9539 +Query 1/1: Action query time = 4.446 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9558 +t=154: Selected seed 195 with value = 0.9558 +Query 1/1: Action query time = 5.047 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9487 +t=170: Selected seed 195 with value = 0.9487 +Query 1/1: Action query time = 4.228 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9099 +t=186: Selected seed 195 with value = 0.9099 +Query 1/1: Action query time = 4.222 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8967 +t=202: Selected seed 195 with value = 0.8967 +Query 1/1: Action query time = 2.618 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9035 +t=218: Selected seed 195 with value = 0.9035 +Query 1/1: Action query time = 4.276 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9091 +t=234: Selected seed 195 with value = 0.9091 +Query 1/1: Action query time = 2.800 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9118 +t=250: Selected seed 195 with value = 0.9118 +Query 1/1: Action query time = 4.136 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9161 +t=266: Selected seed 195 with value = 0.9161 +Query 1/1: Action query time = 3.256 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9183 +t=282: Selected seed 195 with value = 0.9183 +Query 1/1: Action query time = 3.863 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9186 +t=298: Selected seed 195 with value = 0.9186 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=13--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 13 +# successes: 4 (30.8%) +Current task success rate: 0.3076923076923077 +Current total success rate: 0.3076923076923077 +Final results: +Total episodes: 13 +Total successes: 4 +Overall success rate: 0.3077 (30.8%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_46--t7rc800x50_t1_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_46--t7rc800x50_t1_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..eda1c64a01eaed3d235b7bc226b7bb39460dfcf1 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_46--t7rc800x50_t1_s0.txt @@ -0,0 +1,747 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t1_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,4,8,12,16,20,24,28,32,36,40,44,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 4.180 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5221 +t=10: Selected seed 195 with value = 0.5221 +Query 1/1: Action query time = 3.765 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5910 +t=26: Selected seed 195 with value = 0.5910 +Query 1/1: Action query time = 6.790 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7002 +t=42: Selected seed 195 with value = 0.7002 +Query 1/1: Action query time = 6.005 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8048 +t=58: Selected seed 195 with value = 0.8048 +Query 1/1: Action query time = 6.707 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9644 +t=74: Selected seed 195 with value = 0.9644 +Query 1/1: Action query time = 6.020 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9794 +t=90: Selected seed 195 with value = 0.9794 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 5.823 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5134 +t=10: Selected seed 195 with value = 0.5134 +Query 1/1: Action query time = 6.024 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5889 +t=26: Selected seed 195 with value = 0.5889 +Query 1/1: Action query time = 6.400 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7365 +t=42: Selected seed 195 with value = 0.7365 +Query 1/1: Action query time = 5.981 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9601 +t=58: Selected seed 195 with value = 0.9601 +Query 1/1: Action query time = 6.176 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9674 +t=74: Selected seed 195 with value = 0.9674 +Query 1/1: Action query time = 6.197 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.424 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9789 +t=106: Selected seed 195 with value = 0.9789 +Query 1/1: Action query time = 7.981 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9759 +t=122: Selected seed 195 with value = 0.9759 +Query 1/1: Action query time = 5.461 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9897 +t=138: Selected seed 195 with value = 0.9897 +Query 1/1: Action query time = 4.195 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9887 +t=154: Selected seed 195 with value = 0.9887 +Query 1/1: Action query time = 4.768 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=170: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 3.730 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.739 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9953 +t=202: Selected seed 195 with value = 0.9953 +Query 1/1: Action query time = 4.933 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9436 +t=218: Selected seed 195 with value = 0.9436 +Query 1/1: Action query time = 4.627 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.590 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9505 +t=250: Selected seed 195 with value = 0.9505 +Query 1/1: Action query time = 5.462 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9530 +t=266: Selected seed 195 with value = 0.9530 +Query 1/1: Action query time = 4.448 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9675 +t=282: Selected seed 195 with value = 0.9675 +Query 1/1: Action query time = 5.472 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9699 +t=298: Selected seed 195 with value = 0.9699 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 6.045 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4594 +t=10: Selected seed 195 with value = 0.4594 +Query 1/1: Action query time = 6.476 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5961 +t=26: Selected seed 195 with value = 0.5961 +Query 1/1: Action query time = 5.944 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6569 +t=42: Selected seed 195 with value = 0.6569 +Query 1/1: Action query time = 5.880 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7819 +t=58: Selected seed 195 with value = 0.7819 +Query 1/1: Action query time = 6.870 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9617 +t=74: Selected seed 195 with value = 0.9617 +Query 1/1: Action query time = 5.686 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7474 +t=90: Selected seed 195 with value = 0.7474 +Query 1/1: Action query time = 7.809 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8682 +t=106: Selected seed 195 with value = 0.8682 +Query 1/1: Action query time = 7.417 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9281 +t=122: Selected seed 195 with value = 0.9281 +Query 1/1: Action query time = 5.573 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8954 +t=138: Selected seed 195 with value = 0.8954 +Query 1/1: Action query time = 4.129 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9738 +t=154: Selected seed 195 with value = 0.9738 +Query 1/1: Action query time = 5.264 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9220 +t=170: Selected seed 195 with value = 0.9220 +Query 1/1: Action query time = 5.433 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9271 +t=186: Selected seed 195 with value = 0.9271 +Query 1/1: Action query time = 2.989 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9417 +t=202: Selected seed 195 with value = 0.9417 +Query 1/1: Action query time = 4.115 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9402 +t=218: Selected seed 195 with value = 0.9402 +Query 1/1: Action query time = 5.105 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8912 +t=234: Selected seed 195 with value = 0.8912 +Query 1/1: Action query time = 4.288 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9475 +t=250: Selected seed 195 with value = 0.9475 +Query 1/1: Action query time = 4.948 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9385 +t=266: Selected seed 195 with value = 0.9385 +Query 1/1: Action query time = 5.831 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9421 +t=282: Selected seed 195 with value = 0.9421 +Query 1/1: Action query time = 7.021 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9367 +t=298: Selected seed 195 with value = 0.9367 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 5.822 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5206 +t=10: Selected seed 195 with value = 0.5206 +Query 1/1: Action query time = 5.289 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6141 +t=26: Selected seed 195 with value = 0.6141 +Query 1/1: Action query time = 4.512 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6976 +t=42: Selected seed 195 with value = 0.6976 +Query 1/1: Action query time = 6.023 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8905 +t=58: Selected seed 195 with value = 0.8905 +Query 1/1: Action query time = 6.213 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9484 +t=74: Selected seed 195 with value = 0.9484 +Query 1/1: Action query time = 5.468 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=90: Selected seed 195 with value = 0.9809 +Query 1/1: Action query time = 5.394 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9772 +t=106: Selected seed 195 with value = 0.9772 +Query 1/1: Action query time = 6.123 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9333 +t=122: Selected seed 195 with value = 0.9333 +Query 1/1: Action query time = 6.894 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9738 +t=138: Selected seed 195 with value = 0.9738 +Query 1/1: Action query time = 5.924 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9834 +t=154: Selected seed 195 with value = 0.9834 +Query 1/1: Action query time = 5.881 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9013 +t=170: Selected seed 195 with value = 0.9013 +Query 1/1: Action query time = 5.330 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9328 +t=186: Selected seed 195 with value = 0.9328 +Query 1/1: Action query time = 4.835 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8299 +t=202: Selected seed 195 with value = 0.8299 +Query 1/1: Action query time = 5.966 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8071 +t=218: Selected seed 195 with value = 0.8071 +Query 1/1: Action query time = 5.560 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8025 +t=234: Selected seed 195 with value = 0.8025 +Query 1/1: Action query time = 4.201 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7981 +t=250: Selected seed 195 with value = 0.7981 +Query 1/1: Action query time = 6.017 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8775 +t=266: Selected seed 195 with value = 0.8775 +Query 1/1: Action query time = 7.609 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8181 +t=282: Selected seed 195 with value = 0.8181 +Query 1/1: Action query time = 7.697 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8333 +t=298: Selected seed 195 with value = 0.8333 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 4 +# successes: 1 (25.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 3.972 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5248 +t=10: Selected seed 195 with value = 0.5248 +Query 1/1: Action query time = 5.848 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5881 +t=26: Selected seed 195 with value = 0.5881 +Query 1/1: Action query time = 6.293 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7166 +t=42: Selected seed 195 with value = 0.7166 +Query 1/1: Action query time = 4.671 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8605 +t=58: Selected seed 195 with value = 0.8605 +Query 1/1: Action query time = 3.597 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9738 +t=74: Selected seed 195 with value = 0.9738 +Query 1/1: Action query time = 5.777 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9927 +t=90: Selected seed 195 with value = 0.9927 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=5--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 2 (40.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 5.849 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4848 +t=10: Selected seed 195 with value = 0.4848 +Query 1/1: Action query time = 5.005 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5850 +t=26: Selected seed 195 with value = 0.5850 +Query 1/1: Action query time = 3.755 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6871 +t=42: Selected seed 195 with value = 0.6871 +Query 1/1: Action query time = 4.773 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7703 +t=58: Selected seed 195 with value = 0.7703 +Query 1/1: Action query time = 7.522 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8753 +t=74: Selected seed 195 with value = 0.8753 +Query 1/1: Action query time = 4.648 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9921 +t=90: Selected seed 195 with value = 0.9921 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=6--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 3 (50.0%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 5.692 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5243 +t=10: Selected seed 195 with value = 0.5243 +Query 1/1: Action query time = 3.896 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6201 +t=26: Selected seed 195 with value = 0.6201 +Query 1/1: Action query time = 5.123 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7320 +t=42: Selected seed 195 with value = 0.7320 +Query 1/1: Action query time = 5.423 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9675 +t=58: Selected seed 195 with value = 0.9675 +Query 1/1: Action query time = 6.015 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9503 +t=74: Selected seed 195 with value = 0.9503 +Query 1/1: Action query time = 6.002 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.041 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8095 +t=106: Selected seed 195 with value = 0.8095 +Query 1/1: Action query time = 5.661 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9408 +t=122: Selected seed 195 with value = 0.9408 +Query 1/1: Action query time = 6.416 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8222 +t=138: Selected seed 195 with value = 0.8222 +Query 1/1: Action query time = 6.601 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9155 +t=154: Selected seed 195 with value = 0.9155 +Query 1/1: Action query time = 5.510 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9338 +t=170: Selected seed 195 with value = 0.9338 +Query 1/1: Action query time = 3.594 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9540 +t=186: Selected seed 195 with value = 0.9540 +Query 1/1: Action query time = 6.738 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9715 +t=202: Selected seed 195 with value = 0.9715 +Query 1/1: Action query time = 5.647 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9810 +t=218: Selected seed 195 with value = 0.9810 +Query 1/1: Action query time = 5.330 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9925 +t=234: Selected seed 195 with value = 0.9925 +Query 1/1: Action query time = 5.107 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=250: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 4.910 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.708 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=282: Selected seed 195 with value = 0.9971 +Query 1/1: Action query time = 4.396 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9858 +t=298: Selected seed 195 with value = 0.9858 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=7--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 7 +# successes: 3 (42.9%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 4.529 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5066 +t=10: Selected seed 195 with value = 0.5066 +Query 1/1: Action query time = 5.762 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6107 +t=26: Selected seed 195 with value = 0.6107 +Query 1/1: Action query time = 5.737 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6966 +t=42: Selected seed 195 with value = 0.6966 +Query 1/1: Action query time = 6.124 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7762 +t=58: Selected seed 195 with value = 0.7762 +Query 1/1: Action query time = 7.770 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8801 +t=74: Selected seed 195 with value = 0.8801 +Query 1/1: Action query time = 6.105 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9778 +t=90: Selected seed 195 with value = 0.9778 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=8--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 4 (50.0%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 5.388 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5257 +t=10: Selected seed 195 with value = 0.5257 +Query 1/1: Action query time = 5.416 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6144 +t=26: Selected seed 195 with value = 0.6144 +Query 1/1: Action query time = 6.112 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6994 +t=42: Selected seed 195 with value = 0.6994 +Query 1/1: Action query time = 5.985 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8449 +t=58: Selected seed 195 with value = 0.8449 +Query 1/1: Action query time = 5.308 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9634 +t=74: Selected seed 195 with value = 0.9634 +Query 1/1: Action query time = 6.935 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9571 +t=90: Selected seed 195 with value = 0.9571 +Query 1/1: Action query time = 5.174 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9972 +t=106: Selected seed 195 with value = 0.9972 +Query 1/1: Action query time = 7.038 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8229 +t=122: Selected seed 195 with value = 0.8229 +Query 1/1: Action query time = 6.289 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9412 +t=138: Selected seed 195 with value = 0.9412 +Query 1/1: Action query time = 6.318 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8855 +t=154: Selected seed 195 with value = 0.8855 +Query 1/1: Action query time = 4.210 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9817 +t=170: Selected seed 195 with value = 0.9817 +Query 1/1: Action query time = 4.968 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8490 +t=186: Selected seed 195 with value = 0.8490 +Query 1/1: Action query time = 3.797 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9637 +t=202: Selected seed 195 with value = 0.9637 +Query 1/1: Action query time = 5.433 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9622 +t=218: Selected seed 195 with value = 0.9622 +Query 1/1: Action query time = 4.630 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9784 +t=234: Selected seed 195 with value = 0.9784 +Query 1/1: Action query time = 5.165 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9526 +t=250: Selected seed 195 with value = 0.9526 +Query 1/1: Action query time = 4.982 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=266: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 5.764 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9950 +t=282: Selected seed 195 with value = 0.9950 +Query 1/1: Action query time = 6.279 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9785 +t=298: Selected seed 195 with value = 0.9785 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 9 +# successes: 4 (44.4%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 5.229 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5472 +t=10: Selected seed 195 with value = 0.5472 +Query 1/1: Action query time = 6.144 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6148 +t=26: Selected seed 195 with value = 0.6148 +Query 1/1: Action query time = 3.995 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7485 +t=42: Selected seed 195 with value = 0.7485 +Query 1/1: Action query time = 5.725 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8058 +t=58: Selected seed 195 with value = 0.8058 +Query 1/1: Action query time = 4.646 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9365 +t=74: Selected seed 195 with value = 0.9365 +Query 1/1: Action query time = 5.252 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9784 +t=90: Selected seed 195 with value = 0.9784 +Query 1/1: Action query time = 6.374 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=10--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 5 (50.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 4.043 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5352 +t=10: Selected seed 195 with value = 0.5352 +Query 1/1: Action query time = 4.915 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6009 +t=26: Selected seed 195 with value = 0.6009 +Query 1/1: Action query time = 4.973 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6783 +t=42: Selected seed 195 with value = 0.6783 +Query 1/1: Action query time = 5.179 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8190 +t=58: Selected seed 195 with value = 0.8190 +Query 1/1: Action query time = 5.181 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9478 +t=74: Selected seed 195 with value = 0.9478 +Query 1/1: Action query time = 5.441 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9063 +t=90: Selected seed 195 with value = 0.9063 +Query 1/1: Action query time = 5.644 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9133 +t=106: Selected seed 195 with value = 0.9133 +Query 1/1: Action query time = 5.175 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9286 +t=122: Selected seed 195 with value = 0.9286 +Query 1/1: Action query time = 7.614 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9604 +t=138: Selected seed 195 with value = 0.9604 +Query 1/1: Action query time = 6.509 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=154: Selected seed 195 with value = 0.9909 +Query 1/1: Action query time = 6.622 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9073 +t=170: Selected seed 195 with value = 0.9073 +Query 1/1: Action query time = 6.913 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9654 +t=186: Selected seed 195 with value = 0.9654 +Query 1/1: Action query time = 7.574 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9464 +t=202: Selected seed 195 with value = 0.9464 +Query 1/1: Action query time = 7.275 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9614 +t=218: Selected seed 195 with value = 0.9614 +Query 1/1: Action query time = 7.409 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9445 +t=234: Selected seed 195 with value = 0.9445 +Query 1/1: Action query time = 5.302 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9189 +t=250: Selected seed 195 with value = 0.9189 +Query 1/1: Action query time = 6.434 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8910 +t=266: Selected seed 195 with value = 0.8910 +Query 1/1: Action query time = 4.430 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9028 +t=282: Selected seed 195 with value = 0.9028 +Query 1/1: Action query time = 4.594 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9297 +t=298: Selected seed 195 with value = 0.9297 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=11--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 11 +# successes: 5 (45.5%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 5.569 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4759 +t=10: Selected seed 195 with value = 0.4759 +Query 1/1: Action query time = 5.711 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5825 +t=26: Selected seed 195 with value = 0.5825 +Query 1/1: Action query time = 3.710 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6611 +t=42: Selected seed 195 with value = 0.6611 +Query 1/1: Action query time = 3.935 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7430 +t=58: Selected seed 195 with value = 0.7430 +Query 1/1: Action query time = 4.467 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9619 +t=74: Selected seed 195 with value = 0.9619 +Query 1/1: Action query time = 5.180 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8656 +t=90: Selected seed 195 with value = 0.8656 +Query 1/1: Action query time = 3.792 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8904 +t=106: Selected seed 195 with value = 0.8904 +Query 1/1: Action query time = 6.504 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9240 +t=122: Selected seed 195 with value = 0.9240 +Query 1/1: Action query time = 5.838 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9639 +t=138: Selected seed 195 with value = 0.9639 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=12--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 6 (50.0%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 7.424 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4665 +t=10: Selected seed 195 with value = 0.4665 +Query 1/1: Action query time = 5.689 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6024 +t=26: Selected seed 195 with value = 0.6024 +Query 1/1: Action query time = 5.224 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6884 +t=42: Selected seed 195 with value = 0.6884 +Query 1/1: Action query time = 6.637 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7606 +t=58: Selected seed 195 with value = 0.7606 +Query 1/1: Action query time = 6.101 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9280 +t=74: Selected seed 195 with value = 0.9280 +Query 1/1: Action query time = 4.419 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8634 +t=90: Selected seed 195 with value = 0.8634 +Query 1/1: Action query time = 5.306 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9489 +t=106: Selected seed 195 with value = 0.9489 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s0/2026_08_03-01_16_46--with_future_img--episode=13--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 13 +# successes: 7 (53.8%) +Current task success rate: 0.5384615384615384 +Current total success rate: 0.5384615384615384 +Final results: +Total episodes: 13 +Total successes: 7 +Overall success rate: 0.5385 (53.8%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_46--t7rc800x50_t1_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_46--t7rc800x50_t1_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..747546e6136b8a31312226e5444af810913a9e04 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_46--t7rc800x50_t1_s1.txt @@ -0,0 +1,927 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t1_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 5.525 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5256 +t=10: Selected seed 195 with value = 0.5256 +Query 1/1: Action query time = 4.584 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6314 +t=26: Selected seed 195 with value = 0.6314 +Query 1/1: Action query time = 4.184 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7340 +t=42: Selected seed 195 with value = 0.7340 +Query 1/1: Action query time = 6.429 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8988 +t=58: Selected seed 195 with value = 0.8988 +Query 1/1: Action query time = 5.990 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9733 +t=74: Selected seed 195 with value = 0.9733 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 6.006 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5123 +t=10: Selected seed 195 with value = 0.5123 +Query 1/1: Action query time = 6.684 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5801 +t=26: Selected seed 195 with value = 0.5801 +Query 1/1: Action query time = 6.232 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6641 +t=42: Selected seed 195 with value = 0.6641 +Query 1/1: Action query time = 6.156 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7786 +t=58: Selected seed 195 with value = 0.7786 +Query 1/1: Action query time = 5.311 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9574 +t=74: Selected seed 195 with value = 0.9574 +Query 1/1: Action query time = 5.016 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8699 +t=90: Selected seed 195 with value = 0.8699 +Query 1/1: Action query time = 6.166 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9032 +t=106: Selected seed 195 with value = 0.9032 +Query 1/1: Action query time = 3.725 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9184 +t=122: Selected seed 195 with value = 0.9184 +Query 1/1: Action query time = 6.127 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8520 +t=138: Selected seed 195 with value = 0.8520 +Query 1/1: Action query time = 6.222 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8686 +t=154: Selected seed 195 with value = 0.8686 +Query 1/1: Action query time = 4.149 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8541 +t=170: Selected seed 195 with value = 0.8541 +Query 1/1: Action query time = 3.294 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8771 +t=186: Selected seed 195 with value = 0.8771 +Query 1/1: Action query time = 3.307 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8689 +t=202: Selected seed 195 with value = 0.8689 +Query 1/1: Action query time = 4.317 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8309 +t=218: Selected seed 195 with value = 0.8309 +Query 1/1: Action query time = 3.085 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8401 +t=234: Selected seed 195 with value = 0.8401 +Query 1/1: Action query time = 3.218 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9290 +t=250: Selected seed 195 with value = 0.9290 +Query 1/1: Action query time = 4.378 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8744 +t=266: Selected seed 195 with value = 0.8744 +Query 1/1: Action query time = 4.783 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9329 +t=282: Selected seed 195 with value = 0.9329 +Query 1/1: Action query time = 5.713 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9186 +t=298: Selected seed 195 with value = 0.9186 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.659 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5098 +t=10: Selected seed 195 with value = 0.5098 +Query 1/1: Action query time = 5.567 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5998 +t=26: Selected seed 195 with value = 0.5998 +Query 1/1: Action query time = 5.085 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6758 +t=42: Selected seed 195 with value = 0.6758 +Query 1/1: Action query time = 6.087 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7656 +t=58: Selected seed 195 with value = 0.7656 +Query 1/1: Action query time = 5.296 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8371 +t=74: Selected seed 195 with value = 0.8371 +Query 1/1: Action query time = 6.580 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8980 +t=90: Selected seed 195 with value = 0.8980 +Query 1/1: Action query time = 4.954 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9859 +t=106: Selected seed 195 with value = 0.9859 +Query 1/1: Action query time = 6.355 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=122: Selected seed 195 with value = 0.9997 +Query 1/1: Action query time = 2.919 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9380 +t=138: Selected seed 195 with value = 0.9380 +Query 1/1: Action query time = 6.107 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9905 +t=154: Selected seed 195 with value = 0.9905 +Query 1/1: Action query time = 5.379 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9255 +t=170: Selected seed 195 with value = 0.9255 +Query 1/1: Action query time = 5.499 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8783 +t=186: Selected seed 195 with value = 0.8783 +Query 1/1: Action query time = 4.810 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8633 +t=202: Selected seed 195 with value = 0.8633 +Query 1/1: Action query time = 2.415 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8831 +t=218: Selected seed 195 with value = 0.8831 +Query 1/1: Action query time = 3.156 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9002 +t=234: Selected seed 195 with value = 0.9002 +Query 1/1: Action query time = 4.063 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9043 +t=250: Selected seed 195 with value = 0.9043 +Query 1/1: Action query time = 5.233 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8195 +t=266: Selected seed 195 with value = 0.8195 +Query 1/1: Action query time = 4.803 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9521 +t=282: Selected seed 195 with value = 0.9521 +Query 1/1: Action query time = 5.712 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8742 +t=298: Selected seed 195 with value = 0.8742 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 5.158 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5161 +t=10: Selected seed 195 with value = 0.5161 +Query 1/1: Action query time = 3.887 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6008 +t=26: Selected seed 195 with value = 0.6008 +Query 1/1: Action query time = 5.463 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6731 +t=42: Selected seed 195 with value = 0.6731 +Query 1/1: Action query time = 5.720 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8351 +t=58: Selected seed 195 with value = 0.8351 +Query 1/1: Action query time = 3.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9504 +t=74: Selected seed 195 with value = 0.9504 +Query 1/1: Action query time = 4.717 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9848 +t=90: Selected seed 195 with value = 0.9848 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=4--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 2 (50.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 5.318 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5181 +t=10: Selected seed 195 with value = 0.5181 +Query 1/1: Action query time = 5.146 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6462 +t=26: Selected seed 195 with value = 0.6462 +Query 1/1: Action query time = 5.011 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7338 +t=42: Selected seed 195 with value = 0.7338 +Query 1/1: Action query time = 4.382 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9577 +t=58: Selected seed 195 with value = 0.9577 +Query 1/1: Action query time = 2.665 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9612 +t=74: Selected seed 195 with value = 0.9612 +Query 1/1: Action query time = 3.526 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=90: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 3.313 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7893 +t=106: Selected seed 195 with value = 0.7893 +Query 1/1: Action query time = 6.017 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9755 +t=122: Selected seed 195 with value = 0.9755 +Query 1/1: Action query time = 6.654 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9887 +t=138: Selected seed 195 with value = 0.9887 +Query 1/1: Action query time = 5.908 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7904 +t=154: Selected seed 195 with value = 0.7904 +Query 1/1: Action query time = 6.599 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8862 +t=170: Selected seed 195 with value = 0.8862 +Query 1/1: Action query time = 6.699 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8887 +t=186: Selected seed 195 with value = 0.8887 +Query 1/1: Action query time = 5.539 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9277 +t=202: Selected seed 195 with value = 0.9277 +Query 1/1: Action query time = 3.411 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9521 +t=218: Selected seed 195 with value = 0.9521 +Query 1/1: Action query time = 5.337 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9466 +t=234: Selected seed 195 with value = 0.9466 +Query 1/1: Action query time = 4.184 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9621 +t=250: Selected seed 195 with value = 0.9621 +Query 1/1: Action query time = 6.409 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9840 +t=266: Selected seed 195 with value = 0.9840 +Query 1/1: Action query time = 6.523 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=282: Selected seed 195 with value = 0.9965 +Query 1/1: Action query time = 5.209 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=298: Selected seed 195 with value = 0.9983 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=5--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 5 +# successes: 2 (40.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 5.013 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5050 +t=10: Selected seed 195 with value = 0.5050 +Query 1/1: Action query time = 4.871 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5872 +t=26: Selected seed 195 with value = 0.5872 +Query 1/1: Action query time = 4.013 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6575 +t=42: Selected seed 195 with value = 0.6575 +Query 1/1: Action query time = 4.592 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7893 +t=58: Selected seed 195 with value = 0.7893 +Query 1/1: Action query time = 4.214 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9318 +t=74: Selected seed 195 with value = 0.9318 +Query 1/1: Action query time = 3.178 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7609 +t=90: Selected seed 195 with value = 0.7609 +Query 1/1: Action query time = 5.259 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8517 +t=106: Selected seed 195 with value = 0.8517 +Query 1/1: Action query time = 4.295 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9092 +t=122: Selected seed 195 with value = 0.9092 +Query 1/1: Action query time = 6.106 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9400 +t=138: Selected seed 195 with value = 0.9400 +Query 1/1: Action query time = 4.571 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.038 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7848 +t=170: Selected seed 195 with value = 0.7848 +Query 1/1: Action query time = 6.103 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7726 +t=186: Selected seed 195 with value = 0.7726 +Query 1/1: Action query time = 5.090 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8624 +t=202: Selected seed 195 with value = 0.8624 +Query 1/1: Action query time = 6.477 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9646 +t=218: Selected seed 195 with value = 0.9646 +Query 1/1: Action query time = 5.111 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=234: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 3.387 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9862 +t=250: Selected seed 195 with value = 0.9862 +Query 1/1: Action query time = 4.855 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=266: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 4.963 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=282: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 6.313 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9933 +t=298: Selected seed 195 with value = 0.9933 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=6--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 6 +# successes: 2 (33.3%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 4.133 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5323 +t=10: Selected seed 195 with value = 0.5323 +Query 1/1: Action query time = 3.043 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6244 +t=26: Selected seed 195 with value = 0.6244 +Query 1/1: Action query time = 4.294 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7212 +t=42: Selected seed 195 with value = 0.7212 +Query 1/1: Action query time = 4.202 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8892 +t=58: Selected seed 195 with value = 0.8892 +Query 1/1: Action query time = 5.014 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9683 +t=74: Selected seed 195 with value = 0.9683 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=7--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 3 (42.9%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 5.618 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5554 +t=10: Selected seed 195 with value = 0.5554 +Query 1/1: Action query time = 5.046 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5917 +t=26: Selected seed 195 with value = 0.5917 +Query 1/1: Action query time = 3.143 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6818 +t=42: Selected seed 195 with value = 0.6818 +Query 1/1: Action query time = 4.959 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7666 +t=58: Selected seed 195 with value = 0.7666 +Query 1/1: Action query time = 4.593 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8328 +t=74: Selected seed 195 with value = 0.8328 +Query 1/1: Action query time = 5.549 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9390 +t=90: Selected seed 195 with value = 0.9390 +Query 1/1: Action query time = 4.959 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9626 +t=106: Selected seed 195 with value = 0.9626 +Query 1/1: Action query time = 4.776 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9933 +t=122: Selected seed 195 with value = 0.9933 +Query 1/1: Action query time = 5.014 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9175 +t=138: Selected seed 195 with value = 0.9175 +Query 1/1: Action query time = 4.455 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=154: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 3.987 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8186 +t=170: Selected seed 195 with value = 0.8186 +Query 1/1: Action query time = 4.983 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8936 +t=186: Selected seed 195 with value = 0.8936 +Query 1/1: Action query time = 4.854 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9779 +t=202: Selected seed 195 with value = 0.9779 +Query 1/1: Action query time = 5.518 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9367 +t=218: Selected seed 195 with value = 0.9367 +Query 1/1: Action query time = 6.101 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9759 +t=234: Selected seed 195 with value = 0.9759 +Query 1/1: Action query time = 5.944 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9825 +t=250: Selected seed 195 with value = 0.9825 +Query 1/1: Action query time = 5.432 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9893 +t=266: Selected seed 195 with value = 0.9893 +Query 1/1: Action query time = 4.493 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.914 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=298: Selected seed 195 with value = 0.9978 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=8--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 8 +# successes: 3 (37.5%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 4.991 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4894 +t=10: Selected seed 195 with value = 0.4894 +Query 1/1: Action query time = 5.204 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5893 +t=26: Selected seed 195 with value = 0.5893 +Query 1/1: Action query time = 4.008 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6901 +t=42: Selected seed 195 with value = 0.6901 +Query 1/1: Action query time = 5.757 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7682 +t=58: Selected seed 195 with value = 0.7682 +Query 1/1: Action query time = 5.812 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8594 +t=74: Selected seed 195 with value = 0.8594 +Query 1/1: Action query time = 6.306 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9738 +t=90: Selected seed 195 with value = 0.9738 +Query 1/1: Action query time = 4.294 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9895 +t=106: Selected seed 195 with value = 0.9895 +Query 1/1: Action query time = 5.312 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9959 +t=122: Selected seed 195 with value = 0.9959 +Query 1/1: Action query time = 5.727 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9090 +t=138: Selected seed 195 with value = 0.9090 +Query 1/1: Action query time = 4.301 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=154: Selected seed 195 with value = 0.9867 +Query 1/1: Action query time = 4.221 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8826 +t=170: Selected seed 195 with value = 0.8826 +Query 1/1: Action query time = 3.898 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8617 +t=186: Selected seed 195 with value = 0.8617 +Query 1/1: Action query time = 5.217 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8265 +t=202: Selected seed 195 with value = 0.8265 +Query 1/1: Action query time = 4.687 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8220 +t=218: Selected seed 195 with value = 0.8220 +Query 1/1: Action query time = 5.956 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9139 +t=234: Selected seed 195 with value = 0.9139 +Query 1/1: Action query time = 3.984 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8833 +t=250: Selected seed 195 with value = 0.8833 +Query 1/1: Action query time = 3.193 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8975 +t=266: Selected seed 195 with value = 0.8975 +Query 1/1: Action query time = 5.918 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9479 +t=282: Selected seed 195 with value = 0.9479 +Query 1/1: Action query time = 4.077 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9643 +t=298: Selected seed 195 with value = 0.9643 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 9 +# successes: 3 (33.3%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 3.737 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5067 +t=10: Selected seed 195 with value = 0.5067 +Query 1/1: Action query time = 5.088 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6168 +t=26: Selected seed 195 with value = 0.6168 +Query 1/1: Action query time = 5.156 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7071 +t=42: Selected seed 195 with value = 0.7071 +Query 1/1: Action query time = 4.711 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8886 +t=58: Selected seed 195 with value = 0.8886 +Query 1/1: Action query time = 5.682 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9619 +t=74: Selected seed 195 with value = 0.9619 +Query 1/1: Action query time = 5.007 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9715 +t=90: Selected seed 195 with value = 0.9715 +Query 1/1: Action query time = 4.113 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9228 +t=106: Selected seed 195 with value = 0.9228 +Query 1/1: Action query time = 5.334 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9449 +t=122: Selected seed 195 with value = 0.9449 +Query 1/1: Action query time = 4.910 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9008 +t=138: Selected seed 195 with value = 0.9008 +Query 1/1: Action query time = 4.005 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8579 +t=154: Selected seed 195 with value = 0.8579 +Query 1/1: Action query time = 4.560 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8151 +t=170: Selected seed 195 with value = 0.8151 +Query 1/1: Action query time = 5.169 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7919 +t=186: Selected seed 195 with value = 0.7919 +Query 1/1: Action query time = 5.512 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8012 +t=202: Selected seed 195 with value = 0.8012 +Query 1/1: Action query time = 4.461 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7953 +t=218: Selected seed 195 with value = 0.7953 +Query 1/1: Action query time = 6.430 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8879 +t=234: Selected seed 195 with value = 0.8879 +Query 1/1: Action query time = 5.052 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8692 +t=250: Selected seed 195 with value = 0.8692 +Query 1/1: Action query time = 5.390 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8273 +t=266: Selected seed 195 with value = 0.8273 +Query 1/1: Action query time = 4.322 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8379 +t=282: Selected seed 195 with value = 0.8379 +Query 1/1: Action query time = 5.443 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8305 +t=298: Selected seed 195 with value = 0.8305 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=10--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 10 +# successes: 3 (30.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 4.464 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5352 +t=10: Selected seed 195 with value = 0.5352 +Query 1/1: Action query time = 4.844 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6460 +t=26: Selected seed 195 with value = 0.6460 +Query 1/1: Action query time = 5.078 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7785 +t=42: Selected seed 195 with value = 0.7785 +Query 1/1: Action query time = 4.766 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9422 +t=58: Selected seed 195 with value = 0.9422 +Query 1/1: Action query time = 5.551 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.399 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9936 +t=90: Selected seed 195 with value = 0.9936 +Query 1/1: Action query time = 5.487 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7899 +t=106: Selected seed 195 with value = 0.7899 +Query 1/1: Action query time = 4.279 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7875 +t=122: Selected seed 195 with value = 0.7875 +Query 1/1: Action query time = 4.812 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8591 +t=138: Selected seed 195 with value = 0.8591 +Query 1/1: Action query time = 6.608 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9607 +t=154: Selected seed 195 with value = 0.9607 +Query 1/1: Action query time = 4.754 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9866 +t=170: Selected seed 195 with value = 0.9866 +Query 1/1: Action query time = 5.123 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=186: Selected seed 195 with value = 0.9912 +Query 1/1: Action query time = 5.707 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9866 +t=202: Selected seed 195 with value = 0.9866 +Query 1/1: Action query time = 3.756 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9747 +t=218: Selected seed 195 with value = 0.9747 +Query 1/1: Action query time = 4.339 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9408 +t=234: Selected seed 195 with value = 0.9408 +Query 1/1: Action query time = 4.609 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9120 +t=250: Selected seed 195 with value = 0.9120 +Query 1/1: Action query time = 4.333 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8798 +t=266: Selected seed 195 with value = 0.8798 +Query 1/1: Action query time = 2.940 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8725 +t=282: Selected seed 195 with value = 0.8725 +Query 1/1: Action query time = 4.630 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8605 +t=298: Selected seed 195 with value = 0.8605 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=11--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 11 +# successes: 3 (27.3%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 4.442 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4882 +t=10: Selected seed 195 with value = 0.4882 +Query 1/1: Action query time = 4.169 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6030 +t=26: Selected seed 195 with value = 0.6030 +Query 1/1: Action query time = 3.798 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6659 +t=42: Selected seed 195 with value = 0.6659 +Query 1/1: Action query time = 5.486 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8342 +t=58: Selected seed 195 with value = 0.8342 +Query 1/1: Action query time = 4.932 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9708 +t=74: Selected seed 195 with value = 0.9708 +Query 1/1: Action query time = 4.371 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7629 +t=90: Selected seed 195 with value = 0.7629 +Query 1/1: Action query time = 5.988 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8928 +t=106: Selected seed 195 with value = 0.8928 +Query 1/1: Action query time = 5.503 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9042 +t=122: Selected seed 195 with value = 0.9042 +Query 1/1: Action query time = 4.181 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9307 +t=138: Selected seed 195 with value = 0.9307 +Query 1/1: Action query time = 6.144 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9074 +t=154: Selected seed 195 with value = 0.9074 +Query 1/1: Action query time = 6.199 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=170: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 6.244 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8054 +t=186: Selected seed 195 with value = 0.8054 +Query 1/1: Action query time = 3.957 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9239 +t=202: Selected seed 195 with value = 0.9239 +Query 1/1: Action query time = 4.817 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9338 +t=218: Selected seed 195 with value = 0.9338 +Query 1/1: Action query time = 5.850 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8340 +t=234: Selected seed 195 with value = 0.8340 +Query 1/1: Action query time = 3.920 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8519 +t=250: Selected seed 195 with value = 0.8519 +Query 1/1: Action query time = 4.653 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9008 +t=266: Selected seed 195 with value = 0.9008 +Query 1/1: Action query time = 3.880 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9571 +t=282: Selected seed 195 with value = 0.9571 +Query 1/1: Action query time = 4.117 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9750 +t=298: Selected seed 195 with value = 0.9750 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=12--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 12 +# successes: 3 (25.0%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 3.548 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5180 +t=10: Selected seed 195 with value = 0.5180 +Query 1/1: Action query time = 4.699 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6058 +t=26: Selected seed 195 with value = 0.6058 +Query 1/1: Action query time = 4.413 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6880 +t=42: Selected seed 195 with value = 0.6880 +Query 1/1: Action query time = 4.990 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7723 +t=58: Selected seed 195 with value = 0.7723 +Query 1/1: Action query time = 4.288 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8330 +t=74: Selected seed 195 with value = 0.8330 +Query 1/1: Action query time = 4.928 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7603 +t=90: Selected seed 195 with value = 0.7603 +Query 1/1: Action query time = 5.208 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9292 +t=106: Selected seed 195 with value = 0.9292 +Query 1/1: Action query time = 4.603 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9587 +t=122: Selected seed 195 with value = 0.9587 +Query 1/1: Action query time = 5.484 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.273 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9381 +t=154: Selected seed 195 with value = 0.9381 +Query 1/1: Action query time = 3.006 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9849 +t=170: Selected seed 195 with value = 0.9849 +Query 1/1: Action query time = 3.940 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9686 +t=186: Selected seed 195 with value = 0.9686 +Query 1/1: Action query time = 3.536 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.181 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7740 +t=218: Selected seed 195 with value = 0.7740 +Query 1/1: Action query time = 3.824 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7534 +t=234: Selected seed 195 with value = 0.7534 +Query 1/1: Action query time = 4.018 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7696 +t=250: Selected seed 195 with value = 0.7696 +Query 1/1: Action query time = 3.801 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8496 +t=266: Selected seed 195 with value = 0.8496 +Query 1/1: Action query time = 3.681 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9518 +t=282: Selected seed 195 with value = 0.9518 +Query 1/1: Action query time = 3.243 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=298: Selected seed 195 with value = 0.9867 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t1_s1/2026_08_03-01_16_46--with_future_img--episode=13--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 13 +# successes: 3 (23.1%) +Current task success rate: 0.23076923076923078 +Current total success rate: 0.23076923076923078 +Final results: +Total episodes: 13 +Total successes: 3 +Overall success rate: 0.2308 (23.1%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_48--t7dc800x50_t1_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_48--t7dc800x50_t1_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..fdc837640a9779e5af500f2ad0fb98efc015a5a5 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_48--t7dc800x50_t1_s3.txt @@ -0,0 +1,852 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_demochan800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7dc800x50_t1_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,7,11,15,19,23,27,31,35,39,43,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 6.332 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5164 +t=10: Selected seed 195 with value = 0.5164 +Query 1/1: Action query time = 6.181 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6138 +t=26: Selected seed 195 with value = 0.6138 +Query 1/1: Action query time = 5.662 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7505 +t=42: Selected seed 195 with value = 0.7505 +Query 1/1: Action query time = 5.384 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8402 +t=58: Selected seed 195 with value = 0.8402 +Query 1/1: Action query time = 7.072 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9815 +t=74: Selected seed 195 with value = 0.9815 +Query 1/1: Action query time = 5.779 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9737 +t=90: Selected seed 195 with value = 0.9737 +Query 1/1: Action query time = 5.959 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9897 +t=106: Selected seed 195 with value = 0.9897 +Query 1/1: Action query time = 5.675 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9920 +t=122: Selected seed 195 with value = 0.9920 +Query 1/1: Action query time = 7.145 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9830 +t=138: Selected seed 195 with value = 0.9830 +Query 1/1: Action query time = 6.680 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9524 +t=154: Selected seed 195 with value = 0.9524 +Query 1/1: Action query time = 5.747 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9834 +t=170: Selected seed 195 with value = 0.9834 +Query 1/1: Action query time = 3.263 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9718 +t=186: Selected seed 195 with value = 0.9718 +Query 1/1: Action query time = 5.299 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9657 +t=202: Selected seed 195 with value = 0.9657 +Query 1/1: Action query time = 6.602 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9621 +t=218: Selected seed 195 with value = 0.9621 +Query 1/1: Action query time = 5.689 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9610 +t=234: Selected seed 195 with value = 0.9610 +Query 1/1: Action query time = 4.623 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9553 +t=250: Selected seed 195 with value = 0.9553 +Query 1/1: Action query time = 4.544 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9503 +t=266: Selected seed 195 with value = 0.9503 +Query 1/1: Action query time = 5.451 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9433 +t=282: Selected seed 195 with value = 0.9433 +Query 1/1: Action query time = 5.251 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9470 +t=298: Selected seed 195 with value = 0.9470 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 3.473 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5240 +t=10: Selected seed 195 with value = 0.5240 +Query 1/1: Action query time = 3.790 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6153 +t=26: Selected seed 195 with value = 0.6153 +Query 1/1: Action query time = 4.700 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7398 +t=42: Selected seed 195 with value = 0.7398 +Query 1/1: Action query time = 6.816 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8518 +t=58: Selected seed 195 with value = 0.8518 +Query 1/1: Action query time = 5.422 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=74: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 5.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9935 +t=90: Selected seed 195 with value = 0.9935 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 7.690 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5119 +t=10: Selected seed 195 with value = 0.5119 +Query 1/1: Action query time = 6.082 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6009 +t=26: Selected seed 195 with value = 0.6009 +Query 1/1: Action query time = 7.697 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7191 +t=42: Selected seed 195 with value = 0.7191 +Query 1/1: Action query time = 5.639 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9236 +t=58: Selected seed 195 with value = 0.9236 +Query 1/1: Action query time = 7.137 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9597 +t=74: Selected seed 195 with value = 0.9597 +Query 1/1: Action query time = 4.550 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9813 +t=90: Selected seed 195 with value = 0.9813 +Query 1/1: Action query time = 5.992 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=106: Selected seed 195 with value = 0.9870 +Query 1/1: Action query time = 5.716 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9769 +t=122: Selected seed 195 with value = 0.9769 +Query 1/1: Action query time = 6.008 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9635 +t=138: Selected seed 195 with value = 0.9635 +Query 1/1: Action query time = 4.149 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9609 +t=154: Selected seed 195 with value = 0.9609 +Query 1/1: Action query time = 4.059 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9713 +t=170: Selected seed 195 with value = 0.9713 +Query 1/1: Action query time = 6.322 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9767 +t=186: Selected seed 195 with value = 0.9767 +Query 1/1: Action query time = 5.814 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9720 +t=202: Selected seed 195 with value = 0.9720 +Query 1/1: Action query time = 4.503 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9719 +t=218: Selected seed 195 with value = 0.9719 +Query 1/1: Action query time = 3.375 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9833 +t=234: Selected seed 195 with value = 0.9833 +Query 1/1: Action query time = 2.686 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9743 +t=250: Selected seed 195 with value = 0.9743 +Query 1/1: Action query time = 3.563 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9660 +t=266: Selected seed 195 with value = 0.9660 +Query 1/1: Action query time = 6.048 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9666 +t=282: Selected seed 195 with value = 0.9666 +Query 1/1: Action query time = 6.753 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9548 +t=298: Selected seed 195 with value = 0.9548 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 5.652 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5109 +t=10: Selected seed 195 with value = 0.5109 +Query 1/1: Action query time = 5.143 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6003 +t=26: Selected seed 195 with value = 0.6003 +Query 1/1: Action query time = 5.909 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6720 +t=42: Selected seed 195 with value = 0.6720 +Query 1/1: Action query time = 7.133 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9217 +t=58: Selected seed 195 with value = 0.9217 +Query 1/1: Action query time = 6.828 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9645 +t=74: Selected seed 195 with value = 0.9645 +Query 1/1: Action query time = 4.690 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9591 +t=90: Selected seed 195 with value = 0.9591 +Query 1/1: Action query time = 6.023 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9679 +t=106: Selected seed 195 with value = 0.9679 +Query 1/1: Action query time = 6.041 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9621 +t=122: Selected seed 195 with value = 0.9621 +Query 1/1: Action query time = 5.668 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9493 +t=138: Selected seed 195 with value = 0.9493 +Query 1/1: Action query time = 4.923 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9422 +t=154: Selected seed 195 with value = 0.9422 +Query 1/1: Action query time = 6.829 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8812 +t=170: Selected seed 195 with value = 0.8812 +Query 1/1: Action query time = 5.237 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8566 +t=186: Selected seed 195 with value = 0.8566 +Query 1/1: Action query time = 5.058 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8859 +t=202: Selected seed 195 with value = 0.8859 +Query 1/1: Action query time = 5.696 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9312 +t=218: Selected seed 195 with value = 0.9312 +Query 1/1: Action query time = 5.796 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9032 +t=234: Selected seed 195 with value = 0.9032 +Query 1/1: Action query time = 3.573 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9076 +t=250: Selected seed 195 with value = 0.9076 +Query 1/1: Action query time = 3.175 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9224 +t=266: Selected seed 195 with value = 0.9224 +Query 1/1: Action query time = 3.108 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9210 +t=282: Selected seed 195 with value = 0.9210 +Query 1/1: Action query time = 7.253 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9278 +t=298: Selected seed 195 with value = 0.9278 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 4 +# successes: 1 (25.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 5.244 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5224 +t=10: Selected seed 195 with value = 0.5224 +Query 1/1: Action query time = 5.517 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6182 +t=26: Selected seed 195 with value = 0.6182 +Query 1/1: Action query time = 6.575 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7096 +t=42: Selected seed 195 with value = 0.7096 +Query 1/1: Action query time = 5.990 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9684 +t=58: Selected seed 195 with value = 0.9684 +Query 1/1: Action query time = 7.134 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9934 +t=74: Selected seed 195 with value = 0.9934 +Query 1/1: Action query time = 5.159 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9876 +t=90: Selected seed 195 with value = 0.9876 +Query 1/1: Action query time = 5.100 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9898 +t=106: Selected seed 195 with value = 0.9898 +Query 1/1: Action query time = 5.618 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=122: Selected seed 195 with value = 0.9912 +Query 1/1: Action query time = 5.529 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=138: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 6.911 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9916 +t=154: Selected seed 195 with value = 0.9916 +Query 1/1: Action query time = 6.701 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9923 +t=170: Selected seed 195 with value = 0.9923 +Query 1/1: Action query time = 4.074 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=186: Selected seed 195 with value = 0.9926 +Query 1/1: Action query time = 5.146 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=202: Selected seed 195 with value = 0.9914 +Query 1/1: Action query time = 4.752 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9904 +t=218: Selected seed 195 with value = 0.9904 +Query 1/1: Action query time = 3.480 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=234: Selected seed 195 with value = 0.9886 +Query 1/1: Action query time = 5.355 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=250: Selected seed 195 with value = 0.9870 +Query 1/1: Action query time = 3.425 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9866 +t=266: Selected seed 195 with value = 0.9866 +Query 1/1: Action query time = 5.228 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9853 +t=282: Selected seed 195 with value = 0.9853 +Query 1/1: Action query time = 5.745 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9841 +t=298: Selected seed 195 with value = 0.9841 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=5--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 5 +# successes: 1 (20.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 6.882 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4833 +t=10: Selected seed 195 with value = 0.4833 +Query 1/1: Action query time = 6.915 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5667 +t=26: Selected seed 195 with value = 0.5667 +Query 1/1: Action query time = 6.737 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6630 +t=42: Selected seed 195 with value = 0.6630 +Query 1/1: Action query time = 6.059 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7859 +t=58: Selected seed 195 with value = 0.7859 +Query 1/1: Action query time = 4.100 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8829 +t=74: Selected seed 195 with value = 0.8829 +Query 1/1: Action query time = 6.221 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8683 +t=90: Selected seed 195 with value = 0.8683 +Query 1/1: Action query time = 6.952 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8974 +t=106: Selected seed 195 with value = 0.8974 +Query 1/1: Action query time = 5.943 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8642 +t=122: Selected seed 195 with value = 0.8642 +Query 1/1: Action query time = 6.019 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9545 +t=138: Selected seed 195 with value = 0.9545 +Query 1/1: Action query time = 5.603 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9531 +t=154: Selected seed 195 with value = 0.9531 +Query 1/1: Action query time = 5.214 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9623 +t=170: Selected seed 195 with value = 0.9623 +Query 1/1: Action query time = 4.298 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9624 +t=186: Selected seed 195 with value = 0.9624 +Query 1/1: Action query time = 5.427 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9527 +t=202: Selected seed 195 with value = 0.9527 +Query 1/1: Action query time = 3.162 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9448 +t=218: Selected seed 195 with value = 0.9448 +Query 1/1: Action query time = 4.166 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9410 +t=234: Selected seed 195 with value = 0.9410 +Query 1/1: Action query time = 5.052 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9371 +t=250: Selected seed 195 with value = 0.9371 +Query 1/1: Action query time = 2.841 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9517 +t=266: Selected seed 195 with value = 0.9517 +Query 1/1: Action query time = 4.264 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9571 +t=282: Selected seed 195 with value = 0.9571 +Query 1/1: Action query time = 5.113 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9585 +t=298: Selected seed 195 with value = 0.9585 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=6--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 6 +# successes: 1 (16.7%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 5.762 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5234 +t=10: Selected seed 195 with value = 0.5234 +Query 1/1: Action query time = 6.907 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6047 +t=26: Selected seed 195 with value = 0.6047 +Query 1/1: Action query time = 6.987 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6823 +t=42: Selected seed 195 with value = 0.6823 +Query 1/1: Action query time = 6.329 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8200 +t=58: Selected seed 195 with value = 0.8200 +Query 1/1: Action query time = 6.869 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9562 +t=74: Selected seed 195 with value = 0.9562 +Query 1/1: Action query time = 4.504 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9921 +t=90: Selected seed 195 with value = 0.9921 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=7--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 2 (28.6%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 5.589 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5046 +t=10: Selected seed 195 with value = 0.5046 +Query 1/1: Action query time = 5.490 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6156 +t=26: Selected seed 195 with value = 0.6156 +Query 1/1: Action query time = 5.948 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7463 +t=42: Selected seed 195 with value = 0.7463 +Query 1/1: Action query time = 6.524 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9762 +t=58: Selected seed 195 with value = 0.9762 +Query 1/1: Action query time = 6.656 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9834 +t=74: Selected seed 195 with value = 0.9834 +Query 1/1: Action query time = 3.995 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=90: Selected seed 195 with value = 0.9870 +Query 1/1: Action query time = 4.673 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9890 +t=106: Selected seed 195 with value = 0.9890 +Query 1/1: Action query time = 4.912 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=122: Selected seed 195 with value = 0.9912 +Query 1/1: Action query time = 4.052 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=138: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 3.937 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=154: Selected seed 195 with value = 0.9870 +Query 1/1: Action query time = 5.344 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9756 +t=170: Selected seed 195 with value = 0.9756 +Query 1/1: Action query time = 5.825 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9799 +t=186: Selected seed 195 with value = 0.9799 +Query 1/1: Action query time = 6.917 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=202: Selected seed 195 with value = 0.9809 +Query 1/1: Action query time = 4.979 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9772 +t=218: Selected seed 195 with value = 0.9772 +Query 1/1: Action query time = 5.722 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9728 +t=234: Selected seed 195 with value = 0.9728 +Query 1/1: Action query time = 5.884 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9693 +t=250: Selected seed 195 with value = 0.9693 +Query 1/1: Action query time = 6.428 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9555 +t=266: Selected seed 195 with value = 0.9555 +Query 1/1: Action query time = 5.145 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9413 +t=282: Selected seed 195 with value = 0.9413 +Query 1/1: Action query time = 5.807 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9310 +t=298: Selected seed 195 with value = 0.9310 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=8--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 8 +# successes: 2 (25.0%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 4.149 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5433 +t=10: Selected seed 195 with value = 0.5433 +Query 1/1: Action query time = 7.406 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6388 +t=26: Selected seed 195 with value = 0.6388 +Query 1/1: Action query time = 6.176 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7703 +t=42: Selected seed 195 with value = 0.7703 +Query 1/1: Action query time = 5.772 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9799 +t=58: Selected seed 195 with value = 0.9799 +Query 1/1: Action query time = 6.294 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9804 +t=74: Selected seed 195 with value = 0.9804 +Query 1/1: Action query time = 7.777 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9801 +t=90: Selected seed 195 with value = 0.9801 +Query 1/1: Action query time = 5.602 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9829 +t=106: Selected seed 195 with value = 0.9829 +Query 1/1: Action query time = 5.660 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9824 +t=122: Selected seed 195 with value = 0.9824 +Query 1/1: Action query time = 4.358 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9770 +t=138: Selected seed 195 with value = 0.9770 +Query 1/1: Action query time = 3.053 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9670 +t=154: Selected seed 195 with value = 0.9670 +Query 1/1: Action query time = 4.403 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9664 +t=170: Selected seed 195 with value = 0.9664 +Query 1/1: Action query time = 3.736 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9467 +t=186: Selected seed 195 with value = 0.9467 +Query 1/1: Action query time = 5.091 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9240 +t=202: Selected seed 195 with value = 0.9240 +Query 1/1: Action query time = 6.334 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9276 +t=218: Selected seed 195 with value = 0.9276 +Query 1/1: Action query time = 6.312 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9387 +t=234: Selected seed 195 with value = 0.9387 +Query 1/1: Action query time = 5.776 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9243 +t=250: Selected seed 195 with value = 0.9243 +Query 1/1: Action query time = 4.403 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9265 +t=266: Selected seed 195 with value = 0.9265 +Query 1/1: Action query time = 5.739 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9364 +t=282: Selected seed 195 with value = 0.9364 +Query 1/1: Action query time = 4.435 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9452 +t=298: Selected seed 195 with value = 0.9452 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 9 +# successes: 2 (22.2%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 5.696 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5106 +t=10: Selected seed 195 with value = 0.5106 +Query 1/1: Action query time = 5.046 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5509 +t=26: Selected seed 195 with value = 0.5509 +Query 1/1: Action query time = 6.489 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6519 +t=42: Selected seed 195 with value = 0.6519 +Query 1/1: Action query time = 5.595 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7740 +t=58: Selected seed 195 with value = 0.7740 +Query 1/1: Action query time = 5.727 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8828 +t=74: Selected seed 195 with value = 0.8828 +Query 1/1: Action query time = 4.042 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8600 +t=90: Selected seed 195 with value = 0.8600 +Query 1/1: Action query time = 4.814 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8918 +t=106: Selected seed 195 with value = 0.8918 +Query 1/1: Action query time = 5.414 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8727 +t=122: Selected seed 195 with value = 0.8727 +Query 1/1: Action query time = 4.049 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9706 +t=138: Selected seed 195 with value = 0.9706 +Query 1/1: Action query time = 3.914 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9680 +t=154: Selected seed 195 with value = 0.9680 +Query 1/1: Action query time = 4.030 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9314 +t=170: Selected seed 195 with value = 0.9314 +Query 1/1: Action query time = 3.114 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9549 +t=186: Selected seed 195 with value = 0.9549 +Query 1/1: Action query time = 3.241 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9576 +t=202: Selected seed 195 with value = 0.9576 +Query 1/1: Action query time = 2.745 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9456 +t=218: Selected seed 195 with value = 0.9456 +Query 1/1: Action query time = 3.134 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9411 +t=234: Selected seed 195 with value = 0.9411 +Query 1/1: Action query time = 3.741 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9553 +t=250: Selected seed 195 with value = 0.9553 +Query 1/1: Action query time = 6.438 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9524 +t=266: Selected seed 195 with value = 0.9524 +Query 1/1: Action query time = 6.987 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9482 +t=282: Selected seed 195 with value = 0.9482 +Query 1/1: Action query time = 3.354 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9491 +t=298: Selected seed 195 with value = 0.9491 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=10--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 10 +# successes: 2 (20.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 5.545 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4890 +t=10: Selected seed 195 with value = 0.4890 +Query 1/1: Action query time = 6.310 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6068 +t=26: Selected seed 195 with value = 0.6068 +Query 1/1: Action query time = 6.410 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7034 +t=42: Selected seed 195 with value = 0.7034 +Query 1/1: Action query time = 6.085 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8251 +t=58: Selected seed 195 with value = 0.8251 +Query 1/1: Action query time = 6.657 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9581 +t=74: Selected seed 195 with value = 0.9581 +Query 1/1: Action query time = 5.780 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9919 +t=90: Selected seed 195 with value = 0.9919 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=11--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 3 (27.3%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 4.548 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5492 +t=10: Selected seed 195 with value = 0.5492 +Query 1/1: Action query time = 5.636 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6306 +t=26: Selected seed 195 with value = 0.6306 +Query 1/1: Action query time = 5.327 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7603 +t=42: Selected seed 195 with value = 0.7603 +Query 1/1: Action query time = 4.616 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9516 +t=58: Selected seed 195 with value = 0.9516 +Query 1/1: Action query time = 4.629 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9538 +t=74: Selected seed 195 with value = 0.9538 +Query 1/1: Action query time = 3.868 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=90: Selected seed 195 with value = 0.9886 +Query 1/1: Action query time = 4.442 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9938 +t=106: Selected seed 195 with value = 0.9938 +Query 1/1: Action query time = 3.656 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9907 +t=122: Selected seed 195 with value = 0.9907 +Query 1/1: Action query time = 3.557 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9647 +t=138: Selected seed 195 with value = 0.9647 +Query 1/1: Action query time = 4.121 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9471 +t=154: Selected seed 195 with value = 0.9471 +Query 1/1: Action query time = 4.233 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9287 +t=170: Selected seed 195 with value = 0.9287 +Query 1/1: Action query time = 4.791 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9638 +t=186: Selected seed 195 with value = 0.9638 +Query 1/1: Action query time = 4.477 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9625 +t=202: Selected seed 195 with value = 0.9625 +Query 1/1: Action query time = 5.188 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9494 +t=218: Selected seed 195 with value = 0.9494 +Query 1/1: Action query time = 2.583 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9344 +t=234: Selected seed 195 with value = 0.9344 +Query 1/1: Action query time = 3.547 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9199 +t=250: Selected seed 195 with value = 0.9199 +Query 1/1: Action query time = 4.145 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9361 +t=266: Selected seed 195 with value = 0.9361 +Query 1/1: Action query time = 3.222 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9465 +t=282: Selected seed 195 with value = 0.9465 +Query 1/1: Action query time = 3.481 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9505 +t=298: Selected seed 195 with value = 0.9505 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t1_s3/2026_08_03-01_16_48--with_future_img--episode=12--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 12 +# successes: 3 (25.0%) +Current task success rate: 0.25 +Current total success rate: 0.25 +Final results: +Total episodes: 12 +Total successes: 3 +Overall success rate: 0.2500 (25.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_48--t7dc800x50_t2_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_48--t7dc800x50_t2_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..91ef7317c9176a149376d1e922356f3a7b0b0a4a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_48--t7dc800x50_t2_s3.txt @@ -0,0 +1,1008 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_demochan800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7dc800x50_t2_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,7,11,15,19,23,27,31,35,39,43,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 10.265 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5693 +t=10: Selected seed 195 with value = 0.5693 +Query 1/1: Action query time = 6.992 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6822 +t=26: Selected seed 195 with value = 0.6822 +Query 1/1: Action query time = 5.680 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7277 +t=42: Selected seed 195 with value = 0.7277 +Query 1/1: Action query time = 5.800 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8709 +t=58: Selected seed 195 with value = 0.8709 +Query 1/1: Action query time = 5.719 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8791 +t=74: Selected seed 195 with value = 0.8791 +Query 1/1: Action query time = 4.592 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8873 +t=90: Selected seed 195 with value = 0.8873 +Query 1/1: Action query time = 7.053 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9046 +t=106: Selected seed 195 with value = 0.9046 +Query 1/1: Action query time = 4.445 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9212 +t=122: Selected seed 195 with value = 0.9212 +Query 1/1: Action query time = 6.683 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9360 +t=138: Selected seed 195 with value = 0.9360 +Query 1/1: Action query time = 5.969 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9419 +t=154: Selected seed 195 with value = 0.9419 +Query 1/1: Action query time = 4.880 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9340 +t=170: Selected seed 195 with value = 0.9340 +Query 1/1: Action query time = 6.072 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9333 +t=186: Selected seed 195 with value = 0.9333 +Query 1/1: Action query time = 5.832 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9428 +t=202: Selected seed 195 with value = 0.9428 +Query 1/1: Action query time = 5.760 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9247 +t=218: Selected seed 195 with value = 0.9247 +Query 1/1: Action query time = 6.332 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9043 +t=234: Selected seed 195 with value = 0.9043 +Query 1/1: Action query time = 6.548 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8334 +t=250: Selected seed 195 with value = 0.8334 +Query 1/1: Action query time = 5.087 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8257 +t=266: Selected seed 195 with value = 0.8257 +Query 1/1: Action query time = 4.313 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8237 +t=282: Selected seed 195 with value = 0.8237 +Query 1/1: Action query time = 3.558 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8251 +t=298: Selected seed 195 with value = 0.8251 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 6.765 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5767 +t=10: Selected seed 195 with value = 0.5767 +Query 1/1: Action query time = 6.447 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7196 +t=26: Selected seed 195 with value = 0.7196 +Query 1/1: Action query time = 6.921 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8936 +t=42: Selected seed 195 with value = 0.8936 +Query 1/1: Action query time = 6.303 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9261 +t=58: Selected seed 195 with value = 0.9261 +Query 1/1: Action query time = 4.121 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8891 +t=74: Selected seed 195 with value = 0.8891 +Query 1/1: Action query time = 6.017 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8837 +t=90: Selected seed 195 with value = 0.8837 +Query 1/1: Action query time = 6.658 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9069 +t=106: Selected seed 195 with value = 0.9069 +Query 1/1: Action query time = 7.156 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8886 +t=122: Selected seed 195 with value = 0.8886 +Query 1/1: Action query time = 6.087 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8966 +t=138: Selected seed 195 with value = 0.8966 +Query 1/1: Action query time = 7.283 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9083 +t=154: Selected seed 195 with value = 0.9083 +Query 1/1: Action query time = 6.837 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8840 +t=170: Selected seed 195 with value = 0.8840 +Query 1/1: Action query time = 5.653 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9167 +t=186: Selected seed 195 with value = 0.9167 +Query 1/1: Action query time = 5.652 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9125 +t=202: Selected seed 195 with value = 0.9125 +Query 1/1: Action query time = 5.676 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8046 +t=218: Selected seed 195 with value = 0.8046 +Query 1/1: Action query time = 6.838 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8268 +t=234: Selected seed 195 with value = 0.8268 +Query 1/1: Action query time = 4.969 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8359 +t=250: Selected seed 195 with value = 0.8359 +Query 1/1: Action query time = 5.192 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7898 +t=266: Selected seed 195 with value = 0.7898 +Query 1/1: Action query time = 3.269 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7674 +t=282: Selected seed 195 with value = 0.7674 +Query 1/1: Action query time = 2.849 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7658 +t=298: Selected seed 195 with value = 0.7658 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.381 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5890 +t=10: Selected seed 195 with value = 0.5890 +Query 1/1: Action query time = 6.364 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7047 +t=26: Selected seed 195 with value = 0.7047 +Query 1/1: Action query time = 6.334 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8813 +t=42: Selected seed 195 with value = 0.8813 +Query 1/1: Action query time = 5.856 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9822 +t=58: Selected seed 195 with value = 0.9822 +Query 1/1: Action query time = 6.102 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.156 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.002 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.973 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 8.992 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.208 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.087 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=170: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 5.082 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9940 +t=186: Selected seed 195 with value = 0.9940 +Query 1/1: Action query time = 6.217 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9834 +t=202: Selected seed 195 with value = 0.9834 +Query 1/1: Action query time = 5.301 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9531 +t=218: Selected seed 195 with value = 0.9531 +Query 1/1: Action query time = 5.717 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9426 +t=234: Selected seed 195 with value = 0.9426 +Query 1/1: Action query time = 6.140 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9255 +t=250: Selected seed 195 with value = 0.9255 +Query 1/1: Action query time = 5.175 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9131 +t=266: Selected seed 195 with value = 0.9131 +Query 1/1: Action query time = 4.654 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9106 +t=282: Selected seed 195 with value = 0.9106 +Query 1/1: Action query time = 2.059 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8999 +t=298: Selected seed 195 with value = 0.8999 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=3--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 5.242 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5657 +t=10: Selected seed 195 with value = 0.5657 +Query 1/1: Action query time = 5.712 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6380 +t=26: Selected seed 195 with value = 0.6380 +Query 1/1: Action query time = 5.751 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7680 +t=42: Selected seed 195 with value = 0.7680 +Query 1/1: Action query time = 6.151 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8899 +t=58: Selected seed 195 with value = 0.8899 +Query 1/1: Action query time = 5.415 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9542 +t=74: Selected seed 195 with value = 0.9542 +Query 1/1: Action query time = 3.938 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9709 +t=90: Selected seed 195 with value = 0.9709 +Query 1/1: Action query time = 5.732 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9743 +t=106: Selected seed 195 with value = 0.9743 +Query 1/1: Action query time = 6.299 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9650 +t=122: Selected seed 195 with value = 0.9650 +Query 1/1: Action query time = 5.947 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9661 +t=138: Selected seed 195 with value = 0.9661 +Query 1/1: Action query time = 5.952 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9665 +t=154: Selected seed 195 with value = 0.9665 +Query 1/1: Action query time = 5.661 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9847 +t=170: Selected seed 195 with value = 0.9847 +Query 1/1: Action query time = 7.683 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9815 +t=186: Selected seed 195 with value = 0.9815 +Query 1/1: Action query time = 7.081 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=202: Selected seed 195 with value = 0.9809 +Query 1/1: Action query time = 5.366 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9781 +t=218: Selected seed 195 with value = 0.9781 +Query 1/1: Action query time = 6.041 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9846 +t=234: Selected seed 195 with value = 0.9846 +Query 1/1: Action query time = 5.199 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8784 +t=250: Selected seed 195 with value = 0.8784 +Query 1/1: Action query time = 4.490 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=266: Selected seed 195 with value = 0.9864 +Query 1/1: Action query time = 5.239 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8771 +t=282: Selected seed 195 with value = 0.8771 +Query 1/1: Action query time = 1.829 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=298: Selected seed 195 with value = 0.9864 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=4--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 5... +Query 1/1: Action query time = 5.694 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5824 +t=10: Selected seed 195 with value = 0.5824 +Query 1/1: Action query time = 5.134 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6652 +t=26: Selected seed 195 with value = 0.6652 +Query 1/1: Action query time = 6.593 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8879 +t=42: Selected seed 195 with value = 0.8879 +Query 1/1: Action query time = 5.354 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8212 +t=58: Selected seed 195 with value = 0.8212 +Query 1/1: Action query time = 5.695 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8218 +t=74: Selected seed 195 with value = 0.8218 +Query 1/1: Action query time = 7.063 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8892 +t=90: Selected seed 195 with value = 0.8892 +Query 1/1: Action query time = 3.428 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9288 +t=106: Selected seed 195 with value = 0.9288 +Query 1/1: Action query time = 5.714 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9456 +t=122: Selected seed 195 with value = 0.9456 +Query 1/1: Action query time = 6.134 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9448 +t=138: Selected seed 195 with value = 0.9448 +Query 1/1: Action query time = 6.010 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9293 +t=154: Selected seed 195 with value = 0.9293 +Query 1/1: Action query time = 5.547 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8625 +t=170: Selected seed 195 with value = 0.8625 +Query 1/1: Action query time = 7.318 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8309 +t=186: Selected seed 195 with value = 0.8309 +Query 1/1: Action query time = 5.354 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9379 +t=202: Selected seed 195 with value = 0.9379 +Query 1/1: Action query time = 6.761 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9489 +t=218: Selected seed 195 with value = 0.9489 +Query 1/1: Action query time = 7.440 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9606 +t=234: Selected seed 195 with value = 0.9606 +Query 1/1: Action query time = 4.690 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9718 +t=250: Selected seed 195 with value = 0.9718 +Query 1/1: Action query time = 3.451 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9711 +t=266: Selected seed 195 with value = 0.9711 +Query 1/1: Action query time = 4.547 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9717 +t=282: Selected seed 195 with value = 0.9717 +Query 1/1: Action query time = 2.984 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8027 +t=298: Selected seed 195 with value = 0.8027 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=5--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 5 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 6... +Query 1/1: Action query time = 6.810 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5756 +t=10: Selected seed 195 with value = 0.5756 +Query 1/1: Action query time = 7.111 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6225 +t=26: Selected seed 195 with value = 0.6225 +Query 1/1: Action query time = 5.421 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7439 +t=42: Selected seed 195 with value = 0.7439 +Query 1/1: Action query time = 4.747 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8209 +t=58: Selected seed 195 with value = 0.8209 +Query 1/1: Action query time = 5.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9097 +t=74: Selected seed 195 with value = 0.9097 +Query 1/1: Action query time = 5.056 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9286 +t=90: Selected seed 195 with value = 0.9286 +Query 1/1: Action query time = 7.590 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9428 +t=106: Selected seed 195 with value = 0.9428 +Query 1/1: Action query time = 5.931 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9450 +t=122: Selected seed 195 with value = 0.9450 +Query 1/1: Action query time = 4.800 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9425 +t=138: Selected seed 195 with value = 0.9425 +Query 1/1: Action query time = 4.866 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9245 +t=154: Selected seed 195 with value = 0.9245 +Query 1/1: Action query time = 6.559 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8773 +t=170: Selected seed 195 with value = 0.8773 +Query 1/1: Action query time = 5.272 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9332 +t=186: Selected seed 195 with value = 0.9332 +Query 1/1: Action query time = 5.796 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9430 +t=202: Selected seed 195 with value = 0.9430 +Query 1/1: Action query time = 6.361 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9532 +t=218: Selected seed 195 with value = 0.9532 +Query 1/1: Action query time = 6.943 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9660 +t=234: Selected seed 195 with value = 0.9660 +Query 1/1: Action query time = 4.700 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9655 +t=250: Selected seed 195 with value = 0.9655 +Query 1/1: Action query time = 4.402 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9714 +t=266: Selected seed 195 with value = 0.9714 +Query 1/1: Action query time = 4.803 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8511 +t=282: Selected seed 195 with value = 0.8511 +Query 1/1: Action query time = 4.551 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8273 +t=298: Selected seed 195 with value = 0.8273 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=6--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 6 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 7... +Query 1/1: Action query time = 6.021 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5818 +t=10: Selected seed 195 with value = 0.5818 +Query 1/1: Action query time = 6.504 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7142 +t=26: Selected seed 195 with value = 0.7142 +Query 1/1: Action query time = 5.538 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9085 +t=42: Selected seed 195 with value = 0.9085 +Query 1/1: Action query time = 4.760 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9904 +t=58: Selected seed 195 with value = 0.9904 +Query 1/1: Action query time = 5.590 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.092 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.726 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.169 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.975 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.432 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.019 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=170: Selected seed 195 with value = 0.9989 +Query 1/1: Action query time = 6.367 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=186: Selected seed 195 with value = 0.9966 +Query 1/1: Action query time = 5.203 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9905 +t=202: Selected seed 195 with value = 0.9905 +Query 1/1: Action query time = 4.766 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9763 +t=218: Selected seed 195 with value = 0.9763 +Query 1/1: Action query time = 4.434 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9536 +t=234: Selected seed 195 with value = 0.9536 +Query 1/1: Action query time = 5.519 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9608 +t=250: Selected seed 195 with value = 0.9608 +Query 1/1: Action query time = 6.789 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9208 +t=266: Selected seed 195 with value = 0.9208 +Query 1/1: Action query time = 6.525 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9113 +t=282: Selected seed 195 with value = 0.9113 +Query 1/1: Action query time = 7.030 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9072 +t=298: Selected seed 195 with value = 0.9072 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=7--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 7 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 8... +Query 1/1: Action query time = 7.098 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5663 +t=10: Selected seed 195 with value = 0.5663 +Query 1/1: Action query time = 5.895 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6252 +t=26: Selected seed 195 with value = 0.6252 +Query 1/1: Action query time = 4.354 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7283 +t=42: Selected seed 195 with value = 0.7283 +Query 1/1: Action query time = 5.515 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8485 +t=58: Selected seed 195 with value = 0.8485 +Query 1/1: Action query time = 5.849 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8512 +t=74: Selected seed 195 with value = 0.8512 +Query 1/1: Action query time = 5.437 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8997 +t=90: Selected seed 195 with value = 0.8997 +Query 1/1: Action query time = 7.579 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9199 +t=106: Selected seed 195 with value = 0.9199 +Query 1/1: Action query time = 6.660 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8891 +t=122: Selected seed 195 with value = 0.8891 +Query 1/1: Action query time = 6.786 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8942 +t=138: Selected seed 195 with value = 0.8942 +Query 1/1: Action query time = 6.429 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8914 +t=154: Selected seed 195 with value = 0.8914 +Query 1/1: Action query time = 5.890 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9023 +t=170: Selected seed 195 with value = 0.9023 +Query 1/1: Action query time = 6.127 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8977 +t=186: Selected seed 195 with value = 0.8977 +Query 1/1: Action query time = 6.648 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8932 +t=202: Selected seed 195 with value = 0.8932 +Query 1/1: Action query time = 6.138 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8910 +t=218: Selected seed 195 with value = 0.8910 +Query 1/1: Action query time = 4.661 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9029 +t=234: Selected seed 195 with value = 0.9029 +Query 1/1: Action query time = 5.244 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9066 +t=250: Selected seed 195 with value = 0.9066 +Query 1/1: Action query time = 3.808 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9332 +t=266: Selected seed 195 with value = 0.9332 +Query 1/1: Action query time = 4.592 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9423 +t=282: Selected seed 195 with value = 0.9423 +Query 1/1: Action query time = 4.780 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9482 +t=298: Selected seed 195 with value = 0.9482 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=8--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 8 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 9... +Query 1/1: Action query time = 7.182 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5676 +t=10: Selected seed 195 with value = 0.5676 +Query 1/1: Action query time = 5.841 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6525 +t=26: Selected seed 195 with value = 0.6525 +Query 1/1: Action query time = 5.994 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7366 +t=42: Selected seed 195 with value = 0.7366 +Query 1/1: Action query time = 5.776 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8981 +t=58: Selected seed 195 with value = 0.8981 +Query 1/1: Action query time = 5.303 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8893 +t=74: Selected seed 195 with value = 0.8893 +Query 1/1: Action query time = 5.648 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8790 +t=90: Selected seed 195 with value = 0.8790 +Query 1/1: Action query time = 6.678 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8942 +t=106: Selected seed 195 with value = 0.8942 +Query 1/1: Action query time = 5.561 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9176 +t=122: Selected seed 195 with value = 0.9176 +Query 1/1: Action query time = 3.831 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9337 +t=138: Selected seed 195 with value = 0.9337 +Query 1/1: Action query time = 4.734 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9305 +t=154: Selected seed 195 with value = 0.9305 +Query 1/1: Action query time = 5.349 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9153 +t=170: Selected seed 195 with value = 0.9153 +Query 1/1: Action query time = 5.918 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9024 +t=186: Selected seed 195 with value = 0.9024 +Query 1/1: Action query time = 4.804 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8691 +t=202: Selected seed 195 with value = 0.8691 +Query 1/1: Action query time = 5.910 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8545 +t=218: Selected seed 195 with value = 0.8545 +Query 1/1: Action query time = 4.600 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8419 +t=234: Selected seed 195 with value = 0.8419 +Query 1/1: Action query time = 4.416 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8470 +t=250: Selected seed 195 with value = 0.8470 +Query 1/1: Action query time = 5.231 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8653 +t=266: Selected seed 195 with value = 0.8653 +Query 1/1: Action query time = 4.593 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8705 +t=282: Selected seed 195 with value = 0.8705 +Query 1/1: Action query time = 4.533 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8787 +t=298: Selected seed 195 with value = 0.8787 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=9--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 9 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 10... +Query 1/1: Action query time = 4.602 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5290 +t=10: Selected seed 195 with value = 0.5290 +Query 1/1: Action query time = 5.735 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7191 +t=26: Selected seed 195 with value = 0.7191 +Query 1/1: Action query time = 5.039 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8605 +t=42: Selected seed 195 with value = 0.8605 +Query 1/1: Action query time = 5.285 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9013 +t=58: Selected seed 195 with value = 0.9013 +Query 1/1: Action query time = 5.758 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8928 +t=74: Selected seed 195 with value = 0.8928 +Query 1/1: Action query time = 3.106 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8805 +t=90: Selected seed 195 with value = 0.8805 +Query 1/1: Action query time = 4.495 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8827 +t=106: Selected seed 195 with value = 0.8827 +Query 1/1: Action query time = 4.942 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8918 +t=122: Selected seed 195 with value = 0.8918 +Query 1/1: Action query time = 6.454 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9042 +t=138: Selected seed 195 with value = 0.9042 +Query 1/1: Action query time = 5.944 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9140 +t=154: Selected seed 195 with value = 0.9140 +Query 1/1: Action query time = 4.818 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9199 +t=170: Selected seed 195 with value = 0.9199 +Query 1/1: Action query time = 4.793 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9041 +t=186: Selected seed 195 with value = 0.9041 +Query 1/1: Action query time = 4.105 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9065 +t=202: Selected seed 195 with value = 0.9065 +Query 1/1: Action query time = 4.690 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9118 +t=218: Selected seed 195 with value = 0.9118 +Query 1/1: Action query time = 4.536 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9380 +t=234: Selected seed 195 with value = 0.9380 +Query 1/1: Action query time = 4.187 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9227 +t=250: Selected seed 195 with value = 0.9227 +Query 1/1: Action query time = 4.956 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8244 +t=266: Selected seed 195 with value = 0.8244 +Query 1/1: Action query time = 5.676 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8315 +t=282: Selected seed 195 with value = 0.8315 +Query 1/1: Action query time = 3.737 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8343 +t=298: Selected seed 195 with value = 0.8343 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=10--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 10 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 11... +Query 1/1: Action query time = 3.556 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5680 +t=10: Selected seed 195 with value = 0.5680 +Query 1/1: Action query time = 5.043 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6767 +t=26: Selected seed 195 with value = 0.6767 +Query 1/1: Action query time = 4.535 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7772 +t=42: Selected seed 195 with value = 0.7772 +Query 1/1: Action query time = 5.228 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8934 +t=58: Selected seed 195 with value = 0.8934 +Query 1/1: Action query time = 5.116 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9173 +t=74: Selected seed 195 with value = 0.9173 +Query 1/1: Action query time = 3.747 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9426 +t=90: Selected seed 195 with value = 0.9426 +Query 1/1: Action query time = 4.061 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9686 +t=106: Selected seed 195 with value = 0.9686 +Query 1/1: Action query time = 4.373 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9712 +t=122: Selected seed 195 with value = 0.9712 +Query 1/1: Action query time = 4.551 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9715 +t=138: Selected seed 195 with value = 0.9715 +Query 1/1: Action query time = 4.347 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9785 +t=154: Selected seed 195 with value = 0.9785 +Query 1/1: Action query time = 4.847 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=170: Selected seed 195 with value = 0.9903 +Query 1/1: Action query time = 4.700 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9925 +t=186: Selected seed 195 with value = 0.9925 +Query 1/1: Action query time = 3.756 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9902 +t=202: Selected seed 195 with value = 0.9902 +Query 1/1: Action query time = 2.664 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9857 +t=218: Selected seed 195 with value = 0.9857 +Query 1/1: Action query time = 3.550 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9820 +t=234: Selected seed 195 with value = 0.9820 +Query 1/1: Action query time = 4.331 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9791 +t=250: Selected seed 195 with value = 0.9791 +Query 1/1: Action query time = 4.018 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9784 +t=266: Selected seed 195 with value = 0.9784 +Query 1/1: Action query time = 3.639 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9758 +t=282: Selected seed 195 with value = 0.9758 +Query 1/1: Action query time = 3.221 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9747 +t=298: Selected seed 195 with value = 0.9747 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=11--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 11 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 12... +Query 1/1: Action query time = 3.231 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5791 +t=10: Selected seed 195 with value = 0.5791 +Query 1/1: Action query time = 3.630 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7068 +t=26: Selected seed 195 with value = 0.7068 +Query 1/1: Action query time = 3.454 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8885 +t=42: Selected seed 195 with value = 0.8885 +Query 1/1: Action query time = 4.171 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9634 +t=58: Selected seed 195 with value = 0.9634 +Query 1/1: Action query time = 4.263 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=74: Selected seed 195 with value = 0.9926 +Query 1/1: Action query time = 3.743 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.267 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=106: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 2.909 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.902 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.018 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=154: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 3.279 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9934 +t=170: Selected seed 195 with value = 0.9934 +Query 1/1: Action query time = 3.656 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9898 +t=186: Selected seed 195 with value = 0.9898 +Query 1/1: Action query time = 3.212 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9831 +t=202: Selected seed 195 with value = 0.9831 +Query 1/1: Action query time = 2.895 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9614 +t=218: Selected seed 195 with value = 0.9614 +Query 1/1: Action query time = 3.136 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9100 +t=234: Selected seed 195 with value = 0.9100 +Query 1/1: Action query time = 2.250 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9158 +t=250: Selected seed 195 with value = 0.9158 +Query 1/1: Action query time = 2.347 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9253 +t=266: Selected seed 195 with value = 0.9253 +Query 1/1: Action query time = 2.326 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9075 +t=282: Selected seed 195 with value = 0.9075 +Query 1/1: Action query time = 3.497 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9158 +t=298: Selected seed 195 with value = 0.9158 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=12--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 12 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 12 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_48--t7rc800x50_t2_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_48--t7rc800x50_t2_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..0c5f2627fba6ec26fbf1cf7b49ff7f1eb7e66b1e --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_48--t7rc800x50_t2_s3.txt @@ -0,0 +1,956 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t2_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,7,11,15,19,23,27,31,35,39,43,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.404 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5677 +t=10: Selected seed 195 with value = 0.5677 +Query 1/1: Action query time = 5.816 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6627 +t=26: Selected seed 195 with value = 0.6627 +Query 1/1: Action query time = 5.082 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6469 +t=42: Selected seed 195 with value = 0.6469 +Query 1/1: Action query time = 6.630 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8252 +t=58: Selected seed 195 with value = 0.8252 +Query 1/1: Action query time = 7.593 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7447 +t=74: Selected seed 195 with value = 0.7447 +Query 1/1: Action query time = 5.612 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7645 +t=90: Selected seed 195 with value = 0.7645 +Query 1/1: Action query time = 4.791 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9090 +t=106: Selected seed 195 with value = 0.9090 +Query 1/1: Action query time = 6.287 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9243 +t=122: Selected seed 195 with value = 0.9243 +Query 1/1: Action query time = 7.453 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9361 +t=138: Selected seed 195 with value = 0.9361 +Query 1/1: Action query time = 6.910 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9329 +t=154: Selected seed 195 with value = 0.9329 +Query 1/1: Action query time = 6.636 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9032 +t=170: Selected seed 195 with value = 0.9032 +Query 1/1: Action query time = 6.007 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9147 +t=186: Selected seed 195 with value = 0.9147 +Query 1/1: Action query time = 7.116 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9370 +t=202: Selected seed 195 with value = 0.9370 +Query 1/1: Action query time = 6.121 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8447 +t=218: Selected seed 195 with value = 0.8447 +Query 1/1: Action query time = 7.022 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8779 +t=234: Selected seed 195 with value = 0.8779 +Query 1/1: Action query time = 5.521 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8909 +t=250: Selected seed 195 with value = 0.8909 +Query 1/1: Action query time = 5.280 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8726 +t=266: Selected seed 195 with value = 0.8726 +Query 1/1: Action query time = 3.927 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8739 +t=282: Selected seed 195 with value = 0.8739 +Query 1/1: Action query time = 3.403 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8514 +t=298: Selected seed 195 with value = 0.8514 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=1--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 6.332 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5767 +t=10: Selected seed 195 with value = 0.5767 +Query 1/1: Action query time = 7.030 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6946 +t=26: Selected seed 195 with value = 0.6946 +Query 1/1: Action query time = 5.419 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6931 +t=42: Selected seed 195 with value = 0.6931 +Query 1/1: Action query time = 6.523 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9100 +t=58: Selected seed 195 with value = 0.9100 +Query 1/1: Action query time = 5.050 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7684 +t=74: Selected seed 195 with value = 0.7684 +Query 1/1: Action query time = 6.601 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8687 +t=90: Selected seed 195 with value = 0.8687 +Query 1/1: Action query time = 5.736 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8277 +t=106: Selected seed 195 with value = 0.8277 +Query 1/1: Action query time = 5.199 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8571 +t=122: Selected seed 195 with value = 0.8571 +Query 1/1: Action query time = 5.449 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8346 +t=138: Selected seed 195 with value = 0.8346 +Query 1/1: Action query time = 4.867 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8142 +t=154: Selected seed 195 with value = 0.8142 +Query 1/1: Action query time = 6.498 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8320 +t=170: Selected seed 195 with value = 0.8320 +Query 1/1: Action query time = 7.306 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8341 +t=186: Selected seed 195 with value = 0.8341 +Query 1/1: Action query time = 5.635 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8525 +t=202: Selected seed 195 with value = 0.8525 +Query 1/1: Action query time = 4.463 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8654 +t=218: Selected seed 195 with value = 0.8654 +Query 1/1: Action query time = 5.776 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8537 +t=234: Selected seed 195 with value = 0.8537 +Query 1/1: Action query time = 6.217 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8492 +t=250: Selected seed 195 with value = 0.8492 +Query 1/1: Action query time = 4.941 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8533 +t=266: Selected seed 195 with value = 0.8533 +Query 1/1: Action query time = 5.249 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8785 +t=282: Selected seed 195 with value = 0.8785 +Query 1/1: Action query time = 4.921 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8968 +t=298: Selected seed 195 with value = 0.8968 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=2--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.438 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5837 +t=10: Selected seed 195 with value = 0.5837 +Query 1/1: Action query time = 5.629 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7089 +t=26: Selected seed 195 with value = 0.7089 +Query 1/1: Action query time = 5.461 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8443 +t=42: Selected seed 195 with value = 0.8443 +Query 1/1: Action query time = 5.355 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9163 +t=58: Selected seed 195 with value = 0.9163 +Query 1/1: Action query time = 6.216 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9862 +t=74: Selected seed 195 with value = 0.9862 +Query 1/1: Action query time = 6.153 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 1 (33.3%) + +Task: put the wine bottle on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 6.771 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5670 +t=10: Selected seed 195 with value = 0.5670 +Query 1/1: Action query time = 6.653 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6190 +t=26: Selected seed 195 with value = 0.6190 +Query 1/1: Action query time = 6.944 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7350 +t=42: Selected seed 195 with value = 0.7350 +Query 1/1: Action query time = 5.735 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8032 +t=58: Selected seed 195 with value = 0.8032 +Query 1/1: Action query time = 7.849 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9419 +t=74: Selected seed 195 with value = 0.9419 +Query 1/1: Action query time = 6.177 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.127 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.947 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=122: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 4.739 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.071 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.746 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.947 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.390 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.406 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.022 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.518 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.182 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.822 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.992 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=4--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 4 +# successes: 1 (25.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 5... +Query 1/1: Action query time = 6.580 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5800 +t=10: Selected seed 195 with value = 0.5800 +Query 1/1: Action query time = 6.232 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6999 +t=26: Selected seed 195 with value = 0.6999 +Query 1/1: Action query time = 6.278 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8791 +t=42: Selected seed 195 with value = 0.8791 +Query 1/1: Action query time = 5.452 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7720 +t=58: Selected seed 195 with value = 0.7720 +Query 1/1: Action query time = 6.832 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8950 +t=74: Selected seed 195 with value = 0.8950 +Query 1/1: Action query time = 8.271 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8086 +t=90: Selected seed 195 with value = 0.8086 +Query 1/1: Action query time = 7.731 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8580 +t=106: Selected seed 195 with value = 0.8580 +Query 1/1: Action query time = 6.519 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9155 +t=122: Selected seed 195 with value = 0.9155 +Query 1/1: Action query time = 5.225 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9210 +t=138: Selected seed 195 with value = 0.9210 +Query 1/1: Action query time = 6.923 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9178 +t=154: Selected seed 195 with value = 0.9178 +Query 1/1: Action query time = 5.506 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9159 +t=170: Selected seed 195 with value = 0.9159 +Query 1/1: Action query time = 3.688 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9400 +t=186: Selected seed 195 with value = 0.9400 +Query 1/1: Action query time = 4.633 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9233 +t=202: Selected seed 195 with value = 0.9233 +Query 1/1: Action query time = 3.254 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9299 +t=218: Selected seed 195 with value = 0.9299 +Query 1/1: Action query time = 2.290 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8664 +t=234: Selected seed 195 with value = 0.8664 +Query 1/1: Action query time = 3.535 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8753 +t=250: Selected seed 195 with value = 0.8753 +Query 1/1: Action query time = 5.136 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9067 +t=266: Selected seed 195 with value = 0.9067 +Query 1/1: Action query time = 5.665 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8856 +t=282: Selected seed 195 with value = 0.8856 +Query 1/1: Action query time = 4.281 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8560 +t=298: Selected seed 195 with value = 0.8560 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=5--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 5 +# successes: 1 (20.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 6... +Query 1/1: Action query time = 6.148 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5712 +t=10: Selected seed 195 with value = 0.5712 +Query 1/1: Action query time = 6.320 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5852 +t=26: Selected seed 195 with value = 0.5852 +Query 1/1: Action query time = 5.823 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7072 +t=42: Selected seed 195 with value = 0.7072 +Query 1/1: Action query time = 5.438 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8263 +t=58: Selected seed 195 with value = 0.8263 +Query 1/1: Action query time = 5.416 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7910 +t=74: Selected seed 195 with value = 0.7910 +Query 1/1: Action query time = 5.130 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8351 +t=90: Selected seed 195 with value = 0.8351 +Query 1/1: Action query time = 6.295 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8836 +t=106: Selected seed 195 with value = 0.8836 +Query 1/1: Action query time = 4.095 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9086 +t=122: Selected seed 195 with value = 0.9086 +Query 1/1: Action query time = 6.570 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9037 +t=138: Selected seed 195 with value = 0.9037 +Query 1/1: Action query time = 5.542 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9116 +t=154: Selected seed 195 with value = 0.9116 +Query 1/1: Action query time = 5.779 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9031 +t=170: Selected seed 195 with value = 0.9031 +Query 1/1: Action query time = 4.940 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8976 +t=186: Selected seed 195 with value = 0.8976 +Query 1/1: Action query time = 3.956 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8029 +t=202: Selected seed 195 with value = 0.8029 +Query 1/1: Action query time = 4.189 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8384 +t=218: Selected seed 195 with value = 0.8384 +Query 1/1: Action query time = 2.809 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8637 +t=234: Selected seed 195 with value = 0.8637 +Query 1/1: Action query time = 3.468 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8827 +t=250: Selected seed 195 with value = 0.8827 +Query 1/1: Action query time = 5.809 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8820 +t=266: Selected seed 195 with value = 0.8820 +Query 1/1: Action query time = 5.440 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8604 +t=282: Selected seed 195 with value = 0.8604 +Query 1/1: Action query time = 6.382 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8491 +t=298: Selected seed 195 with value = 0.8491 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=6--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 6 +# successes: 1 (16.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 7... +Query 1/1: Action query time = 5.761 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=10: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 4.692 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7261 +t=26: Selected seed 195 with value = 0.7261 +Query 1/1: Action query time = 4.762 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9049 +t=42: Selected seed 195 with value = 0.9049 +Query 1/1: Action query time = 4.720 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8451 +t=58: Selected seed 195 with value = 0.8451 +Query 1/1: Action query time = 4.906 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8534 +t=74: Selected seed 195 with value = 0.8534 +Query 1/1: Action query time = 4.670 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8556 +t=90: Selected seed 195 with value = 0.8556 +Query 1/1: Action query time = 4.638 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8586 +t=106: Selected seed 195 with value = 0.8586 +Query 1/1: Action query time = 3.600 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8600 +t=122: Selected seed 195 with value = 0.8600 +Query 1/1: Action query time = 4.823 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9048 +t=138: Selected seed 195 with value = 0.9048 +Query 1/1: Action query time = 4.361 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9916 +t=154: Selected seed 195 with value = 0.9916 +Query 1/1: Action query time = 4.591 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.169 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9803 +t=186: Selected seed 195 with value = 0.9803 +Query 1/1: Action query time = 4.554 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=202: Selected seed 195 with value = 0.9966 +Query 1/1: Action query time = 4.538 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9919 +t=218: Selected seed 195 with value = 0.9919 +Query 1/1: Action query time = 6.901 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.964 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9900 +t=250: Selected seed 195 with value = 0.9900 +Query 1/1: Action query time = 3.305 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.364 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.812 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=7--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 7 +# successes: 1 (14.3%) + +Task: put the wine bottle on top of the cabinet +Starting episode 8... +Query 1/1: Action query time = 3.381 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5612 +t=10: Selected seed 195 with value = 0.5612 +Query 1/1: Action query time = 3.099 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6388 +t=26: Selected seed 195 with value = 0.6388 +Query 1/1: Action query time = 4.213 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7057 +t=42: Selected seed 195 with value = 0.7057 +Query 1/1: Action query time = 4.477 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8844 +t=58: Selected seed 195 with value = 0.8844 +Query 1/1: Action query time = 5.137 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7660 +t=74: Selected seed 195 with value = 0.7660 +Query 1/1: Action query time = 3.663 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7920 +t=90: Selected seed 195 with value = 0.7920 +Query 1/1: Action query time = 5.730 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8040 +t=106: Selected seed 195 with value = 0.8040 +Query 1/1: Action query time = 4.970 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9239 +t=122: Selected seed 195 with value = 0.9239 +Query 1/1: Action query time = 2.725 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9417 +t=138: Selected seed 195 with value = 0.9417 +Query 1/1: Action query time = 3.745 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9283 +t=154: Selected seed 195 with value = 0.9283 +Query 1/1: Action query time = 5.147 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9266 +t=170: Selected seed 195 with value = 0.9266 +Query 1/1: Action query time = 3.802 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8992 +t=186: Selected seed 195 with value = 0.8992 +Query 1/1: Action query time = 5.279 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8346 +t=202: Selected seed 195 with value = 0.8346 +Query 1/1: Action query time = 4.723 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8473 +t=218: Selected seed 195 with value = 0.8473 +Query 1/1: Action query time = 3.661 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9051 +t=234: Selected seed 195 with value = 0.9051 +Query 1/1: Action query time = 5.192 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9302 +t=250: Selected seed 195 with value = 0.9302 +Query 1/1: Action query time = 4.739 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9060 +t=266: Selected seed 195 with value = 0.9060 +Query 1/1: Action query time = 3.860 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8620 +t=282: Selected seed 195 with value = 0.8620 +Query 1/1: Action query time = 4.799 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8421 +t=298: Selected seed 195 with value = 0.8421 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=8--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 8 +# successes: 1 (12.5%) + +Task: put the wine bottle on top of the cabinet +Starting episode 9... +Query 1/1: Action query time = 3.667 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5608 +t=10: Selected seed 195 with value = 0.5608 +Query 1/1: Action query time = 4.828 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6727 +t=26: Selected seed 195 with value = 0.6727 +Query 1/1: Action query time = 5.330 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7043 +t=42: Selected seed 195 with value = 0.7043 +Query 1/1: Action query time = 4.740 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8785 +t=58: Selected seed 195 with value = 0.8785 +Query 1/1: Action query time = 4.859 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7564 +t=74: Selected seed 195 with value = 0.7564 +Query 1/1: Action query time = 3.600 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8641 +t=90: Selected seed 195 with value = 0.8641 +Query 1/1: Action query time = 3.957 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9108 +t=106: Selected seed 195 with value = 0.9108 +Query 1/1: Action query time = 4.087 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9357 +t=122: Selected seed 195 with value = 0.9357 +Query 1/1: Action query time = 2.833 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9571 +t=138: Selected seed 195 with value = 0.9571 +Query 1/1: Action query time = 4.019 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9322 +t=154: Selected seed 195 with value = 0.9322 +Query 1/1: Action query time = 2.742 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9085 +t=170: Selected seed 195 with value = 0.9085 +Query 1/1: Action query time = 3.584 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8633 +t=186: Selected seed 195 with value = 0.8633 +Query 1/1: Action query time = 2.394 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7788 +t=202: Selected seed 195 with value = 0.7788 +Query 1/1: Action query time = 3.442 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8596 +t=218: Selected seed 195 with value = 0.8596 +Query 1/1: Action query time = 4.235 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8870 +t=234: Selected seed 195 with value = 0.8870 +Query 1/1: Action query time = 3.821 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8917 +t=250: Selected seed 195 with value = 0.8917 +Query 1/1: Action query time = 3.725 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8668 +t=266: Selected seed 195 with value = 0.8668 +Query 1/1: Action query time = 3.951 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8530 +t=282: Selected seed 195 with value = 0.8530 +Query 1/1: Action query time = 4.034 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8440 +t=298: Selected seed 195 with value = 0.8440 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=9--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 9 +# successes: 1 (11.1%) + +Task: put the wine bottle on top of the cabinet +Starting episode 10... +Query 1/1: Action query time = 3.911 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5271 +t=10: Selected seed 195 with value = 0.5271 +Query 1/1: Action query time = 5.092 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7162 +t=26: Selected seed 195 with value = 0.7162 +Query 1/1: Action query time = 4.194 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6664 +t=42: Selected seed 195 with value = 0.6664 +Query 1/1: Action query time = 5.203 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8450 +t=58: Selected seed 195 with value = 0.8450 +Query 1/1: Action query time = 4.251 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7377 +t=74: Selected seed 195 with value = 0.7377 +Query 1/1: Action query time = 4.407 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8622 +t=90: Selected seed 195 with value = 0.8622 +Query 1/1: Action query time = 4.900 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8947 +t=106: Selected seed 195 with value = 0.8947 +Query 1/1: Action query time = 4.031 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9175 +t=122: Selected seed 195 with value = 0.9175 +Query 1/1: Action query time = 5.364 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9261 +t=138: Selected seed 195 with value = 0.9261 +Query 1/1: Action query time = 3.819 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9142 +t=154: Selected seed 195 with value = 0.9142 +Query 1/1: Action query time = 4.010 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9297 +t=170: Selected seed 195 with value = 0.9297 +Query 1/1: Action query time = 4.737 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9158 +t=186: Selected seed 195 with value = 0.9158 +Query 1/1: Action query time = 2.865 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9273 +t=202: Selected seed 195 with value = 0.9273 +Query 1/1: Action query time = 3.932 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9103 +t=218: Selected seed 195 with value = 0.9103 +Query 1/1: Action query time = 3.351 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8520 +t=234: Selected seed 195 with value = 0.8520 +Query 1/1: Action query time = 3.467 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8794 +t=250: Selected seed 195 with value = 0.8794 +Query 1/1: Action query time = 3.114 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9083 +t=266: Selected seed 195 with value = 0.9083 +Query 1/1: Action query time = 3.151 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8535 +t=282: Selected seed 195 with value = 0.8535 +Query 1/1: Action query time = 2.818 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8272 +t=298: Selected seed 195 with value = 0.8272 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=10--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 10 +# successes: 1 (10.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 11... +Query 1/1: Action query time = 3.178 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5811 +t=10: Selected seed 195 with value = 0.5811 +Query 1/1: Action query time = 5.000 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6935 +t=26: Selected seed 195 with value = 0.6935 +Query 1/1: Action query time = 4.359 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7585 +t=42: Selected seed 195 with value = 0.7585 +Query 1/1: Action query time = 3.825 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8749 +t=58: Selected seed 195 with value = 0.8749 +Query 1/1: Action query time = 4.049 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9500 +t=74: Selected seed 195 with value = 0.9500 +Query 1/1: Action query time = 3.728 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.931 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.466 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.089 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.807 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.849 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.625 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.995 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.833 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.735 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.059 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.420 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.463 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.715 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=11--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 11 +# successes: 1 (9.1%) + +Task: put the wine bottle on top of the cabinet +Starting episode 12... +Query 1/1: Action query time = 3.835 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5742 +t=10: Selected seed 195 with value = 0.5742 +Query 1/1: Action query time = 4.523 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6717 +t=26: Selected seed 195 with value = 0.6717 +Query 1/1: Action query time = 2.775 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7799 +t=42: Selected seed 195 with value = 0.7799 +Query 1/1: Action query time = 3.188 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8175 +t=58: Selected seed 195 with value = 0.8175 +Query 1/1: Action query time = 3.118 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9640 +t=74: Selected seed 195 with value = 0.9640 +Query 1/1: Action query time = 3.507 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.866 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.526 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.776 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.269 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.350 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.942 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.466 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.389 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.514 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.623 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.012 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.860 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.699 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t2_s3/2026_08_03-01_16_48--with_future_img--episode=12--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 12 +# successes: 1 (8.3%) +Current task success rate: 0.08333333333333333 +Current total success rate: 0.08333333333333333 +Final results: +Total episodes: 12 +Total successes: 1 +Overall success rate: 0.0833 (8.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_49--t7rc800x50_t3_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_49--t7rc800x50_t3_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..24ef0c3b6465ff5bf89e9faf083339d0aaf1788e --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_16_49--t7rc800x50_t3_s0.txt @@ -0,0 +1,1091 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t3_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,4,8,12,16,20,24,28,32,36,40,44,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 6.513 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5129 +t=10: Selected seed 195 with value = 0.5129 +Query 1/1: Action query time = 4.298 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7504 +t=26: Selected seed 195 with value = 0.7504 +Query 1/1: Action query time = 6.623 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8328 +t=42: Selected seed 195 with value = 0.8328 +Query 1/1: Action query time = 6.703 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8597 +t=58: Selected seed 195 with value = 0.8597 +Query 1/1: Action query time = 6.455 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8853 +t=74: Selected seed 195 with value = 0.8853 +Query 1/1: Action query time = 6.863 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7890 +t=90: Selected seed 195 with value = 0.7890 +Query 1/1: Action query time = 6.913 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9157 +t=106: Selected seed 195 with value = 0.9157 +Query 1/1: Action query time = 8.097 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9665 +t=122: Selected seed 195 with value = 0.9665 +Query 1/1: Action query time = 5.257 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8929 +t=138: Selected seed 195 with value = 0.8929 +Query 1/1: Action query time = 7.221 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8316 +t=154: Selected seed 195 with value = 0.8316 +Query 1/1: Action query time = 6.514 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8480 +t=170: Selected seed 195 with value = 0.8480 +Query 1/1: Action query time = 6.782 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8622 +t=186: Selected seed 195 with value = 0.8622 +Query 1/1: Action query time = 7.313 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8702 +t=202: Selected seed 195 with value = 0.8702 +Query 1/1: Action query time = 5.744 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8698 +t=218: Selected seed 195 with value = 0.8698 +Query 1/1: Action query time = 7.429 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8654 +t=234: Selected seed 195 with value = 0.8654 +Query 1/1: Action query time = 6.289 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8837 +t=250: Selected seed 195 with value = 0.8837 +Query 1/1: Action query time = 5.789 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9130 +t=266: Selected seed 195 with value = 0.9130 +Query 1/1: Action query time = 3.958 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9184 +t=282: Selected seed 195 with value = 0.9184 +Query 1/1: Action query time = 3.466 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8882 +t=298: Selected seed 195 with value = 0.8882 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=1--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.455 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6099 +t=10: Selected seed 195 with value = 0.6099 +Query 1/1: Action query time = 7.440 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8436 +t=26: Selected seed 195 with value = 0.8436 +Query 1/1: Action query time = 6.043 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8552 +t=42: Selected seed 195 with value = 0.8552 +Query 1/1: Action query time = 5.244 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8440 +t=58: Selected seed 195 with value = 0.8440 +Query 1/1: Action query time = 4.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8760 +t=74: Selected seed 195 with value = 0.8760 +Query 1/1: Action query time = 6.067 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8805 +t=90: Selected seed 195 with value = 0.8805 +Query 1/1: Action query time = 6.784 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9575 +t=106: Selected seed 195 with value = 0.9575 +Query 1/1: Action query time = 5.831 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9064 +t=122: Selected seed 195 with value = 0.9064 +Query 1/1: Action query time = 5.698 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7219 +t=138: Selected seed 195 with value = 0.7219 +Query 1/1: Action query time = 5.777 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9327 +t=154: Selected seed 195 with value = 0.9327 +Query 1/1: Action query time = 6.969 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9503 +t=170: Selected seed 195 with value = 0.9503 +Query 1/1: Action query time = 6.686 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9202 +t=186: Selected seed 195 with value = 0.9202 +Query 1/1: Action query time = 6.802 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7786 +t=202: Selected seed 195 with value = 0.7786 +Query 1/1: Action query time = 5.733 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9180 +t=218: Selected seed 195 with value = 0.9180 +Query 1/1: Action query time = 6.329 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9566 +t=234: Selected seed 195 with value = 0.9566 +Query 1/1: Action query time = 6.196 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9901 +t=250: Selected seed 195 with value = 0.9901 +Query 1/1: Action query time = 2.806 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.678 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9734 +t=282: Selected seed 195 with value = 0.9734 +Query 1/1: Action query time = 5.671 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9458 +t=298: Selected seed 195 with value = 0.9458 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 5.504 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4765 +t=10: Selected seed 195 with value = 0.4765 +Query 1/1: Action query time = 6.283 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6995 +t=26: Selected seed 195 with value = 0.6995 +Query 1/1: Action query time = 5.920 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8219 +t=42: Selected seed 195 with value = 0.8219 +Query 1/1: Action query time = 6.109 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9020 +t=58: Selected seed 195 with value = 0.9020 +Query 1/1: Action query time = 5.835 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8530 +t=74: Selected seed 195 with value = 0.8530 +Query 1/1: Action query time = 6.959 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8674 +t=90: Selected seed 195 with value = 0.8674 +Query 1/1: Action query time = 5.600 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8569 +t=106: Selected seed 195 with value = 0.8569 +Query 1/1: Action query time = 6.753 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8759 +t=122: Selected seed 195 with value = 0.8759 +Query 1/1: Action query time = 6.805 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7684 +t=138: Selected seed 195 with value = 0.7684 +Query 1/1: Action query time = 6.689 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8954 +t=154: Selected seed 195 with value = 0.8954 +Query 1/1: Action query time = 6.313 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7294 +t=170: Selected seed 195 with value = 0.7294 +Query 1/1: Action query time = 6.414 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8764 +t=186: Selected seed 195 with value = 0.8764 +Query 1/1: Action query time = 5.718 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7418 +t=202: Selected seed 195 with value = 0.7418 +Query 1/1: Action query time = 5.876 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8815 +t=218: Selected seed 195 with value = 0.8815 +Query 1/1: Action query time = 4.690 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8667 +t=234: Selected seed 195 with value = 0.8667 +Query 1/1: Action query time = 5.187 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8301 +t=250: Selected seed 195 with value = 0.8301 +Query 1/1: Action query time = 5.586 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8178 +t=266: Selected seed 195 with value = 0.8178 +Query 1/1: Action query time = 5.357 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8147 +t=282: Selected seed 195 with value = 0.8147 +Query 1/1: Action query time = 4.101 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7617 +t=298: Selected seed 195 with value = 0.7617 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 5.882 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4752 +t=10: Selected seed 195 with value = 0.4752 +Query 1/1: Action query time = 5.662 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6878 +t=26: Selected seed 195 with value = 0.6878 +Query 1/1: Action query time = 6.147 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8094 +t=42: Selected seed 195 with value = 0.8094 +Query 1/1: Action query time = 5.631 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8619 +t=58: Selected seed 195 with value = 0.8619 +Query 1/1: Action query time = 6.391 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8307 +t=74: Selected seed 195 with value = 0.8307 +Query 1/1: Action query time = 5.662 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8292 +t=90: Selected seed 195 with value = 0.8292 +Query 1/1: Action query time = 5.084 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8750 +t=106: Selected seed 195 with value = 0.8750 +Query 1/1: Action query time = 5.737 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=122: Selected seed 195 with value = 0.9928 +Query 1/1: Action query time = 6.410 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9541 +t=138: Selected seed 195 with value = 0.9541 +Query 1/1: Action query time = 6.223 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8143 +t=154: Selected seed 195 with value = 0.8143 +Query 1/1: Action query time = 6.256 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9851 +t=170: Selected seed 195 with value = 0.9851 +Query 1/1: Action query time = 4.327 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8476 +t=186: Selected seed 195 with value = 0.8476 +Query 1/1: Action query time = 6.055 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8646 +t=202: Selected seed 195 with value = 0.8646 +Query 1/1: Action query time = 5.523 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9431 +t=218: Selected seed 195 with value = 0.9431 +Query 1/1: Action query time = 7.751 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9791 +t=234: Selected seed 195 with value = 0.9791 +Query 1/1: Action query time = 5.569 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9698 +t=250: Selected seed 195 with value = 0.9698 +Query 1/1: Action query time = 4.803 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9464 +t=266: Selected seed 195 with value = 0.9464 +Query 1/1: Action query time = 6.859 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9651 +t=282: Selected seed 195 with value = 0.9651 +Query 1/1: Action query time = 6.111 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8670 +t=298: Selected seed 195 with value = 0.8670 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=4--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 5... +Query 1/1: Action query time = 4.217 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4747 +t=10: Selected seed 195 with value = 0.4747 +Query 1/1: Action query time = 5.471 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7458 +t=26: Selected seed 195 with value = 0.7458 +Query 1/1: Action query time = 6.563 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8229 +t=42: Selected seed 195 with value = 0.8229 +Query 1/1: Action query time = 5.420 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9031 +t=58: Selected seed 195 with value = 0.9031 +Query 1/1: Action query time = 5.014 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8452 +t=74: Selected seed 195 with value = 0.8452 +Query 1/1: Action query time = 4.699 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8423 +t=90: Selected seed 195 with value = 0.8423 +Query 1/1: Action query time = 5.233 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9144 +t=106: Selected seed 195 with value = 0.9144 +Query 1/1: Action query time = 5.585 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8503 +t=122: Selected seed 195 with value = 0.8503 +Query 1/1: Action query time = 5.598 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8315 +t=138: Selected seed 195 with value = 0.8315 +Query 1/1: Action query time = 4.766 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8830 +t=154: Selected seed 195 with value = 0.8830 +Query 1/1: Action query time = 6.091 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8955 +t=170: Selected seed 195 with value = 0.8955 +Query 1/1: Action query time = 5.956 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8554 +t=186: Selected seed 195 with value = 0.8554 +Query 1/1: Action query time = 7.613 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8699 +t=202: Selected seed 195 with value = 0.8699 +Query 1/1: Action query time = 5.995 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8562 +t=218: Selected seed 195 with value = 0.8562 +Query 1/1: Action query time = 5.238 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8288 +t=234: Selected seed 195 with value = 0.8288 +Query 1/1: Action query time = 6.657 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7765 +t=250: Selected seed 195 with value = 0.7765 +Query 1/1: Action query time = 5.702 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7715 +t=266: Selected seed 195 with value = 0.7715 +Query 1/1: Action query time = 5.372 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7768 +t=282: Selected seed 195 with value = 0.7768 +Query 1/1: Action query time = 5.471 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7655 +t=298: Selected seed 195 with value = 0.7655 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=5--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 5 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 6... +Query 1/1: Action query time = 5.319 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7272 +t=10: Selected seed 195 with value = 0.7272 +Query 1/1: Action query time = 4.768 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7255 +t=26: Selected seed 195 with value = 0.7255 +Query 1/1: Action query time = 5.356 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7539 +t=42: Selected seed 195 with value = 0.7539 +Query 1/1: Action query time = 6.762 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9126 +t=58: Selected seed 195 with value = 0.9126 +Query 1/1: Action query time = 4.644 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9556 +t=74: Selected seed 195 with value = 0.9556 +Query 1/1: Action query time = 4.339 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9459 +t=90: Selected seed 195 with value = 0.9459 +Query 1/1: Action query time = 4.272 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9500 +t=106: Selected seed 195 with value = 0.9500 +Query 1/1: Action query time = 3.807 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9585 +t=122: Selected seed 195 with value = 0.9585 +Query 1/1: Action query time = 7.138 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9654 +t=138: Selected seed 195 with value = 0.9654 +Query 1/1: Action query time = 5.760 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9753 +t=154: Selected seed 195 with value = 0.9753 +Query 1/1: Action query time = 4.948 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9804 +t=170: Selected seed 195 with value = 0.9804 +Query 1/1: Action query time = 6.355 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.025 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.784 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=218: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 7.866 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=234: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 7.096 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=250: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 5.142 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=266: Selected seed 195 with value = 0.9968 +Query 1/1: Action query time = 5.082 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.320 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=298: Selected seed 195 with value = 0.9984 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=6--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 6 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 7... +Query 1/1: Action query time = 2.887 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4819 +t=10: Selected seed 195 with value = 0.4819 +Query 1/1: Action query time = 3.682 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5983 +t=26: Selected seed 195 with value = 0.5983 +Query 1/1: Action query time = 4.787 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8186 +t=42: Selected seed 195 with value = 0.8186 +Query 1/1: Action query time = 4.036 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8932 +t=58: Selected seed 195 with value = 0.8932 +Query 1/1: Action query time = 5.916 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8020 +t=74: Selected seed 195 with value = 0.8020 +Query 1/1: Action query time = 7.124 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9205 +t=90: Selected seed 195 with value = 0.9205 +Query 1/1: Action query time = 4.253 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8774 +t=106: Selected seed 195 with value = 0.8774 +Query 1/1: Action query time = 5.422 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8328 +t=122: Selected seed 195 with value = 0.8328 +Query 1/1: Action query time = 6.206 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8864 +t=138: Selected seed 195 with value = 0.8864 +Query 1/1: Action query time = 7.138 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8041 +t=154: Selected seed 195 with value = 0.8041 +Query 1/1: Action query time = 4.186 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8650 +t=170: Selected seed 195 with value = 0.8650 +Query 1/1: Action query time = 5.316 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7448 +t=186: Selected seed 195 with value = 0.7448 +Query 1/1: Action query time = 5.248 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8418 +t=202: Selected seed 195 with value = 0.8418 +Query 1/1: Action query time = 6.879 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8130 +t=218: Selected seed 195 with value = 0.8130 +Query 1/1: Action query time = 6.937 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8210 +t=234: Selected seed 195 with value = 0.8210 +Query 1/1: Action query time = 5.969 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7305 +t=250: Selected seed 195 with value = 0.7305 +Query 1/1: Action query time = 5.344 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8075 +t=266: Selected seed 195 with value = 0.8075 +Query 1/1: Action query time = 5.385 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7941 +t=282: Selected seed 195 with value = 0.7941 +Query 1/1: Action query time = 6.583 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9080 +t=298: Selected seed 195 with value = 0.9080 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=7--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 7 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 8... +Query 1/1: Action query time = 7.547 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5373 +t=10: Selected seed 195 with value = 0.5373 +Query 1/1: Action query time = 4.690 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5977 +t=26: Selected seed 195 with value = 0.5977 +Query 1/1: Action query time = 3.572 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8098 +t=42: Selected seed 195 with value = 0.8098 +Query 1/1: Action query time = 4.139 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8856 +t=58: Selected seed 195 with value = 0.8856 +Query 1/1: Action query time = 4.157 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8906 +t=74: Selected seed 195 with value = 0.8906 +Query 1/1: Action query time = 6.718 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7672 +t=90: Selected seed 195 with value = 0.7672 +Query 1/1: Action query time = 5.470 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9413 +t=106: Selected seed 195 with value = 0.9413 +Query 1/1: Action query time = 6.147 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9547 +t=122: Selected seed 195 with value = 0.9547 +Query 1/1: Action query time = 4.993 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9821 +t=138: Selected seed 195 with value = 0.9821 +Query 1/1: Action query time = 4.741 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9581 +t=154: Selected seed 195 with value = 0.9581 +Query 1/1: Action query time = 7.812 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9352 +t=170: Selected seed 195 with value = 0.9352 +Query 1/1: Action query time = 6.214 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8975 +t=186: Selected seed 195 with value = 0.8975 +Query 1/1: Action query time = 4.723 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9078 +t=202: Selected seed 195 with value = 0.9078 +Query 1/1: Action query time = 4.841 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8867 +t=218: Selected seed 195 with value = 0.8867 +Query 1/1: Action query time = 5.862 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8877 +t=234: Selected seed 195 with value = 0.8877 +Query 1/1: Action query time = 6.597 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8807 +t=250: Selected seed 195 with value = 0.8807 +Query 1/1: Action query time = 5.704 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8622 +t=266: Selected seed 195 with value = 0.8622 +Query 1/1: Action query time = 7.169 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8184 +t=282: Selected seed 195 with value = 0.8184 +Query 1/1: Action query time = 5.306 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8186 +t=298: Selected seed 195 with value = 0.8186 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=8--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 8 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 9... +Query 1/1: Action query time = 5.409 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4749 +t=10: Selected seed 195 with value = 0.4749 +Query 1/1: Action query time = 3.231 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7364 +t=26: Selected seed 195 with value = 0.7364 +Query 1/1: Action query time = 3.827 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8290 +t=42: Selected seed 195 with value = 0.8290 +Query 1/1: Action query time = 3.485 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8515 +t=58: Selected seed 195 with value = 0.8515 +Query 1/1: Action query time = 3.921 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8359 +t=74: Selected seed 195 with value = 0.8359 +Query 1/1: Action query time = 6.742 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8685 +t=90: Selected seed 195 with value = 0.8685 +Query 1/1: Action query time = 6.103 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8749 +t=106: Selected seed 195 with value = 0.8749 +Query 1/1: Action query time = 3.774 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8856 +t=122: Selected seed 195 with value = 0.8856 +Query 1/1: Action query time = 5.007 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9574 +t=138: Selected seed 195 with value = 0.9574 +Query 1/1: Action query time = 4.637 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8664 +t=154: Selected seed 195 with value = 0.8664 +Query 1/1: Action query time = 4.723 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8764 +t=170: Selected seed 195 with value = 0.8764 +Query 1/1: Action query time = 6.147 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9086 +t=186: Selected seed 195 with value = 0.9086 +Query 1/1: Action query time = 5.795 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8784 +t=202: Selected seed 195 with value = 0.8784 +Query 1/1: Action query time = 6.599 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8406 +t=218: Selected seed 195 with value = 0.8406 +Query 1/1: Action query time = 6.496 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8942 +t=234: Selected seed 195 with value = 0.8942 +Query 1/1: Action query time = 5.432 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8164 +t=250: Selected seed 195 with value = 0.8164 +Query 1/1: Action query time = 5.818 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8940 +t=266: Selected seed 195 with value = 0.8940 +Query 1/1: Action query time = 5.792 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8034 +t=282: Selected seed 195 with value = 0.8034 +Query 1/1: Action query time = 3.464 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8219 +t=298: Selected seed 195 with value = 0.8219 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=9--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 9 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 10... +Query 1/1: Action query time = 6.123 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4733 +t=10: Selected seed 195 with value = 0.4733 +Query 1/1: Action query time = 5.500 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6483 +t=26: Selected seed 195 with value = 0.6483 +Query 1/1: Action query time = 3.431 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7909 +t=42: Selected seed 195 with value = 0.7909 +Query 1/1: Action query time = 5.310 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9037 +t=58: Selected seed 195 with value = 0.9037 +Query 1/1: Action query time = 2.858 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9507 +t=74: Selected seed 195 with value = 0.9507 +Query 1/1: Action query time = 3.267 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8781 +t=90: Selected seed 195 with value = 0.8781 +Query 1/1: Action query time = 5.162 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8586 +t=106: Selected seed 195 with value = 0.8586 +Query 1/1: Action query time = 4.951 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8921 +t=122: Selected seed 195 with value = 0.8921 +Query 1/1: Action query time = 3.351 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8867 +t=138: Selected seed 195 with value = 0.8867 +Query 1/1: Action query time = 5.060 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8082 +t=154: Selected seed 195 with value = 0.8082 +Query 1/1: Action query time = 4.205 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8939 +t=170: Selected seed 195 with value = 0.8939 +Query 1/1: Action query time = 5.978 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7681 +t=186: Selected seed 195 with value = 0.7681 +Query 1/1: Action query time = 5.167 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8087 +t=202: Selected seed 195 with value = 0.8087 +Query 1/1: Action query time = 5.533 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8406 +t=218: Selected seed 195 with value = 0.8406 +Query 1/1: Action query time = 4.697 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8371 +t=234: Selected seed 195 with value = 0.8371 +Query 1/1: Action query time = 5.668 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7079 +t=250: Selected seed 195 with value = 0.7079 +Query 1/1: Action query time = 4.648 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8293 +t=266: Selected seed 195 with value = 0.8293 +Query 1/1: Action query time = 4.799 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8732 +t=282: Selected seed 195 with value = 0.8732 +Query 1/1: Action query time = 4.336 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9044 +t=298: Selected seed 195 with value = 0.9044 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=10--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 10 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 11... +Query 1/1: Action query time = 5.148 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4820 +t=10: Selected seed 195 with value = 0.4820 +Query 1/1: Action query time = 4.048 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6799 +t=26: Selected seed 195 with value = 0.6799 +Query 1/1: Action query time = 4.171 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8150 +t=42: Selected seed 195 with value = 0.8150 +Query 1/1: Action query time = 4.765 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8306 +t=58: Selected seed 195 with value = 0.8306 +Query 1/1: Action query time = 4.127 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9067 +t=74: Selected seed 195 with value = 0.9067 +Query 1/1: Action query time = 3.563 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8844 +t=90: Selected seed 195 with value = 0.8844 +Query 1/1: Action query time = 3.701 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8502 +t=106: Selected seed 195 with value = 0.8502 +Query 1/1: Action query time = 5.090 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7872 +t=122: Selected seed 195 with value = 0.7872 +Query 1/1: Action query time = 5.127 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8685 +t=138: Selected seed 195 with value = 0.8685 +Query 1/1: Action query time = 4.292 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8726 +t=154: Selected seed 195 with value = 0.8726 +Query 1/1: Action query time = 4.360 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9215 +t=170: Selected seed 195 with value = 0.9215 +Query 1/1: Action query time = 3.470 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8639 +t=186: Selected seed 195 with value = 0.8639 +Query 1/1: Action query time = 4.813 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8864 +t=202: Selected seed 195 with value = 0.8864 +Query 1/1: Action query time = 5.410 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9380 +t=218: Selected seed 195 with value = 0.9380 +Query 1/1: Action query time = 4.437 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8805 +t=234: Selected seed 195 with value = 0.8805 +Query 1/1: Action query time = 4.406 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8376 +t=250: Selected seed 195 with value = 0.8376 +Query 1/1: Action query time = 4.502 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8693 +t=266: Selected seed 195 with value = 0.8693 +Query 1/1: Action query time = 5.871 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8733 +t=282: Selected seed 195 with value = 0.8733 +Query 1/1: Action query time = 4.717 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8160 +t=298: Selected seed 195 with value = 0.8160 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=11--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 11 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 12... +Query 1/1: Action query time = 4.973 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5068 +t=10: Selected seed 195 with value = 0.5068 +Query 1/1: Action query time = 3.302 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6519 +t=26: Selected seed 195 with value = 0.6519 +Query 1/1: Action query time = 4.985 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8022 +t=42: Selected seed 195 with value = 0.8022 +Query 1/1: Action query time = 2.669 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8783 +t=58: Selected seed 195 with value = 0.8783 +Query 1/1: Action query time = 2.195 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8979 +t=74: Selected seed 195 with value = 0.8979 +Query 1/1: Action query time = 2.813 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9345 +t=90: Selected seed 195 with value = 0.9345 +Query 1/1: Action query time = 2.721 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9061 +t=106: Selected seed 195 with value = 0.9061 +Query 1/1: Action query time = 3.470 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9193 +t=122: Selected seed 195 with value = 0.9193 +Query 1/1: Action query time = 3.603 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9061 +t=138: Selected seed 195 with value = 0.9061 +Query 1/1: Action query time = 3.867 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8360 +t=154: Selected seed 195 with value = 0.8360 +Query 1/1: Action query time = 3.432 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9158 +t=170: Selected seed 195 with value = 0.9158 +Query 1/1: Action query time = 2.873 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7756 +t=186: Selected seed 195 with value = 0.7756 +Query 1/1: Action query time = 3.173 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8730 +t=202: Selected seed 195 with value = 0.8730 +Query 1/1: Action query time = 4.131 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8260 +t=218: Selected seed 195 with value = 0.8260 +Query 1/1: Action query time = 4.102 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8671 +t=234: Selected seed 195 with value = 0.8671 +Query 1/1: Action query time = 3.600 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8251 +t=250: Selected seed 195 with value = 0.8251 +Query 1/1: Action query time = 3.818 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8682 +t=266: Selected seed 195 with value = 0.8682 +Query 1/1: Action query time = 4.007 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8538 +t=282: Selected seed 195 with value = 0.8538 +Query 1/1: Action query time = 3.901 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8812 +t=298: Selected seed 195 with value = 0.8812 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=12--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 12 +# successes: 0 (0.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 13... +Query 1/1: Action query time = 3.972 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4941 +t=10: Selected seed 195 with value = 0.4941 +Query 1/1: Action query time = 3.778 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7054 +t=26: Selected seed 195 with value = 0.7054 +Query 1/1: Action query time = 3.497 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7868 +t=42: Selected seed 195 with value = 0.7868 +Query 1/1: Action query time = 3.533 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8691 +t=58: Selected seed 195 with value = 0.8691 +Query 1/1: Action query time = 2.363 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8597 +t=74: Selected seed 195 with value = 0.8597 +Query 1/1: Action query time = 1.821 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8268 +t=90: Selected seed 195 with value = 0.8268 +Query 1/1: Action query time = 1.979 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9277 +t=106: Selected seed 195 with value = 0.9277 +Query 1/1: Action query time = 1.791 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9013 +t=122: Selected seed 195 with value = 0.9013 +Query 1/1: Action query time = 2.518 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7521 +t=138: Selected seed 195 with value = 0.7521 +Query 1/1: Action query time = 2.502 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9025 +t=154: Selected seed 195 with value = 0.9025 +Query 1/1: Action query time = 1.888 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8572 +t=170: Selected seed 195 with value = 0.8572 +Query 1/1: Action query time = 1.905 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8625 +t=186: Selected seed 195 with value = 0.8625 +Query 1/1: Action query time = 1.802 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7575 +t=202: Selected seed 195 with value = 0.7575 +Query 1/1: Action query time = 1.764 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8629 +t=218: Selected seed 195 with value = 0.8629 +Query 1/1: Action query time = 2.054 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9029 +t=234: Selected seed 195 with value = 0.9029 +Query 1/1: Action query time = 2.083 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8639 +t=250: Selected seed 195 with value = 0.8639 +Query 1/1: Action query time = 1.577 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8154 +t=266: Selected seed 195 with value = 0.8154 +Query 1/1: Action query time = 1.351 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8809 +t=282: Selected seed 195 with value = 0.8809 +Query 1/1: Action query time = 1.815 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8421 +t=298: Selected seed 195 with value = 0.8421 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t3_s0/2026_08_03-01_16_49--with_future_img--episode=13--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 13 +# successes: 0 (0.0%) +Current task success rate: 0.0 +Current total success rate: 0.0 +Final results: +Total episodes: 13 +Total successes: 0 +Overall success rate: 0.0000 (0.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_43--t7rc800x50_t4_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_43--t7rc800x50_t4_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..816f83058bd4faf36e558f55f7e21cfc6113598a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_43--t7rc800x50_t4_s0.txt @@ -0,0 +1,523 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t4_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,4,8,12,16,20,24,28,32,36,40,44,48', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.428 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5438 +t=10: Selected seed 195 with value = 0.5438 +Query 1/1: Action query time = 2.143 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5982 +t=26: Selected seed 195 with value = 0.5982 +Query 1/1: Action query time = 2.817 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7306 +t=42: Selected seed 195 with value = 0.7306 +Query 1/1: Action query time = 3.170 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8185 +t=58: Selected seed 195 with value = 0.8185 +Query 1/1: Action query time = 4.247 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9826 +t=74: Selected seed 195 with value = 0.9826 +Query 1/1: Action query time = 4.833 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=90: Selected seed 195 with value = 0.9985 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.433 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5614 +t=10: Selected seed 195 with value = 0.5614 +Query 1/1: Action query time = 6.429 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6670 +t=26: Selected seed 195 with value = 0.6670 +Query 1/1: Action query time = 5.933 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7706 +t=42: Selected seed 195 with value = 0.7706 +Query 1/1: Action query time = 4.578 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8160 +t=58: Selected seed 195 with value = 0.8160 +Query 1/1: Action query time = 8.123 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9901 +t=74: Selected seed 195 with value = 0.9901 +Query 1/1: Action query time = 7.400 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=90: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 6.164 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5762 +t=10: Selected seed 195 with value = 0.5762 +Query 1/1: Action query time = 4.124 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6605 +t=26: Selected seed 195 with value = 0.6605 +Query 1/1: Action query time = 5.279 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7217 +t=42: Selected seed 195 with value = 0.7217 +Query 1/1: Action query time = 4.999 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8292 +t=58: Selected seed 195 with value = 0.8292 +Query 1/1: Action query time = 4.318 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9442 +t=74: Selected seed 195 with value = 0.9442 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 4.168 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5055 +t=10: Selected seed 195 with value = 0.5055 +Query 1/1: Action query time = 4.656 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6641 +t=26: Selected seed 195 with value = 0.6641 +Query 1/1: Action query time = 4.301 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8011 +t=42: Selected seed 195 with value = 0.8011 +Query 1/1: Action query time = 4.949 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8290 +t=58: Selected seed 195 with value = 0.8290 +Query 1/1: Action query time = 5.894 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9144 +t=74: Selected seed 195 with value = 0.9144 +Query 1/1: Action query time = 6.527 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=4--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 5... +Query 1/1: Action query time = 4.617 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6262 +t=10: Selected seed 195 with value = 0.6262 +Query 1/1: Action query time = 4.530 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7079 +t=26: Selected seed 195 with value = 0.7079 +Query 1/1: Action query time = 4.842 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8373 +t=42: Selected seed 195 with value = 0.8373 +Query 1/1: Action query time = 6.093 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9511 +t=58: Selected seed 195 with value = 0.9511 +Query 1/1: Action query time = 6.052 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9820 +t=74: Selected seed 195 with value = 0.9820 +Query 1/1: Action query time = 4.923 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9821 +t=90: Selected seed 195 with value = 0.9821 +Query 1/1: Action query time = 5.975 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9849 +t=106: Selected seed 195 with value = 0.9849 +Query 1/1: Action query time = 6.072 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=122: Selected seed 195 with value = 0.9867 +Query 1/1: Action query time = 5.961 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9873 +t=138: Selected seed 195 with value = 0.9873 +Query 1/1: Action query time = 5.746 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9894 +t=154: Selected seed 195 with value = 0.9894 +Query 1/1: Action query time = 5.096 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9906 +t=170: Selected seed 195 with value = 0.9906 +Query 1/1: Action query time = 6.615 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9850 +t=186: Selected seed 195 with value = 0.9850 +Query 1/1: Action query time = 5.302 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9848 +t=202: Selected seed 195 with value = 0.9848 +Query 1/1: Action query time = 3.284 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9890 +t=218: Selected seed 195 with value = 0.9890 +Query 1/1: Action query time = 5.784 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9920 +t=234: Selected seed 195 with value = 0.9920 +Query 1/1: Action query time = 6.484 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9942 +t=250: Selected seed 195 with value = 0.9942 +Query 1/1: Action query time = 3.987 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=266: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 3.833 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.381 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=5--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 5 +# successes: 4 (80.0%) + +Task: put the bowl on top of the cabinet +Starting episode 6... +Query 1/1: Action query time = 6.684 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4731 +t=10: Selected seed 195 with value = 0.4731 +Query 1/1: Action query time = 6.285 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6353 +t=26: Selected seed 195 with value = 0.6353 +Query 1/1: Action query time = 5.667 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7255 +t=42: Selected seed 195 with value = 0.7255 +Query 1/1: Action query time = 5.001 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8731 +t=58: Selected seed 195 with value = 0.8731 +Query 1/1: Action query time = 4.466 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9766 +t=74: Selected seed 195 with value = 0.9766 +Query 1/1: Action query time = 5.456 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=90: Selected seed 195 with value = 0.9928 +Query 1/1: Action query time = 5.981 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9955 +t=106: Selected seed 195 with value = 0.9955 +Query 1/1: Action query time = 6.166 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=122: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 5.566 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=138: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 4.381 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=154: Selected seed 195 with value = 0.9962 +Query 1/1: Action query time = 5.526 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9911 +t=170: Selected seed 195 with value = 0.9911 +Query 1/1: Action query time = 4.269 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=186: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 5.080 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=202: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 4.875 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.265 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.782 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.360 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.355 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9674 +t=282: Selected seed 195 with value = 0.9674 +Query 1/1: Action query time = 5.636 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=298: Selected seed 195 with value = 0.9993 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=6--success=False--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: False +# episodes completed so far: 6 +# successes: 4 (66.7%) + +Task: put the bowl on top of the cabinet +Starting episode 7... +Query 1/1: Action query time = 6.245 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5586 +t=10: Selected seed 195 with value = 0.5586 +Query 1/1: Action query time = 7.325 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6381 +t=26: Selected seed 195 with value = 0.6381 +Query 1/1: Action query time = 5.683 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7274 +t=42: Selected seed 195 with value = 0.7274 +Query 1/1: Action query time = 5.658 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8476 +t=58: Selected seed 195 with value = 0.8476 +Query 1/1: Action query time = 5.765 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=74: Selected seed 195 with value = 0.9864 +Query 1/1: Action query time = 6.075 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=90: Selected seed 195 with value = 0.9966 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=7--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 7 +# successes: 5 (71.4%) + +Task: put the bowl on top of the cabinet +Starting episode 8... +Query 1/1: Action query time = 6.814 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5392 +t=10: Selected seed 195 with value = 0.5392 +Query 1/1: Action query time = 5.772 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6336 +t=26: Selected seed 195 with value = 0.6336 +Query 1/1: Action query time = 5.766 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7484 +t=42: Selected seed 195 with value = 0.7484 +Query 1/1: Action query time = 4.536 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8347 +t=58: Selected seed 195 with value = 0.8347 +Query 1/1: Action query time = 4.761 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9889 +t=74: Selected seed 195 with value = 0.9889 +Query 1/1: Action query time = 3.876 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=90: Selected seed 195 with value = 0.9945 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=8--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 8 +# successes: 6 (75.0%) + +Task: put the bowl on top of the cabinet +Starting episode 9... +Query 1/1: Action query time = 4.305 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5922 +t=10: Selected seed 195 with value = 0.5922 +Query 1/1: Action query time = 3.494 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6337 +t=26: Selected seed 195 with value = 0.6337 +Query 1/1: Action query time = 6.394 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7310 +t=42: Selected seed 195 with value = 0.7310 +Query 1/1: Action query time = 5.301 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8395 +t=58: Selected seed 195 with value = 0.8395 +Query 1/1: Action query time = 5.691 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9907 +t=74: Selected seed 195 with value = 0.9907 +Query 1/1: Action query time = 5.204 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9930 +t=90: Selected seed 195 with value = 0.9930 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=9--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 9 +# successes: 7 (77.8%) + +Task: put the bowl on top of the cabinet +Starting episode 10... +Query 1/1: Action query time = 3.985 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5375 +t=10: Selected seed 195 with value = 0.5375 +Query 1/1: Action query time = 4.267 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6240 +t=26: Selected seed 195 with value = 0.6240 +Query 1/1: Action query time = 4.220 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7365 +t=42: Selected seed 195 with value = 0.7365 +Query 1/1: Action query time = 6.044 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8511 +t=58: Selected seed 195 with value = 0.8511 +Query 1/1: Action query time = 3.849 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9889 +t=74: Selected seed 195 with value = 0.9889 +Query 1/1: Action query time = 3.337 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9967 +t=90: Selected seed 195 with value = 0.9967 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=10--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 10 +# successes: 8 (80.0%) + +Task: put the bowl on top of the cabinet +Starting episode 11... +Query 1/1: Action query time = 5.492 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5292 +t=10: Selected seed 195 with value = 0.5292 +Query 1/1: Action query time = 4.790 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6338 +t=26: Selected seed 195 with value = 0.6338 +Query 1/1: Action query time = 5.570 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6926 +t=42: Selected seed 195 with value = 0.6926 +Query 1/1: Action query time = 4.312 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7835 +t=58: Selected seed 195 with value = 0.7835 +Query 1/1: Action query time = 4.788 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9071 +t=74: Selected seed 195 with value = 0.9071 +Query 1/1: Action query time = 4.912 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9443 +t=90: Selected seed 195 with value = 0.9443 +Query 1/1: Action query time = 4.153 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=11--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 11 +# successes: 9 (81.8%) + +Task: put the bowl on top of the cabinet +Starting episode 12... +Query 1/1: Action query time = 3.879 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5608 +t=10: Selected seed 195 with value = 0.5608 +Query 1/1: Action query time = 4.875 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6247 +t=26: Selected seed 195 with value = 0.6247 +Query 1/1: Action query time = 4.844 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7592 +t=42: Selected seed 195 with value = 0.7592 +Query 1/1: Action query time = 5.375 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8581 +t=58: Selected seed 195 with value = 0.8581 +Query 1/1: Action query time = 5.616 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.769 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=90: Selected seed 195 with value = 0.9984 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=12--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 12 +# successes: 10 (83.3%) + +Task: put the bowl on top of the cabinet +Starting episode 13... +Query 1/1: Action query time = 4.952 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5086 +t=10: Selected seed 195 with value = 0.5086 +Query 1/1: Action query time = 3.483 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6105 +t=26: Selected seed 195 with value = 0.6105 +Query 1/1: Action query time = 3.057 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6846 +t=42: Selected seed 195 with value = 0.6846 +Query 1/1: Action query time = 4.395 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7934 +t=58: Selected seed 195 with value = 0.7934 +Query 1/1: Action query time = 4.648 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8600 +t=74: Selected seed 195 with value = 0.8600 +Query 1/1: Action query time = 5.872 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9820 +t=90: Selected seed 195 with value = 0.9820 +Query 1/1: Action query time = 4.435 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9829 +t=106: Selected seed 195 with value = 0.9829 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t4_s0/2026_08_03-01_47_43--with_future_img--episode=13--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 13 +# successes: 11 (84.6%) +Current task success rate: 0.8461538461538461 +Current total success rate: 0.8461538461538461 +Final results: +Total episodes: 13 +Total successes: 11 +Overall success rate: 0.8462 (84.6%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_45--t7rc800x50_t7_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_45--t7rc800x50_t7_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..4c7ed3673e350eaee7e3329309f7180368684b4b --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_45--t7rc800x50_t7_s3.txt @@ -0,0 +1,348 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t7_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,7,11,15,19,23,27,31,35,39,43,47', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 4.176 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4502 +t=10: Selected seed 195 with value = 0.4502 +Query 1/1: Action query time = 3.586 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5710 +t=26: Selected seed 195 with value = 0.5710 +Query 1/1: Action query time = 4.408 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6402 +t=42: Selected seed 195 with value = 0.6402 +Query 1/1: Action query time = 6.263 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8021 +t=58: Selected seed 195 with value = 0.8021 +Query 1/1: Action query time = 6.285 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9651 +t=74: Selected seed 195 with value = 0.9651 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 4.958 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4852 +t=10: Selected seed 195 with value = 0.4852 +Query 1/1: Action query time = 4.563 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5506 +t=26: Selected seed 195 with value = 0.5506 +Query 1/1: Action query time = 6.579 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6662 +t=42: Selected seed 195 with value = 0.6662 +Query 1/1: Action query time = 4.238 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7774 +t=58: Selected seed 195 with value = 0.7774 +Query 1/1: Action query time = 5.412 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8955 +t=74: Selected seed 195 with value = 0.8955 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 6.139 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4784 +t=10: Selected seed 195 with value = 0.4784 +Query 1/1: Action query time = 5.920 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5450 +t=26: Selected seed 195 with value = 0.5450 +Query 1/1: Action query time = 5.188 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6756 +t=42: Selected seed 195 with value = 0.6756 +Query 1/1: Action query time = 6.055 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7880 +t=58: Selected seed 195 with value = 0.7880 +Query 1/1: Action query time = 4.857 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9150 +t=74: Selected seed 195 with value = 0.9150 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 4.758 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5400 +t=10: Selected seed 195 with value = 0.5400 +Query 1/1: Action query time = 5.602 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5964 +t=26: Selected seed 195 with value = 0.5964 +Query 1/1: Action query time = 3.301 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6784 +t=42: Selected seed 195 with value = 0.6784 +Query 1/1: Action query time = 3.579 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8257 +t=58: Selected seed 195 with value = 0.8257 +Query 1/1: Action query time = 4.719 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9774 +t=74: Selected seed 195 with value = 0.9774 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 5.517 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5014 +t=10: Selected seed 195 with value = 0.5014 +Query 1/1: Action query time = 3.948 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5802 +t=26: Selected seed 195 with value = 0.5802 +Query 1/1: Action query time = 4.733 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6761 +t=42: Selected seed 195 with value = 0.6761 +Query 1/1: Action query time = 4.636 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7980 +t=58: Selected seed 195 with value = 0.7980 +Query 1/1: Action query time = 5.381 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9468 +t=74: Selected seed 195 with value = 0.9468 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 5.572 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4721 +t=10: Selected seed 195 with value = 0.4721 +Query 1/1: Action query time = 6.299 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5513 +t=26: Selected seed 195 with value = 0.5513 +Query 1/1: Action query time = 4.591 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6461 +t=42: Selected seed 195 with value = 0.6461 +Query 1/1: Action query time = 4.653 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7788 +t=58: Selected seed 195 with value = 0.7788 +Query 1/1: Action query time = 4.157 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8860 +t=74: Selected seed 195 with value = 0.8860 +Query 1/1: Action query time = 6.384 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: turn on the stove +Starting episode 7... +Query 1/1: Action query time = 5.104 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4853 +t=10: Selected seed 195 with value = 0.4853 +Query 1/1: Action query time = 4.418 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5735 +t=26: Selected seed 195 with value = 0.5735 +Query 1/1: Action query time = 5.138 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6582 +t=42: Selected seed 195 with value = 0.6582 +Query 1/1: Action query time = 5.790 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7699 +t=58: Selected seed 195 with value = 0.7699 +Query 1/1: Action query time = 4.244 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9039 +t=74: Selected seed 195 with value = 0.9039 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=7--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: turn on the stove +Starting episode 8... +Query 1/1: Action query time = 4.624 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4796 +t=10: Selected seed 195 with value = 0.4796 +Query 1/1: Action query time = 5.950 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5690 +t=26: Selected seed 195 with value = 0.5690 +Query 1/1: Action query time = 5.599 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6193 +t=42: Selected seed 195 with value = 0.6193 +Query 1/1: Action query time = 4.947 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7321 +t=58: Selected seed 195 with value = 0.7321 +Query 1/1: Action query time = 4.387 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8507 +t=74: Selected seed 195 with value = 0.8507 +Query 1/1: Action query time = 4.182 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9837 +t=90: Selected seed 195 with value = 0.9837 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=8--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: turn on the stove +Starting episode 9... +Query 1/1: Action query time = 5.280 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5114 +t=10: Selected seed 195 with value = 0.5114 +Query 1/1: Action query time = 3.589 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5982 +t=26: Selected seed 195 with value = 0.5982 +Query 1/1: Action query time = 5.491 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6669 +t=42: Selected seed 195 with value = 0.6669 +Query 1/1: Action query time = 5.301 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7518 +t=58: Selected seed 195 with value = 0.7518 +Query 1/1: Action query time = 5.873 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8669 +t=74: Selected seed 195 with value = 0.8669 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=9--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: turn on the stove +Starting episode 10... +Query 1/1: Action query time = 5.538 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5122 +t=10: Selected seed 195 with value = 0.5122 +Query 1/1: Action query time = 5.948 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5916 +t=26: Selected seed 195 with value = 0.5916 +Query 1/1: Action query time = 4.527 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6869 +t=42: Selected seed 195 with value = 0.6869 +Query 1/1: Action query time = 5.408 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8065 +t=58: Selected seed 195 with value = 0.8065 +Query 1/1: Action query time = 5.281 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9567 +t=74: Selected seed 195 with value = 0.9567 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=10--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: turn on the stove +Starting episode 11... +Query 1/1: Action query time = 5.715 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4842 +t=10: Selected seed 195 with value = 0.4842 +Query 1/1: Action query time = 4.690 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5666 +t=26: Selected seed 195 with value = 0.5666 +Query 1/1: Action query time = 4.230 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6675 +t=42: Selected seed 195 with value = 0.6675 +Query 1/1: Action query time = 4.637 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7657 +t=58: Selected seed 195 with value = 0.7657 +Query 1/1: Action query time = 5.963 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9104 +t=74: Selected seed 195 with value = 0.9104 +Query 1/1: Action query time = 4.847 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9967 +t=90: Selected seed 195 with value = 0.9967 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=11--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: turn on the stove +Starting episode 12... +Query 1/1: Action query time = 5.577 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5182 +t=10: Selected seed 195 with value = 0.5182 +Query 1/1: Action query time = 5.056 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5963 +t=26: Selected seed 195 with value = 0.5963 +Query 1/1: Action query time = 4.284 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6639 +t=42: Selected seed 195 with value = 0.6639 +Query 1/1: Action query time = 2.810 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7903 +t=58: Selected seed 195 with value = 0.7903 +Query 1/1: Action query time = 4.657 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9327 +t=74: Selected seed 195 with value = 0.9327 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t7_s3/2026_08_03-01_47_45--with_future_img--episode=12--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 12 +Total successes: 12 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_46--t7dc800x50_t5_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_46--t7dc800x50_t5_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..cc3ac7dd68cb08b46d716600c5352857542b37fe --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_46--t7dc800x50_t5_s1.txt @@ -0,0 +1,883 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_demochan800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7dc800x50_t5_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 7.363 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4533 +t=10: Selected seed 195 with value = 0.4533 +Query 1/1: Action query time = 6.730 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5322 +t=26: Selected seed 195 with value = 0.5322 +Query 1/1: Action query time = 5.756 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6316 +t=42: Selected seed 195 with value = 0.6316 +Query 1/1: Action query time = 5.496 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8200 +t=58: Selected seed 195 with value = 0.8200 +Query 1/1: Action query time = 5.169 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9256 +t=74: Selected seed 195 with value = 0.9256 +Query 1/1: Action query time = 5.083 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9761 +t=90: Selected seed 195 with value = 0.9761 +Query 1/1: Action query time = 5.511 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.860 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.156 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.856 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.136 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.844 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.331 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.507 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.970 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4515 +t=10: Selected seed 195 with value = 0.4515 +Query 1/1: Action query time = 4.733 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5206 +t=26: Selected seed 195 with value = 0.5206 +Query 1/1: Action query time = 5.077 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6095 +t=42: Selected seed 195 with value = 0.6095 +Query 1/1: Action query time = 2.957 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7627 +t=58: Selected seed 195 with value = 0.7627 +Query 1/1: Action query time = 4.634 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8638 +t=74: Selected seed 195 with value = 0.8638 +Query 1/1: Action query time = 3.402 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9470 +t=90: Selected seed 195 with value = 0.9470 +Query 1/1: Action query time = 3.847 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9935 +t=106: Selected seed 195 with value = 0.9935 +Query 1/1: Action query time = 4.992 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 7.044 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.083 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.683 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=170: Selected seed 195 with value = 0.9986 +Query 1/1: Action query time = 4.092 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9774 +t=186: Selected seed 195 with value = 0.9774 +Query 1/1: Action query time = 5.743 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9727 +t=202: Selected seed 195 with value = 0.9727 +Query 1/1: Action query time = 4.272 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9873 +t=218: Selected seed 195 with value = 0.9873 +Query 1/1: Action query time = 5.590 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=234: Selected seed 195 with value = 0.9903 +Query 1/1: Action query time = 5.774 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=250: Selected seed 195 with value = 0.9954 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 6.489 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4405 +t=10: Selected seed 195 with value = 0.4405 +Query 1/1: Action query time = 2.914 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5149 +t=26: Selected seed 195 with value = 0.5149 +Query 1/1: Action query time = 4.244 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6045 +t=42: Selected seed 195 with value = 0.6045 +Query 1/1: Action query time = 3.600 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6535 +t=58: Selected seed 195 with value = 0.6535 +Query 1/1: Action query time = 5.278 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7777 +t=74: Selected seed 195 with value = 0.7777 +Query 1/1: Action query time = 5.603 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8847 +t=90: Selected seed 195 with value = 0.8847 +Query 1/1: Action query time = 6.456 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9522 +t=106: Selected seed 195 with value = 0.9522 +Query 1/1: Action query time = 4.622 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=122: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 5.577 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.642 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=154: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 5.742 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.182 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 4.599 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4419 +t=10: Selected seed 195 with value = 0.4419 +Query 1/1: Action query time = 5.593 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4892 +t=26: Selected seed 195 with value = 0.4892 +Query 1/1: Action query time = 3.725 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5395 +t=42: Selected seed 195 with value = 0.5395 +Query 1/1: Action query time = 4.906 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6398 +t=58: Selected seed 195 with value = 0.6398 +Query 1/1: Action query time = 5.121 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8632 +t=74: Selected seed 195 with value = 0.8632 +Query 1/1: Action query time = 5.620 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8933 +t=90: Selected seed 195 with value = 0.8933 +Query 1/1: Action query time = 5.446 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9864 +t=106: Selected seed 195 with value = 0.9864 +Query 1/1: Action query time = 6.268 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.585 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9896 +t=138: Selected seed 195 with value = 0.9896 +Query 1/1: Action query time = 4.824 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=154: Selected seed 195 with value = 0.9924 +Query 1/1: Action query time = 3.431 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.214 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9895 +t=186: Selected seed 195 with value = 0.9895 +Query 1/1: Action query time = 4.145 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9814 +t=202: Selected seed 195 with value = 0.9814 +Query 1/1: Action query time = 4.559 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9900 +t=218: Selected seed 195 with value = 0.9900 +Query 1/1: Action query time = 5.027 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=234: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 4.004 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9930 +t=250: Selected seed 195 with value = 0.9930 +Query 1/1: Action query time = 4.672 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9900 +t=266: Selected seed 195 with value = 0.9900 +Query 1/1: Action query time = 4.575 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9901 +t=282: Selected seed 195 with value = 0.9901 +Query 1/1: Action query time = 5.908 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9908 +t=298: Selected seed 195 with value = 0.9908 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=4--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 5... +Query 1/1: Action query time = 5.405 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4375 +t=10: Selected seed 195 with value = 0.4375 +Query 1/1: Action query time = 4.456 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5115 +t=26: Selected seed 195 with value = 0.5115 +Query 1/1: Action query time = 4.745 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6080 +t=42: Selected seed 195 with value = 0.6080 +Query 1/1: Action query time = 5.301 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7136 +t=58: Selected seed 195 with value = 0.7136 +Query 1/1: Action query time = 4.561 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8786 +t=74: Selected seed 195 with value = 0.8786 +Query 1/1: Action query time = 4.373 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9348 +t=90: Selected seed 195 with value = 0.9348 +Query 1/1: Action query time = 3.454 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9806 +t=106: Selected seed 195 with value = 0.9806 +Query 1/1: Action query time = 5.994 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.541 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.654 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.684 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.873 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 6.219 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9800 +t=202: Selected seed 195 with value = 0.9800 +Query 1/1: Action query time = 4.873 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=218: Selected seed 195 with value = 0.9909 +Query 1/1: Action query time = 4.166 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=234: Selected seed 195 with value = 0.9971 +Query 1/1: Action query time = 5.031 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9970 +t=250: Selected seed 195 with value = 0.9970 +Query 1/1: Action query time = 3.455 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=266: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 3.266 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=282: Selected seed 195 with value = 0.9990 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=5--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 6... +Query 1/1: Action query time = 6.093 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4417 +t=10: Selected seed 195 with value = 0.4417 +Query 1/1: Action query time = 5.600 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5122 +t=26: Selected seed 195 with value = 0.5122 +Query 1/1: Action query time = 5.645 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6048 +t=42: Selected seed 195 with value = 0.6048 +Query 1/1: Action query time = 4.852 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6475 +t=58: Selected seed 195 with value = 0.6475 +Query 1/1: Action query time = 3.537 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7645 +t=74: Selected seed 195 with value = 0.7645 +Query 1/1: Action query time = 5.376 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8914 +t=90: Selected seed 195 with value = 0.8914 +Query 1/1: Action query time = 5.312 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9596 +t=106: Selected seed 195 with value = 0.9596 +Query 1/1: Action query time = 3.534 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=122: Selected seed 195 with value = 0.9963 +Query 1/1: Action query time = 3.227 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9936 +t=138: Selected seed 195 with value = 0.9936 +Query 1/1: Action query time = 3.252 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9707 +t=154: Selected seed 195 with value = 0.9707 +Query 1/1: Action query time = 3.652 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9267 +t=170: Selected seed 195 with value = 0.9267 +Query 1/1: Action query time = 4.001 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9398 +t=186: Selected seed 195 with value = 0.9398 +Query 1/1: Action query time = 4.295 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9746 +t=202: Selected seed 195 with value = 0.9746 +Query 1/1: Action query time = 5.815 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=218: Selected seed 195 with value = 0.9867 +Query 1/1: Action query time = 4.883 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9744 +t=234: Selected seed 195 with value = 0.9744 +Query 1/1: Action query time = 4.894 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9694 +t=250: Selected seed 195 with value = 0.9694 +Query 1/1: Action query time = 3.043 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9682 +t=266: Selected seed 195 with value = 0.9682 +Query 1/1: Action query time = 4.097 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9649 +t=282: Selected seed 195 with value = 0.9649 +Query 1/1: Action query time = 5.065 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9625 +t=298: Selected seed 195 with value = 0.9625 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=6--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 6 +# successes: 5 (83.3%) + +Task: push the plate to the front of the stove +Starting episode 7... +Query 1/1: Action query time = 5.311 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4480 +t=10: Selected seed 195 with value = 0.4480 +Query 1/1: Action query time = 4.357 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5219 +t=26: Selected seed 195 with value = 0.5219 +Query 1/1: Action query time = 5.417 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6270 +t=42: Selected seed 195 with value = 0.6270 +Query 1/1: Action query time = 3.510 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6963 +t=58: Selected seed 195 with value = 0.6963 +Query 1/1: Action query time = 3.050 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7180 +t=74: Selected seed 195 with value = 0.7180 +Query 1/1: Action query time = 4.105 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7733 +t=90: Selected seed 195 with value = 0.7733 +Query 1/1: Action query time = 3.748 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8866 +t=106: Selected seed 195 with value = 0.8866 +Query 1/1: Action query time = 3.735 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9526 +t=122: Selected seed 195 with value = 0.9526 +Query 1/1: Action query time = 4.003 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9786 +t=138: Selected seed 195 with value = 0.9786 +Query 1/1: Action query time = 4.596 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.243 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=170: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 5.332 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.249 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.135 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9736 +t=218: Selected seed 195 with value = 0.9736 +Query 1/1: Action query time = 4.725 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9668 +t=234: Selected seed 195 with value = 0.9668 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=7--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 7 +# successes: 6 (85.7%) + +Task: push the plate to the front of the stove +Starting episode 8... +Query 1/1: Action query time = 4.960 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4402 +t=10: Selected seed 195 with value = 0.4402 +Query 1/1: Action query time = 4.663 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5087 +t=26: Selected seed 195 with value = 0.5087 +Query 1/1: Action query time = 4.829 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5836 +t=42: Selected seed 195 with value = 0.5836 +Query 1/1: Action query time = 4.688 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7063 +t=58: Selected seed 195 with value = 0.7063 +Query 1/1: Action query time = 4.133 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8446 +t=74: Selected seed 195 with value = 0.8446 +Query 1/1: Action query time = 4.713 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9649 +t=90: Selected seed 195 with value = 0.9649 +Query 1/1: Action query time = 5.327 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.761 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=8--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 8 +# successes: 7 (87.5%) + +Task: push the plate to the front of the stove +Starting episode 9... +Query 1/1: Action query time = 4.156 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3955 +t=10: Selected seed 195 with value = 0.3955 +Query 1/1: Action query time = 3.511 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5018 +t=26: Selected seed 195 with value = 0.5018 +Query 1/1: Action query time = 5.035 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6043 +t=42: Selected seed 195 with value = 0.6043 +Query 1/1: Action query time = 5.014 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6679 +t=58: Selected seed 195 with value = 0.6679 +Query 1/1: Action query time = 5.404 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8488 +t=74: Selected seed 195 with value = 0.8488 +Query 1/1: Action query time = 4.495 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9377 +t=90: Selected seed 195 with value = 0.9377 +Query 1/1: Action query time = 5.095 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=106: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 3.454 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.741 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.341 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.211 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=9--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 9 +# successes: 8 (88.9%) + +Task: push the plate to the front of the stove +Starting episode 10... +Query 1/1: Action query time = 3.998 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4372 +t=10: Selected seed 195 with value = 0.4372 +Query 1/1: Action query time = 4.722 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4950 +t=26: Selected seed 195 with value = 0.4950 +Query 1/1: Action query time = 4.737 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5644 +t=42: Selected seed 195 with value = 0.5644 +Query 1/1: Action query time = 5.299 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6514 +t=58: Selected seed 195 with value = 0.6514 +Query 1/1: Action query time = 3.030 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8173 +t=74: Selected seed 195 with value = 0.8173 +Query 1/1: Action query time = 2.961 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9193 +t=90: Selected seed 195 with value = 0.9193 +Query 1/1: Action query time = 3.077 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9631 +t=106: Selected seed 195 with value = 0.9631 +Query 1/1: Action query time = 3.569 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9896 +t=122: Selected seed 195 with value = 0.9896 +Query 1/1: Action query time = 4.091 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=138: Selected seed 195 with value = 0.9983 +Query 1/1: Action query time = 4.629 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.598 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=10--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 10 +# successes: 9 (90.0%) + +Task: push the plate to the front of the stove +Starting episode 11... +Query 1/1: Action query time = 3.977 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4565 +t=10: Selected seed 195 with value = 0.4565 +Query 1/1: Action query time = 3.830 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5172 +t=26: Selected seed 195 with value = 0.5172 +Query 1/1: Action query time = 3.800 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6041 +t=42: Selected seed 195 with value = 0.6041 +Query 1/1: Action query time = 3.418 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7021 +t=58: Selected seed 195 with value = 0.7021 +Query 1/1: Action query time = 3.419 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8664 +t=74: Selected seed 195 with value = 0.8664 +Query 1/1: Action query time = 3.230 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9301 +t=90: Selected seed 195 with value = 0.9301 +Query 1/1: Action query time = 3.919 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9814 +t=106: Selected seed 195 with value = 0.9814 +Query 1/1: Action query time = 4.244 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.667 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.350 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.409 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.099 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.256 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.527 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.562 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=234: Selected seed 195 with value = 0.9999 +Query 1/1: Action query time = 4.282 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9936 +t=250: Selected seed 195 with value = 0.9936 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=11--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 11 +# successes: 10 (90.9%) + +Task: push the plate to the front of the stove +Starting episode 12... +Query 1/1: Action query time = 3.745 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4231 +t=10: Selected seed 195 with value = 0.4231 +Query 1/1: Action query time = 2.959 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5296 +t=26: Selected seed 195 with value = 0.5296 +Query 1/1: Action query time = 3.272 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6308 +t=42: Selected seed 195 with value = 0.6308 +Query 1/1: Action query time = 3.488 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7942 +t=58: Selected seed 195 with value = 0.7942 +Query 1/1: Action query time = 2.611 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8882 +t=74: Selected seed 195 with value = 0.8882 +Query 1/1: Action query time = 2.472 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9555 +t=90: Selected seed 195 with value = 0.9555 +Query 1/1: Action query time = 2.713 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9936 +t=106: Selected seed 195 with value = 0.9936 +Query 1/1: Action query time = 1.964 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.102 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.879 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.225 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.666 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.865 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9689 +t=202: Selected seed 195 with value = 0.9689 +Query 1/1: Action query time = 1.769 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9637 +t=218: Selected seed 195 with value = 0.9637 +Query 1/1: Action query time = 2.156 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9682 +t=234: Selected seed 195 with value = 0.9682 +Query 1/1: Action query time = 2.621 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9694 +t=250: Selected seed 195 with value = 0.9694 +Query 1/1: Action query time = 2.169 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9746 +t=266: Selected seed 195 with value = 0.9746 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=12--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 12 +# successes: 11 (91.7%) + +Task: push the plate to the front of the stove +Starting episode 13... +Query 1/1: Action query time = 1.364 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4431 +t=10: Selected seed 195 with value = 0.4431 +Query 1/1: Action query time = 2.022 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5325 +t=26: Selected seed 195 with value = 0.5325 +Query 1/1: Action query time = 2.053 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6159 +t=42: Selected seed 195 with value = 0.6159 +Query 1/1: Action query time = 2.272 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7477 +t=58: Selected seed 195 with value = 0.7477 +Query 1/1: Action query time = 2.044 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7853 +t=74: Selected seed 195 with value = 0.7853 +Query 1/1: Action query time = 1.213 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8866 +t=90: Selected seed 195 with value = 0.8866 +Query 1/1: Action query time = 1.760 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9688 +t=106: Selected seed 195 with value = 0.9688 +Query 1/1: Action query time = 2.074 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9851 +t=122: Selected seed 195 with value = 0.9851 +Query 1/1: Action query time = 2.360 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9724 +t=138: Selected seed 195 with value = 0.9724 +Query 1/1: Action query time = 2.112 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.040 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9313 +t=170: Selected seed 195 with value = 0.9313 +Query 1/1: Action query time = 1.797 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9382 +t=186: Selected seed 195 with value = 0.9382 +Query 1/1: Action query time = 1.680 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9493 +t=202: Selected seed 195 with value = 0.9493 +Query 1/1: Action query time = 1.607 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9491 +t=218: Selected seed 195 with value = 0.9491 +Query 1/1: Action query time = 0.976 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9587 +t=234: Selected seed 195 with value = 0.9587 +Query 1/1: Action query time = 1.018 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9790 +t=250: Selected seed 195 with value = 0.9790 +Query 1/1: Action query time = 0.983 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9741 +t=266: Selected seed 195 with value = 0.9741 +Query 1/1: Action query time = 0.992 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9716 +t=282: Selected seed 195 with value = 0.9716 +Query 1/1: Action query time = 1.007 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9837 +t=298: Selected seed 195 with value = 0.9837 +Saved rollout MP4 at path ./rollouts/t7dc800x50_t5_s1/2026_08_03-01_47_46--with_future_img--episode=13--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 13 +# successes: 11 (84.6%) +Current task success rate: 0.8461538461538461 +Current total success rate: 0.8461538461538461 +Final results: +Total episodes: 13 +Total successes: 11 +Overall success rate: 0.8462 (84.6%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_46--t7rc800x50_t6_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_46--t7rc800x50_t6_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..72969c4d0bc1151f25e5b26cefd01f170c6c10c1 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-01_47_46--t7rc800x50_t6_s1.txt @@ -0,0 +1,859 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='6', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7rc800x50_t6_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, simple_task_max_steps=100, complex_task_max_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [6] +Using default initial states + +Task: put the cream cheese in the bowl +Starting episode 1... +Query 1/1: Action query time = 8.467 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4698 +t=10: Selected seed 195 with value = 0.4698 +Query 1/1: Action query time = 6.841 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5705 +t=26: Selected seed 195 with value = 0.5705 +Query 1/1: Action query time = 5.357 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6767 +t=42: Selected seed 195 with value = 0.6767 +Query 1/1: Action query time = 3.898 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7691 +t=58: Selected seed 195 with value = 0.7691 +Query 1/1: Action query time = 5.683 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7245 +t=74: Selected seed 195 with value = 0.7245 +Query 1/1: Action query time = 3.193 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7843 +t=90: Selected seed 195 with value = 0.7843 +Query 1/1: Action query time = 6.433 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8675 +t=106: Selected seed 195 with value = 0.8675 +Query 1/1: Action query time = 5.299 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9769 +t=122: Selected seed 195 with value = 0.9769 +Query 1/1: Action query time = 5.345 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8316 +t=138: Selected seed 195 with value = 0.8316 +Query 1/1: Action query time = 5.375 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9369 +t=154: Selected seed 195 with value = 0.9369 +Query 1/1: Action query time = 6.003 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9613 +t=170: Selected seed 195 with value = 0.9613 +Query 1/1: Action query time = 4.898 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9788 +t=186: Selected seed 195 with value = 0.9788 +Query 1/1: Action query time = 4.804 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9891 +t=202: Selected seed 195 with value = 0.9891 +Query 1/1: Action query time = 3.970 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9472 +t=218: Selected seed 195 with value = 0.9472 +Query 1/1: Action query time = 4.804 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9893 +t=234: Selected seed 195 with value = 0.9893 +Query 1/1: Action query time = 3.862 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9160 +t=250: Selected seed 195 with value = 0.9160 +Query 1/1: Action query time = 4.049 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9646 +t=266: Selected seed 195 with value = 0.9646 +Query 1/1: Action query time = 5.131 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9128 +t=282: Selected seed 195 with value = 0.9128 +Query 1/1: Action query time = 5.511 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8576 +t=298: Selected seed 195 with value = 0.8576 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=1--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the cream cheese in the bowl +Starting episode 2... +Query 1/1: Action query time = 6.333 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5137 +t=10: Selected seed 195 with value = 0.5137 +Query 1/1: Action query time = 6.348 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6071 +t=26: Selected seed 195 with value = 0.6071 +Query 1/1: Action query time = 5.066 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7126 +t=42: Selected seed 195 with value = 0.7126 +Query 1/1: Action query time = 4.042 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7526 +t=58: Selected seed 195 with value = 0.7526 +Query 1/1: Action query time = 4.453 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8913 +t=74: Selected seed 195 with value = 0.8913 +Query 1/1: Action query time = 5.140 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=2--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 3... +Query 1/1: Action query time = 5.560 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4794 +t=10: Selected seed 195 with value = 0.4794 +Query 1/1: Action query time = 4.722 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5405 +t=26: Selected seed 195 with value = 0.5405 +Query 1/1: Action query time = 5.118 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6747 +t=42: Selected seed 195 with value = 0.6747 +Query 1/1: Action query time = 5.059 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7907 +t=58: Selected seed 195 with value = 0.7907 +Query 1/1: Action query time = 6.131 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8888 +t=74: Selected seed 195 with value = 0.8888 +Query 1/1: Action query time = 4.563 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=90: Selected seed 195 with value = 0.9914 +Query 1/1: Action query time = 5.147 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9925 +t=106: Selected seed 195 with value = 0.9925 +Query 1/1: Action query time = 4.352 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9911 +t=122: Selected seed 195 with value = 0.9911 +Query 1/1: Action query time = 3.655 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9667 +t=138: Selected seed 195 with value = 0.9667 +Query 1/1: Action query time = 6.133 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9556 +t=154: Selected seed 195 with value = 0.9556 +Query 1/1: Action query time = 4.205 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9714 +t=170: Selected seed 195 with value = 0.9714 +Query 1/1: Action query time = 4.576 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9676 +t=186: Selected seed 195 with value = 0.9676 +Query 1/1: Action query time = 5.098 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=202: Selected seed 195 with value = 0.9982 +Query 1/1: Action query time = 5.122 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9735 +t=218: Selected seed 195 with value = 0.9735 +Query 1/1: Action query time = 5.751 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.431 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=250: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 5.071 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9682 +t=266: Selected seed 195 with value = 0.9682 +Query 1/1: Action query time = 5.600 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9054 +t=282: Selected seed 195 with value = 0.9054 +Query 1/1: Action query time = 4.669 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9302 +t=298: Selected seed 195 with value = 0.9302 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=3--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 3 +# successes: 1 (33.3%) + +Task: put the cream cheese in the bowl +Starting episode 4... +Query 1/1: Action query time = 4.893 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4935 +t=10: Selected seed 195 with value = 0.4935 +Query 1/1: Action query time = 5.930 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5560 +t=26: Selected seed 195 with value = 0.5560 +Query 1/1: Action query time = 5.451 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6800 +t=42: Selected seed 195 with value = 0.6800 +Query 1/1: Action query time = 2.491 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7944 +t=58: Selected seed 195 with value = 0.7944 +Query 1/1: Action query time = 5.133 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8967 +t=74: Selected seed 195 with value = 0.8967 +Query 1/1: Action query time = 4.892 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9712 +t=90: Selected seed 195 with value = 0.9712 +Query 1/1: Action query time = 5.457 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.719 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=122: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 4.156 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8230 +t=138: Selected seed 195 with value = 0.8230 +Query 1/1: Action query time = 5.283 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9838 +t=154: Selected seed 195 with value = 0.9838 +Query 1/1: Action query time = 4.533 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9843 +t=170: Selected seed 195 with value = 0.9843 +Query 1/1: Action query time = 2.514 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9922 +t=186: Selected seed 195 with value = 0.9922 +Query 1/1: Action query time = 5.452 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8674 +t=202: Selected seed 195 with value = 0.8674 +Query 1/1: Action query time = 4.798 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9599 +t=218: Selected seed 195 with value = 0.9599 +Query 1/1: Action query time = 3.958 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9921 +t=234: Selected seed 195 with value = 0.9921 +Query 1/1: Action query time = 4.204 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9404 +t=250: Selected seed 195 with value = 0.9404 +Query 1/1: Action query time = 5.833 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.872 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9663 +t=282: Selected seed 195 with value = 0.9663 +Query 1/1: Action query time = 5.080 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=298: Selected seed 195 with value = 0.9997 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=4--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 4 +# successes: 1 (25.0%) + +Task: put the cream cheese in the bowl +Starting episode 5... +Query 1/1: Action query time = 5.374 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4937 +t=10: Selected seed 195 with value = 0.4937 +Query 1/1: Action query time = 4.697 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5690 +t=26: Selected seed 195 with value = 0.5690 +Query 1/1: Action query time = 4.594 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6864 +t=42: Selected seed 195 with value = 0.6864 +Query 1/1: Action query time = 3.638 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7694 +t=58: Selected seed 195 with value = 0.7694 +Query 1/1: Action query time = 5.166 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8881 +t=74: Selected seed 195 with value = 0.8881 +Query 1/1: Action query time = 5.452 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9949 +t=90: Selected seed 195 with value = 0.9949 +Query 1/1: Action query time = 5.135 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=106: Selected seed 195 with value = 0.9963 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=5--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 5 +# successes: 2 (40.0%) + +Task: put the cream cheese in the bowl +Starting episode 6... +Query 1/1: Action query time = 4.285 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4515 +t=10: Selected seed 195 with value = 0.4515 +Query 1/1: Action query time = 3.751 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5499 +t=26: Selected seed 195 with value = 0.5499 +Query 1/1: Action query time = 2.814 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6868 +t=42: Selected seed 195 with value = 0.6868 +Query 1/1: Action query time = 3.679 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7847 +t=58: Selected seed 195 with value = 0.7847 +Query 1/1: Action query time = 3.156 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7324 +t=74: Selected seed 195 with value = 0.7324 +Query 1/1: Action query time = 2.445 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8192 +t=90: Selected seed 195 with value = 0.8192 +Query 1/1: Action query time = 2.047 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9656 +t=106: Selected seed 195 with value = 0.9656 +Query 1/1: Action query time = 4.029 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.171 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=138: Selected seed 195 with value = 0.9871 +Query 1/1: Action query time = 4.114 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9792 +t=154: Selected seed 195 with value = 0.9792 +Query 1/1: Action query time = 3.448 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9720 +t=170: Selected seed 195 with value = 0.9720 +Query 1/1: Action query time = 4.037 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=186: Selected seed 195 with value = 0.9886 +Query 1/1: Action query time = 4.270 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9540 +t=202: Selected seed 195 with value = 0.9540 +Query 1/1: Action query time = 3.694 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9077 +t=218: Selected seed 195 with value = 0.9077 +Query 1/1: Action query time = 4.527 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9169 +t=234: Selected seed 195 with value = 0.9169 +Query 1/1: Action query time = 3.751 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9528 +t=250: Selected seed 195 with value = 0.9528 +Query 1/1: Action query time = 3.075 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9655 +t=266: Selected seed 195 with value = 0.9655 +Query 1/1: Action query time = 3.112 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9338 +t=282: Selected seed 195 with value = 0.9338 +Query 1/1: Action query time = 2.861 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9031 +t=298: Selected seed 195 with value = 0.9031 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=6--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 6 +# successes: 2 (33.3%) + +Task: put the cream cheese in the bowl +Starting episode 7... +Query 1/1: Action query time = 3.228 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4848 +t=10: Selected seed 195 with value = 0.4848 +Query 1/1: Action query time = 3.565 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5701 +t=26: Selected seed 195 with value = 0.5701 +Query 1/1: Action query time = 3.601 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6628 +t=42: Selected seed 195 with value = 0.6628 +Query 1/1: Action query time = 3.418 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7849 +t=58: Selected seed 195 with value = 0.7849 +Query 1/1: Action query time = 3.904 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8923 +t=74: Selected seed 195 with value = 0.8923 +Query 1/1: Action query time = 4.060 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9846 +t=90: Selected seed 195 with value = 0.9846 +Query 1/1: Action query time = 4.202 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.746 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9956 +t=122: Selected seed 195 with value = 0.9956 +Query 1/1: Action query time = 2.859 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9866 +t=138: Selected seed 195 with value = 0.9866 +Query 1/1: Action query time = 2.856 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9784 +t=154: Selected seed 195 with value = 0.9784 +Query 1/1: Action query time = 3.559 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9767 +t=170: Selected seed 195 with value = 0.9767 +Query 1/1: Action query time = 3.420 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9217 +t=186: Selected seed 195 with value = 0.9217 +Query 1/1: Action query time = 4.116 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9270 +t=202: Selected seed 195 with value = 0.9270 +Query 1/1: Action query time = 4.372 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9634 +t=218: Selected seed 195 with value = 0.9634 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=7--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 7 +# successes: 3 (42.9%) + +Task: put the cream cheese in the bowl +Starting episode 8... +Query 1/1: Action query time = 3.626 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4949 +t=10: Selected seed 195 with value = 0.4949 +Query 1/1: Action query time = 3.000 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5719 +t=26: Selected seed 195 with value = 0.5719 +Query 1/1: Action query time = 4.547 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6830 +t=42: Selected seed 195 with value = 0.6830 +Query 1/1: Action query time = 4.409 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7637 +t=58: Selected seed 195 with value = 0.7637 +Query 1/1: Action query time = 3.439 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9011 +t=74: Selected seed 195 with value = 0.9011 +Query 1/1: Action query time = 2.755 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=90: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 2.226 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.797 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=122: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 2.048 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=138: Selected seed 195 with value = 0.9946 +Query 1/1: Action query time = 3.560 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9783 +t=154: Selected seed 195 with value = 0.9783 +Query 1/1: Action query time = 3.461 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.208 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9655 +t=186: Selected seed 195 with value = 0.9655 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=8--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 8 +# successes: 4 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 9... +Query 1/1: Action query time = 3.050 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4952 +t=10: Selected seed 195 with value = 0.4952 +Query 1/1: Action query time = 2.356 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5490 +t=26: Selected seed 195 with value = 0.5490 +Query 1/1: Action query time = 1.993 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6946 +t=42: Selected seed 195 with value = 0.6946 +Query 1/1: Action query time = 3.935 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7691 +t=58: Selected seed 195 with value = 0.7691 +Query 1/1: Action query time = 3.813 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8834 +t=74: Selected seed 195 with value = 0.8834 +Query 1/1: Action query time = 3.552 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9861 +t=90: Selected seed 195 with value = 0.9861 +Query 1/1: Action query time = 3.755 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=106: Selected seed 195 with value = 0.9976 +Query 1/1: Action query time = 3.749 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9889 +t=122: Selected seed 195 with value = 0.9889 +Query 1/1: Action query time = 3.448 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9787 +t=138: Selected seed 195 with value = 0.9787 +Query 1/1: Action query time = 3.638 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=154: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 3.165 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8610 +t=170: Selected seed 195 with value = 0.8610 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=9--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 9 +# successes: 5 (55.6%) + +Task: put the cream cheese in the bowl +Starting episode 10... +Query 1/1: Action query time = 1.979 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4893 +t=10: Selected seed 195 with value = 0.4893 +Query 1/1: Action query time = 2.843 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5653 +t=26: Selected seed 195 with value = 0.5653 +Query 1/1: Action query time = 2.471 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6790 +t=42: Selected seed 195 with value = 0.6790 +Query 1/1: Action query time = 2.581 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7703 +t=58: Selected seed 195 with value = 0.7703 +Query 1/1: Action query time = 3.089 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7233 +t=74: Selected seed 195 with value = 0.7233 +Query 1/1: Action query time = 2.780 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7872 +t=90: Selected seed 195 with value = 0.7872 +Query 1/1: Action query time = 2.129 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8725 +t=106: Selected seed 195 with value = 0.8725 +Query 1/1: Action query time = 3.416 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9082 +t=122: Selected seed 195 with value = 0.9082 +Query 1/1: Action query time = 3.396 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9163 +t=138: Selected seed 195 with value = 0.9163 +Query 1/1: Action query time = 3.033 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9630 +t=154: Selected seed 195 with value = 0.9630 +Query 1/1: Action query time = 2.735 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=170: Selected seed 195 with value = 0.9924 +Query 1/1: Action query time = 2.646 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9970 +t=186: Selected seed 195 with value = 0.9970 +Query 1/1: Action query time = 2.911 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9669 +t=202: Selected seed 195 with value = 0.9669 +Query 1/1: Action query time = 2.739 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9697 +t=218: Selected seed 195 with value = 0.9697 +Query 1/1: Action query time = 3.262 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9682 +t=234: Selected seed 195 with value = 0.9682 +Query 1/1: Action query time = 2.422 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8820 +t=250: Selected seed 195 with value = 0.8820 +Query 1/1: Action query time = 2.494 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9752 +t=266: Selected seed 195 with value = 0.9752 +Query 1/1: Action query time = 2.155 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9738 +t=282: Selected seed 195 with value = 0.9738 +Query 1/1: Action query time = 2.582 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9528 +t=298: Selected seed 195 with value = 0.9528 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=10--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 10 +# successes: 5 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 11... +Query 1/1: Action query time = 2.821 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5102 +t=10: Selected seed 195 with value = 0.5102 +Query 1/1: Action query time = 3.463 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5829 +t=26: Selected seed 195 with value = 0.5829 +Query 1/1: Action query time = 2.296 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6856 +t=42: Selected seed 195 with value = 0.6856 +Query 1/1: Action query time = 3.321 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7622 +t=58: Selected seed 195 with value = 0.7622 +Query 1/1: Action query time = 3.325 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8971 +t=74: Selected seed 195 with value = 0.8971 +Query 1/1: Action query time = 3.573 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=90: Selected seed 195 with value = 0.9924 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=11--success=True--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: True +# episodes completed so far: 11 +# successes: 6 (54.5%) + +Task: put the cream cheese in the bowl +Starting episode 12... +Query 1/1: Action query time = 4.008 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5189 +t=10: Selected seed 195 with value = 0.5189 +Query 1/1: Action query time = 2.978 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5957 +t=26: Selected seed 195 with value = 0.5957 +Query 1/1: Action query time = 2.098 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6841 +t=42: Selected seed 195 with value = 0.6841 +Query 1/1: Action query time = 2.584 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8470 +t=58: Selected seed 195 with value = 0.8470 +Query 1/1: Action query time = 2.741 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8258 +t=74: Selected seed 195 with value = 0.8258 +Query 1/1: Action query time = 2.815 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8224 +t=90: Selected seed 195 with value = 0.8224 +Query 1/1: Action query time = 3.899 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9557 +t=106: Selected seed 195 with value = 0.9557 +Query 1/1: Action query time = 2.349 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9820 +t=122: Selected seed 195 with value = 0.9820 +Query 1/1: Action query time = 3.226 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9940 +t=138: Selected seed 195 with value = 0.9940 +Query 1/1: Action query time = 2.783 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9927 +t=154: Selected seed 195 with value = 0.9927 +Query 1/1: Action query time = 2.781 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9797 +t=170: Selected seed 195 with value = 0.9797 +Query 1/1: Action query time = 2.674 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9359 +t=186: Selected seed 195 with value = 0.9359 +Query 1/1: Action query time = 2.940 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9410 +t=202: Selected seed 195 with value = 0.9410 +Query 1/1: Action query time = 2.428 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9630 +t=218: Selected seed 195 with value = 0.9630 +Query 1/1: Action query time = 2.323 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9634 +t=234: Selected seed 195 with value = 0.9634 +Query 1/1: Action query time = 3.496 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9677 +t=250: Selected seed 195 with value = 0.9677 +Query 1/1: Action query time = 2.669 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9464 +t=266: Selected seed 195 with value = 0.9464 +Query 1/1: Action query time = 2.474 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9350 +t=282: Selected seed 195 with value = 0.9350 +Query 1/1: Action query time = 2.856 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9126 +t=298: Selected seed 195 with value = 0.9126 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=12--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 12 +# successes: 6 (50.0%) + +Task: put the cream cheese in the bowl +Starting episode 13... +Query 1/1: Action query time = 2.130 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5078 +t=10: Selected seed 195 with value = 0.5078 +Query 1/1: Action query time = 3.227 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5911 +t=26: Selected seed 195 with value = 0.5911 +Query 1/1: Action query time = 2.519 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6800 +t=42: Selected seed 195 with value = 0.6800 +Query 1/1: Action query time = 2.308 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8027 +t=58: Selected seed 195 with value = 0.8027 +Query 1/1: Action query time = 2.658 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8485 +t=74: Selected seed 195 with value = 0.8485 +Query 1/1: Action query time = 2.840 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7986 +t=90: Selected seed 195 with value = 0.7986 +Query 1/1: Action query time = 2.294 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8502 +t=106: Selected seed 195 with value = 0.8502 +Query 1/1: Action query time = 2.659 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9602 +t=122: Selected seed 195 with value = 0.9602 +Query 1/1: Action query time = 2.543 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9921 +t=138: Selected seed 195 with value = 0.9921 +Query 1/1: Action query time = 2.783 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9089 +t=154: Selected seed 195 with value = 0.9089 +Query 1/1: Action query time = 2.506 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8666 +t=170: Selected seed 195 with value = 0.8666 +Query 1/1: Action query time = 2.751 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9437 +t=186: Selected seed 195 with value = 0.9437 +Query 1/1: Action query time = 2.253 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9572 +t=202: Selected seed 195 with value = 0.9572 +Query 1/1: Action query time = 1.589 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9676 +t=218: Selected seed 195 with value = 0.9676 +Query 1/1: Action query time = 2.102 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9668 +t=234: Selected seed 195 with value = 0.9668 +Query 1/1: Action query time = 1.643 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9456 +t=250: Selected seed 195 with value = 0.9456 +Query 1/1: Action query time = 1.556 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9247 +t=266: Selected seed 195 with value = 0.9247 +Query 1/1: Action query time = 1.585 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9308 +t=282: Selected seed 195 with value = 0.9308 +Query 1/1: Action query time = 1.729 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8989 +t=298: Selected seed 195 with value = 0.8989 +Saved rollout MP4 at path ./rollouts/t7rc800x50_t6_s1/2026_08_03-01_47_46--with_future_img--episode=13--success=False--task=put_the_cream_cheese_in_the_bowl.mp4 +Success: False +# episodes completed so far: 13 +# successes: 6 (46.2%) +Current task success rate: 0.46153846153846156 +Current total success rate: 0.46153846153846156 +Final results: +Total episodes: 13 +Total successes: 6 +Overall success rate: 0.4615 (46.2%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-10_52_25--t7realcl800x50_t2_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-10_52_25--t7realcl800x50_t2_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..3a8a3b29684c54638fa7818aa39c1f63486028d1 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-10_52_25--t7realcl800x50_t2_s0.txt @@ -0,0 +1,959 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_real_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7realcl800x50_t2_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.948 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4783 +t=10: Selected seed 195 with value = 0.4783 +Query 1/1: Action query time = 4.496 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5732 +t=26: Selected seed 195 with value = 0.5732 +Query 1/1: Action query time = 4.995 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6733 +t=42: Selected seed 195 with value = 0.6733 +Query 1/1: Action query time = 4.454 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7533 +t=58: Selected seed 195 with value = 0.7533 +Query 1/1: Action query time = 3.790 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8456 +t=74: Selected seed 195 with value = 0.8456 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.288 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4272 +t=10: Selected seed 195 with value = 0.4272 +Query 1/1: Action query time = 3.702 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4862 +t=26: Selected seed 195 with value = 0.4862 +Query 1/1: Action query time = 5.079 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5476 +t=42: Selected seed 195 with value = 0.5476 +Query 1/1: Action query time = 4.999 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6431 +t=58: Selected seed 195 with value = 0.6431 +Query 1/1: Action query time = 4.852 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7671 +t=74: Selected seed 195 with value = 0.7671 +Query 1/1: Action query time = 4.966 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8840 +t=90: Selected seed 195 with value = 0.8840 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.216 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4699 +t=10: Selected seed 195 with value = 0.4699 +Query 1/1: Action query time = 3.269 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5585 +t=26: Selected seed 195 with value = 0.5585 +Query 1/1: Action query time = 5.429 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6299 +t=42: Selected seed 195 with value = 0.6299 +Query 1/1: Action query time = 5.735 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7459 +t=58: Selected seed 195 with value = 0.7459 +Query 1/1: Action query time = 6.116 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8505 +t=74: Selected seed 195 with value = 0.8505 +Query 1/1: Action query time = 3.546 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 3.906 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3097 +t=10: Selected seed 195 with value = 0.3097 +Query 1/1: Action query time = 3.662 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3820 +t=26: Selected seed 195 with value = 0.3820 +Query 1/1: Action query time = 5.159 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4503 +t=42: Selected seed 195 with value = 0.4503 +Query 1/1: Action query time = 4.968 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5604 +t=58: Selected seed 195 with value = 0.5604 +Query 1/1: Action query time = 3.967 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5934 +t=74: Selected seed 195 with value = 0.5934 +Query 1/1: Action query time = 3.614 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6967 +t=90: Selected seed 195 with value = 0.6967 +Query 1/1: Action query time = 3.686 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8579 +t=106: Selected seed 195 with value = 0.8579 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=4--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 5... +Query 1/1: Action query time = 5.211 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3557 +t=10: Selected seed 195 with value = 0.3557 +Query 1/1: Action query time = 4.325 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4511 +t=26: Selected seed 195 with value = 0.4511 +Query 1/1: Action query time = 5.366 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5444 +t=42: Selected seed 195 with value = 0.5444 +Query 1/1: Action query time = 5.514 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6700 +t=58: Selected seed 195 with value = 0.6700 +Query 1/1: Action query time = 5.080 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8100 +t=74: Selected seed 195 with value = 0.8100 +Query 1/1: Action query time = 3.479 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9435 +t=90: Selected seed 195 with value = 0.9435 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=5--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 6... +Query 1/1: Action query time = 4.490 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4065 +t=10: Selected seed 195 with value = 0.4065 +Query 1/1: Action query time = 4.232 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4524 +t=26: Selected seed 195 with value = 0.4524 +Query 1/1: Action query time = 4.869 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5567 +t=42: Selected seed 195 with value = 0.5567 +Query 1/1: Action query time = 4.447 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6787 +t=58: Selected seed 195 with value = 0.6787 +Query 1/1: Action query time = 5.200 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8477 +t=74: Selected seed 195 with value = 0.8477 +Query 1/1: Action query time = 4.775 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=90: Selected seed 195 with value = 0.9965 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=6--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 7... +Query 1/1: Action query time = 4.799 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4712 +t=10: Selected seed 195 with value = 0.4712 +Query 1/1: Action query time = 4.609 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5388 +t=26: Selected seed 195 with value = 0.5388 +Query 1/1: Action query time = 4.233 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6287 +t=42: Selected seed 195 with value = 0.6287 +Query 1/1: Action query time = 4.534 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7253 +t=58: Selected seed 195 with value = 0.7253 +Query 1/1: Action query time = 4.395 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8546 +t=74: Selected seed 195 with value = 0.8546 +Query 1/1: Action query time = 3.620 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=7--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 8... +Query 1/1: Action query time = 3.466 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4185 +t=10: Selected seed 195 with value = 0.4185 +Query 1/1: Action query time = 3.743 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5066 +t=26: Selected seed 195 with value = 0.5066 +Query 1/1: Action query time = 4.707 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5886 +t=42: Selected seed 195 with value = 0.5886 +Query 1/1: Action query time = 5.246 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6812 +t=58: Selected seed 195 with value = 0.6812 +Query 1/1: Action query time = 5.089 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8704 +t=74: Selected seed 195 with value = 0.8704 +Query 1/1: Action query time = 5.175 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=90: Selected seed 195 with value = 0.9965 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=8--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 9... +Query 1/1: Action query time = 3.923 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3948 +t=10: Selected seed 195 with value = 0.3948 +Query 1/1: Action query time = 5.321 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4296 +t=26: Selected seed 195 with value = 0.4296 +Query 1/1: Action query time = 4.683 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5304 +t=42: Selected seed 195 with value = 0.5304 +Query 1/1: Action query time = 4.731 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6477 +t=58: Selected seed 195 with value = 0.6477 +Query 1/1: Action query time = 4.140 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7771 +t=74: Selected seed 195 with value = 0.7771 +Query 1/1: Action query time = 3.908 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9157 +t=90: Selected seed 195 with value = 0.9157 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=9--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 10... +Query 1/1: Action query time = 3.879 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4362 +t=10: Selected seed 195 with value = 0.4362 +Query 1/1: Action query time = 4.584 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5056 +t=26: Selected seed 195 with value = 0.5056 +Query 1/1: Action query time = 5.054 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6000 +t=42: Selected seed 195 with value = 0.6000 +Query 1/1: Action query time = 5.090 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7094 +t=58: Selected seed 195 with value = 0.7094 +Query 1/1: Action query time = 5.377 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8671 +t=74: Selected seed 195 with value = 0.8671 +Query 1/1: Action query time = 5.329 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=10--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 11... +Query 1/1: Action query time = 4.466 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4627 +t=10: Selected seed 195 with value = 0.4627 +Query 1/1: Action query time = 5.190 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5557 +t=26: Selected seed 195 with value = 0.5557 +Query 1/1: Action query time = 4.474 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6310 +t=42: Selected seed 195 with value = 0.6310 +Query 1/1: Action query time = 3.167 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7618 +t=58: Selected seed 195 with value = 0.7618 +Query 1/1: Action query time = 4.314 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8879 +t=74: Selected seed 195 with value = 0.8879 +Query 1/1: Action query time = 4.864 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=90: Selected seed 195 with value = 0.9999 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=11--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 12... +Query 1/1: Action query time = 4.448 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4372 +t=10: Selected seed 195 with value = 0.4372 +Query 1/1: Action query time = 5.076 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4776 +t=26: Selected seed 195 with value = 0.4776 +Query 1/1: Action query time = 4.713 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5772 +t=42: Selected seed 195 with value = 0.5772 +Query 1/1: Action query time = 4.387 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6282 +t=58: Selected seed 195 with value = 0.6282 +Query 1/1: Action query time = 4.727 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7308 +t=74: Selected seed 195 with value = 0.7308 +Query 1/1: Action query time = 4.909 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8505 +t=90: Selected seed 195 with value = 0.8505 +Query 1/1: Action query time = 3.636 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9918 +t=106: Selected seed 195 with value = 0.9918 +Query 1/1: Action query time = 5.058 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9953 +t=122: Selected seed 195 with value = 0.9953 +Query 1/1: Action query time = 4.041 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=138: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 3.675 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.156 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.578 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.612 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=202: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 1.558 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=218: Selected seed 195 with value = 0.9976 +Query 1/1: Action query time = 4.646 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=234: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 4.835 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9933 +t=250: Selected seed 195 with value = 0.9933 +Query 1/1: Action query time = 4.577 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=266: Selected seed 195 with value = 0.9903 +Query 1/1: Action query time = 4.870 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9862 +t=282: Selected seed 195 with value = 0.9862 +Query 1/1: Action query time = 5.719 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9830 +t=298: Selected seed 195 with value = 0.9830 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=12--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 12 +# successes: 11 (91.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 13... +Query 1/1: Action query time = 3.723 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3154 +t=10: Selected seed 195 with value = 0.3154 +Query 1/1: Action query time = 4.667 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4393 +t=26: Selected seed 195 with value = 0.4393 +Query 1/1: Action query time = 2.939 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4855 +t=42: Selected seed 195 with value = 0.4855 +Query 1/1: Action query time = 3.812 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5653 +t=58: Selected seed 195 with value = 0.5653 +Query 1/1: Action query time = 4.454 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6780 +t=74: Selected seed 195 with value = 0.6780 +Query 1/1: Action query time = 4.222 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8148 +t=90: Selected seed 195 with value = 0.8148 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=13--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 13 +# successes: 12 (92.3%) + +Task: put the wine bottle on top of the cabinet +Starting episode 14... +Query 1/1: Action query time = 4.280 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3274 +t=10: Selected seed 195 with value = 0.3274 +Query 1/1: Action query time = 4.171 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4173 +t=26: Selected seed 195 with value = 0.4173 +Query 1/1: Action query time = 3.826 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5580 +t=42: Selected seed 195 with value = 0.5580 +Query 1/1: Action query time = 4.993 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6334 +t=58: Selected seed 195 with value = 0.6334 +Query 1/1: Action query time = 4.349 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7446 +t=74: Selected seed 195 with value = 0.7446 +Query 1/1: Action query time = 4.041 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8821 +t=90: Selected seed 195 with value = 0.8821 +Query 1/1: Action query time = 4.786 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9732 +t=106: Selected seed 195 with value = 0.9732 +Query 1/1: Action query time = 4.651 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9949 +t=122: Selected seed 195 with value = 0.9949 +Query 1/1: Action query time = 4.398 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=138: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 5.501 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.189 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=170: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 2.894 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=186: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 5.141 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9935 +t=202: Selected seed 195 with value = 0.9935 +Query 1/1: Action query time = 5.085 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9817 +t=218: Selected seed 195 with value = 0.9817 +Query 1/1: Action query time = 4.076 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8706 +t=234: Selected seed 195 with value = 0.8706 +Query 1/1: Action query time = 3.787 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9842 +t=250: Selected seed 195 with value = 0.9842 +Query 1/1: Action query time = 4.712 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.568 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.806 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=298: Selected seed 195 with value = 0.9982 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=14--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 14 +# successes: 12 (85.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 15... +Query 1/1: Action query time = 5.713 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3530 +t=10: Selected seed 195 with value = 0.3530 +Query 1/1: Action query time = 3.968 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3442 +t=26: Selected seed 195 with value = 0.3442 +Query 1/1: Action query time = 4.200 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4300 +t=42: Selected seed 195 with value = 0.4300 +Query 1/1: Action query time = 3.193 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5164 +t=58: Selected seed 195 with value = 0.5164 +Query 1/1: Action query time = 3.626 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5277 +t=74: Selected seed 195 with value = 0.5277 +Query 1/1: Action query time = 3.986 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5896 +t=90: Selected seed 195 with value = 0.5896 +Query 1/1: Action query time = 5.819 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6865 +t=106: Selected seed 195 with value = 0.6865 +Query 1/1: Action query time = 5.132 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8044 +t=122: Selected seed 195 with value = 0.8044 +Query 1/1: Action query time = 4.372 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9410 +t=138: Selected seed 195 with value = 0.9410 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=15--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 15 +# successes: 13 (86.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 16... +Query 1/1: Action query time = 3.917 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3676 +t=10: Selected seed 195 with value = 0.3676 +Query 1/1: Action query time = 4.899 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3913 +t=26: Selected seed 195 with value = 0.3913 +Query 1/1: Action query time = 4.331 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4815 +t=42: Selected seed 195 with value = 0.4815 +Query 1/1: Action query time = 5.429 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5440 +t=58: Selected seed 195 with value = 0.5440 +Query 1/1: Action query time = 4.135 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5421 +t=74: Selected seed 195 with value = 0.5421 +Query 1/1: Action query time = 5.036 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4706 +t=90: Selected seed 195 with value = 0.4706 +Query 1/1: Action query time = 3.365 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4889 +t=106: Selected seed 195 with value = 0.4889 +Query 1/1: Action query time = 3.409 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5292 +t=122: Selected seed 195 with value = 0.5292 +Query 1/1: Action query time = 5.281 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6488 +t=138: Selected seed 195 with value = 0.6488 +Query 1/1: Action query time = 4.704 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7282 +t=154: Selected seed 195 with value = 0.7282 +Query 1/1: Action query time = 5.214 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6521 +t=170: Selected seed 195 with value = 0.6521 +Query 1/1: Action query time = 5.793 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5469 +t=186: Selected seed 195 with value = 0.5469 +Query 1/1: Action query time = 2.548 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4687 +t=202: Selected seed 195 with value = 0.4687 +Query 1/1: Action query time = 4.304 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4893 +t=218: Selected seed 195 with value = 0.4893 +Query 1/1: Action query time = 4.702 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4839 +t=234: Selected seed 195 with value = 0.4839 +Query 1/1: Action query time = 3.101 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5329 +t=250: Selected seed 195 with value = 0.5329 +Query 1/1: Action query time = 4.501 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4907 +t=266: Selected seed 195 with value = 0.4907 +Query 1/1: Action query time = 4.222 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5643 +t=282: Selected seed 195 with value = 0.5643 +Query 1/1: Action query time = 4.662 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5046 +t=298: Selected seed 195 with value = 0.5046 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=16--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 16 +# successes: 13 (81.2%) + +Task: put the wine bottle on top of the cabinet +Starting episode 17... +Query 1/1: Action query time = 4.275 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4332 +t=10: Selected seed 195 with value = 0.4332 +Query 1/1: Action query time = 4.239 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4889 +t=26: Selected seed 195 with value = 0.4889 +Query 1/1: Action query time = 4.799 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5486 +t=42: Selected seed 195 with value = 0.5486 +Query 1/1: Action query time = 3.635 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6425 +t=58: Selected seed 195 with value = 0.6425 +Query 1/1: Action query time = 4.054 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7916 +t=74: Selected seed 195 with value = 0.7916 +Query 1/1: Action query time = 5.431 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9150 +t=90: Selected seed 195 with value = 0.9150 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=17--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 17 +# successes: 14 (82.4%) + +Task: put the wine bottle on top of the cabinet +Starting episode 18... +Query 1/1: Action query time = 4.493 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3497 +t=10: Selected seed 195 with value = 0.3497 +Query 1/1: Action query time = 3.644 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4162 +t=26: Selected seed 195 with value = 0.4162 +Query 1/1: Action query time = 4.164 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5334 +t=42: Selected seed 195 with value = 0.5334 +Query 1/1: Action query time = 4.827 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6720 +t=58: Selected seed 195 with value = 0.6720 +Query 1/1: Action query time = 5.155 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8512 +t=74: Selected seed 195 with value = 0.8512 +Query 1/1: Action query time = 5.634 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=90: Selected seed 195 with value = 0.9782 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=18--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 18 +# successes: 15 (83.3%) + +Task: put the wine bottle on top of the cabinet +Starting episode 19... +Query 1/1: Action query time = 3.313 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3746 +t=10: Selected seed 195 with value = 0.3746 +Query 1/1: Action query time = 4.835 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4458 +t=26: Selected seed 195 with value = 0.4458 +Query 1/1: Action query time = 4.260 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5217 +t=42: Selected seed 195 with value = 0.5217 +Query 1/1: Action query time = 5.244 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6112 +t=58: Selected seed 195 with value = 0.6112 +Query 1/1: Action query time = 4.500 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7289 +t=74: Selected seed 195 with value = 0.7289 +Query 1/1: Action query time = 5.481 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8435 +t=90: Selected seed 195 with value = 0.8435 +Query 1/1: Action query time = 4.322 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=106: Selected seed 195 with value = 0.9977 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=19--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 19 +# successes: 16 (84.2%) + +Task: put the wine bottle on top of the cabinet +Starting episode 20... +Query 1/1: Action query time = 4.847 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4294 +t=10: Selected seed 195 with value = 0.4294 +Query 1/1: Action query time = 4.343 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4782 +t=26: Selected seed 195 with value = 0.4782 +Query 1/1: Action query time = 3.899 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6098 +t=42: Selected seed 195 with value = 0.6098 +Query 1/1: Action query time = 3.580 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7233 +t=58: Selected seed 195 with value = 0.7233 +Query 1/1: Action query time = 3.647 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8305 +t=74: Selected seed 195 with value = 0.8305 +Query 1/1: Action query time = 2.948 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9682 +t=90: Selected seed 195 with value = 0.9682 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=20--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 20 +# successes: 17 (85.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 21... +Query 1/1: Action query time = 3.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3675 +t=10: Selected seed 195 with value = 0.3675 +Query 1/1: Action query time = 4.165 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3934 +t=26: Selected seed 195 with value = 0.3934 +Query 1/1: Action query time = 2.781 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4617 +t=42: Selected seed 195 with value = 0.4617 +Query 1/1: Action query time = 3.099 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5445 +t=58: Selected seed 195 with value = 0.5445 +Query 1/1: Action query time = 4.034 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6772 +t=74: Selected seed 195 with value = 0.6772 +Query 1/1: Action query time = 2.882 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8043 +t=90: Selected seed 195 with value = 0.8043 +Query 1/1: Action query time = 2.741 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9380 +t=106: Selected seed 195 with value = 0.9380 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=21--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 21 +# successes: 18 (85.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 22... +Query 1/1: Action query time = 2.957 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3999 +t=10: Selected seed 195 with value = 0.3999 +Query 1/1: Action query time = 3.783 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4630 +t=26: Selected seed 195 with value = 0.4630 +Query 1/1: Action query time = 2.314 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5103 +t=42: Selected seed 195 with value = 0.5103 +Query 1/1: Action query time = 3.630 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6227 +t=58: Selected seed 195 with value = 0.6227 +Query 1/1: Action query time = 4.346 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6792 +t=74: Selected seed 195 with value = 0.6792 +Query 1/1: Action query time = 3.374 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7868 +t=90: Selected seed 195 with value = 0.7868 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=22--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 22 +# successes: 19 (86.4%) + +Task: put the wine bottle on top of the cabinet +Starting episode 23... +Query 1/1: Action query time = 2.495 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3167 +t=10: Selected seed 195 with value = 0.3167 +Query 1/1: Action query time = 3.111 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3691 +t=26: Selected seed 195 with value = 0.3691 +Query 1/1: Action query time = 4.132 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5487 +t=42: Selected seed 195 with value = 0.5487 +Query 1/1: Action query time = 3.191 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5789 +t=58: Selected seed 195 with value = 0.5789 +Query 1/1: Action query time = 3.459 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7800 +t=74: Selected seed 195 with value = 0.7800 +Query 1/1: Action query time = 3.255 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8879 +t=90: Selected seed 195 with value = 0.8879 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=23--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 23 +# successes: 20 (87.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 24... +Query 1/1: Action query time = 1.927 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4439 +t=10: Selected seed 195 with value = 0.4439 +Query 1/1: Action query time = 3.040 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5380 +t=26: Selected seed 195 with value = 0.5380 +Query 1/1: Action query time = 3.159 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6198 +t=42: Selected seed 195 with value = 0.6198 +Query 1/1: Action query time = 2.899 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7243 +t=58: Selected seed 195 with value = 0.7243 +Query 1/1: Action query time = 2.680 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8655 +t=74: Selected seed 195 with value = 0.8655 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=24--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 24 +# successes: 21 (87.5%) + +Task: put the wine bottle on top of the cabinet +Starting episode 25... +Query 1/1: Action query time = 3.702 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3600 +t=10: Selected seed 195 with value = 0.3600 +Query 1/1: Action query time = 3.751 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4099 +t=26: Selected seed 195 with value = 0.4099 +Query 1/1: Action query time = 2.292 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5732 +t=42: Selected seed 195 with value = 0.5732 +Query 1/1: Action query time = 2.497 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7216 +t=58: Selected seed 195 with value = 0.7216 +Query 1/1: Action query time = 3.054 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8580 +t=74: Selected seed 195 with value = 0.8580 +Query 1/1: Action query time = 2.486 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=90: Selected seed 195 with value = 0.9965 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t2_s0/2026_08_03-10_52_25--with_future_img--episode=25--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 25 +# successes: 22 (88.0%) +Current task success rate: 0.88 +Current total success rate: 0.88 +Final results: +Total episodes: 25 +Total successes: 22 +Overall success rate: 0.8800 (88.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-10_52_39--t7realcl800x50_t5_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-10_52_39--t7realcl800x50_t5_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..69dab1147543525ffe997f4e70d182869999a4df --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-10_52_39--t7realcl800x50_t5_s1.txt @@ -0,0 +1,1175 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_real_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7realcl800x50_t5_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 3.962 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2295 +t=10: Selected seed 195 with value = 0.2295 +Query 1/1: Action query time = 3.021 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2365 +t=26: Selected seed 195 with value = 0.2365 +Query 1/1: Action query time = 4.121 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2806 +t=42: Selected seed 195 with value = 0.2806 +Query 1/1: Action query time = 4.880 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4197 +t=58: Selected seed 195 with value = 0.4197 +Query 1/1: Action query time = 4.616 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2416 +t=74: Selected seed 195 with value = 0.2416 +Query 1/1: Action query time = 5.472 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6272 +t=90: Selected seed 195 with value = 0.6272 +Query 1/1: Action query time = 5.018 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6962 +t=106: Selected seed 195 with value = 0.6962 +Query 1/1: Action query time = 4.193 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8139 +t=122: Selected seed 195 with value = 0.8139 +Query 1/1: Action query time = 4.825 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9735 +t=138: Selected seed 195 with value = 0.9735 +Query 1/1: Action query time = 3.331 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.803 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9972 +t=170: Selected seed 195 with value = 0.9972 +Query 1/1: Action query time = 4.369 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9887 +t=186: Selected seed 195 with value = 0.9887 +Query 1/1: Action query time = 5.250 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9790 +t=202: Selected seed 195 with value = 0.9790 +Query 1/1: Action query time = 5.119 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9812 +t=218: Selected seed 195 with value = 0.9812 +Query 1/1: Action query time = 4.613 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9823 +t=234: Selected seed 195 with value = 0.9823 +Query 1/1: Action query time = 4.091 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9820 +t=250: Selected seed 195 with value = 0.9820 +Query 1/1: Action query time = 4.743 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9815 +t=266: Selected seed 195 with value = 0.9815 +Query 1/1: Action query time = 3.474 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9818 +t=282: Selected seed 195 with value = 0.9818 +Query 1/1: Action query time = 4.002 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9816 +t=298: Selected seed 195 with value = 0.9816 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=1--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 3.376 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2240 +t=10: Selected seed 195 with value = 0.2240 +Query 1/1: Action query time = 5.597 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2097 +t=26: Selected seed 195 with value = 0.2097 +Query 1/1: Action query time = 5.179 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2121 +t=42: Selected seed 195 with value = 0.2121 +Query 1/1: Action query time = 5.275 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2500 +t=58: Selected seed 195 with value = 0.2500 +Query 1/1: Action query time = 3.832 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3336 +t=74: Selected seed 195 with value = 0.3336 +Query 1/1: Action query time = 5.372 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6760 +t=90: Selected seed 195 with value = 0.6760 +Query 1/1: Action query time = 5.279 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8196 +t=106: Selected seed 195 with value = 0.8196 +Query 1/1: Action query time = 4.203 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9394 +t=122: Selected seed 195 with value = 0.9394 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 4.014 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3241 +t=10: Selected seed 195 with value = 0.3241 +Query 1/1: Action query time = 4.971 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3393 +t=26: Selected seed 195 with value = 0.3393 +Query 1/1: Action query time = 4.174 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4104 +t=42: Selected seed 195 with value = 0.4104 +Query 1/1: Action query time = 3.948 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4799 +t=58: Selected seed 195 with value = 0.4799 +Query 1/1: Action query time = 5.043 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5348 +t=74: Selected seed 195 with value = 0.5348 +Query 1/1: Action query time = 4.666 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3747 +t=90: Selected seed 195 with value = 0.3747 +Query 1/1: Action query time = 4.399 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6807 +t=106: Selected seed 195 with value = 0.6807 +Query 1/1: Action query time = 4.024 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8702 +t=122: Selected seed 195 with value = 0.8702 +Query 1/1: Action query time = 5.778 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 4.253 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3502 +t=10: Selected seed 195 with value = 0.3502 +Query 1/1: Action query time = 4.886 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3751 +t=26: Selected seed 195 with value = 0.3751 +Query 1/1: Action query time = 4.415 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4721 +t=42: Selected seed 195 with value = 0.4721 +Query 1/1: Action query time = 4.080 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5908 +t=58: Selected seed 195 with value = 0.5908 +Query 1/1: Action query time = 4.455 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5829 +t=74: Selected seed 195 with value = 0.5829 +Query 1/1: Action query time = 5.040 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6348 +t=90: Selected seed 195 with value = 0.6348 +Query 1/1: Action query time = 4.031 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6700 +t=106: Selected seed 195 with value = 0.6700 +Query 1/1: Action query time = 3.927 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7824 +t=122: Selected seed 195 with value = 0.7824 +Query 1/1: Action query time = 4.339 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9859 +t=138: Selected seed 195 with value = 0.9859 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=4--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 4 +# successes: 3 (75.0%) + +Task: push the plate to the front of the stove +Starting episode 5... +Query 1/1: Action query time = 4.775 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2855 +t=10: Selected seed 195 with value = 0.2855 +Query 1/1: Action query time = 5.261 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2620 +t=26: Selected seed 195 with value = 0.2620 +Query 1/1: Action query time = 3.540 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4269 +t=42: Selected seed 195 with value = 0.4269 +Query 1/1: Action query time = 5.172 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3350 +t=58: Selected seed 195 with value = 0.3350 +Query 1/1: Action query time = 4.877 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5826 +t=74: Selected seed 195 with value = 0.5826 +Query 1/1: Action query time = 4.467 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2892 +t=90: Selected seed 195 with value = 0.2892 +Query 1/1: Action query time = 3.555 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7259 +t=106: Selected seed 195 with value = 0.7259 +Query 1/1: Action query time = 4.575 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6129 +t=122: Selected seed 195 with value = 0.6129 +Query 1/1: Action query time = 5.028 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7263 +t=138: Selected seed 195 with value = 0.7263 +Query 1/1: Action query time = 5.051 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3851 +t=154: Selected seed 195 with value = 0.3851 +Query 1/1: Action query time = 4.140 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4690 +t=170: Selected seed 195 with value = 0.4690 +Query 1/1: Action query time = 4.477 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7014 +t=186: Selected seed 195 with value = 0.7014 +Query 1/1: Action query time = 3.618 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5955 +t=202: Selected seed 195 with value = 0.5955 +Query 1/1: Action query time = 4.603 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3706 +t=218: Selected seed 195 with value = 0.3706 +Query 1/1: Action query time = 5.045 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3465 +t=234: Selected seed 195 with value = 0.3465 +Query 1/1: Action query time = 5.992 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4326 +t=250: Selected seed 195 with value = 0.4326 +Query 1/1: Action query time = 5.197 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5976 +t=266: Selected seed 195 with value = 0.5976 +Query 1/1: Action query time = 3.411 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6813 +t=282: Selected seed 195 with value = 0.6813 +Query 1/1: Action query time = 4.991 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=5--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 5 +# successes: 3 (60.0%) + +Task: push the plate to the front of the stove +Starting episode 6... +Query 1/1: Action query time = 5.180 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3364 +t=10: Selected seed 195 with value = 0.3364 +Query 1/1: Action query time = 5.718 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3947 +t=26: Selected seed 195 with value = 0.3947 +Query 1/1: Action query time = 4.468 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4810 +t=42: Selected seed 195 with value = 0.4810 +Query 1/1: Action query time = 4.490 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5619 +t=58: Selected seed 195 with value = 0.5619 +Query 1/1: Action query time = 5.081 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6834 +t=74: Selected seed 195 with value = 0.6834 +Query 1/1: Action query time = 5.038 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7801 +t=90: Selected seed 195 with value = 0.7801 +Query 1/1: Action query time = 4.996 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9094 +t=106: Selected seed 195 with value = 0.9094 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=6--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 6 +# successes: 4 (66.7%) + +Task: push the plate to the front of the stove +Starting episode 7... +Query 1/1: Action query time = 4.190 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2433 +t=10: Selected seed 195 with value = 0.2433 +Query 1/1: Action query time = 4.176 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2513 +t=26: Selected seed 195 with value = 0.2513 +Query 1/1: Action query time = 4.024 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4284 +t=42: Selected seed 195 with value = 0.4284 +Query 1/1: Action query time = 3.557 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4759 +t=58: Selected seed 195 with value = 0.4759 +Query 1/1: Action query time = 3.779 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4670 +t=74: Selected seed 195 with value = 0.4670 +Query 1/1: Action query time = 4.158 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6525 +t=90: Selected seed 195 with value = 0.6525 +Query 1/1: Action query time = 4.159 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7801 +t=106: Selected seed 195 with value = 0.7801 +Query 1/1: Action query time = 5.313 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9289 +t=122: Selected seed 195 with value = 0.9289 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=7--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 7 +# successes: 5 (71.4%) + +Task: push the plate to the front of the stove +Starting episode 8... +Query 1/1: Action query time = 4.828 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3455 +t=10: Selected seed 195 with value = 0.3455 +Query 1/1: Action query time = 3.246 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3983 +t=26: Selected seed 195 with value = 0.3983 +Query 1/1: Action query time = 3.134 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4802 +t=42: Selected seed 195 with value = 0.4802 +Query 1/1: Action query time = 4.897 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5568 +t=58: Selected seed 195 with value = 0.5568 +Query 1/1: Action query time = 5.122 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6396 +t=74: Selected seed 195 with value = 0.6396 +Query 1/1: Action query time = 4.618 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7264 +t=90: Selected seed 195 with value = 0.7264 +Query 1/1: Action query time = 5.165 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6042 +t=106: Selected seed 195 with value = 0.6042 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=8--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 8 +# successes: 6 (75.0%) + +Task: push the plate to the front of the stove +Starting episode 9... +Query 1/1: Action query time = 4.322 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2115 +t=10: Selected seed 195 with value = 0.2115 +Query 1/1: Action query time = 4.752 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2008 +t=26: Selected seed 195 with value = 0.2008 +Query 1/1: Action query time = 4.616 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4606 +t=42: Selected seed 195 with value = 0.4606 +Query 1/1: Action query time = 4.170 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5306 +t=58: Selected seed 195 with value = 0.5306 +Query 1/1: Action query time = 5.194 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6315 +t=74: Selected seed 195 with value = 0.6315 +Query 1/1: Action query time = 4.358 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7292 +t=90: Selected seed 195 with value = 0.7292 +Query 1/1: Action query time = 5.269 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8666 +t=106: Selected seed 195 with value = 0.8666 +Query 1/1: Action query time = 4.530 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=122: Selected seed 195 with value = 0.9980 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=9--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 9 +# successes: 7 (77.8%) + +Task: push the plate to the front of the stove +Starting episode 10... +Query 1/1: Action query time = 3.847 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2832 +t=10: Selected seed 195 with value = 0.2832 +Query 1/1: Action query time = 4.380 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3386 +t=26: Selected seed 195 with value = 0.3386 +Query 1/1: Action query time = 5.381 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4160 +t=42: Selected seed 195 with value = 0.4160 +Query 1/1: Action query time = 5.242 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4844 +t=58: Selected seed 195 with value = 0.4844 +Query 1/1: Action query time = 3.777 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5857 +t=74: Selected seed 195 with value = 0.5857 +Query 1/1: Action query time = 3.934 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6606 +t=90: Selected seed 195 with value = 0.6606 +Query 1/1: Action query time = 4.387 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7721 +t=106: Selected seed 195 with value = 0.7721 +Query 1/1: Action query time = 4.258 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8973 +t=122: Selected seed 195 with value = 0.8973 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=10--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 10 +# successes: 8 (80.0%) + +Task: push the plate to the front of the stove +Starting episode 11... +Query 1/1: Action query time = 3.763 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2565 +t=10: Selected seed 195 with value = 0.2565 +Query 1/1: Action query time = 4.980 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2763 +t=26: Selected seed 195 with value = 0.2763 +Query 1/1: Action query time = 3.390 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3043 +t=42: Selected seed 195 with value = 0.3043 +Query 1/1: Action query time = 4.892 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3688 +t=58: Selected seed 195 with value = 0.3688 +Query 1/1: Action query time = 4.582 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3671 +t=74: Selected seed 195 with value = 0.3671 +Query 1/1: Action query time = 3.524 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5158 +t=90: Selected seed 195 with value = 0.5158 +Query 1/1: Action query time = 4.827 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5180 +t=106: Selected seed 195 with value = 0.5180 +Query 1/1: Action query time = 5.007 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8753 +t=122: Selected seed 195 with value = 0.8753 +Query 1/1: Action query time = 4.114 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9794 +t=138: Selected seed 195 with value = 0.9794 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=11--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 11 +# successes: 9 (81.8%) + +Task: push the plate to the front of the stove +Starting episode 12... +Query 1/1: Action query time = 5.331 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1898 +t=10: Selected seed 195 with value = 0.1898 +Query 1/1: Action query time = 4.688 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2271 +t=26: Selected seed 195 with value = 0.2271 +Query 1/1: Action query time = 4.405 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4462 +t=42: Selected seed 195 with value = 0.4462 +Query 1/1: Action query time = 3.384 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2944 +t=58: Selected seed 195 with value = 0.2944 +Query 1/1: Action query time = 3.843 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5580 +t=74: Selected seed 195 with value = 0.5580 +Query 1/1: Action query time = 5.291 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3011 +t=90: Selected seed 195 with value = 0.3011 +Query 1/1: Action query time = 5.336 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3233 +t=106: Selected seed 195 with value = 0.3233 +Query 1/1: Action query time = 4.876 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3741 +t=122: Selected seed 195 with value = 0.3741 +Query 1/1: Action query time = 4.156 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4391 +t=138: Selected seed 195 with value = 0.4391 +Query 1/1: Action query time = 4.530 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5318 +t=154: Selected seed 195 with value = 0.5318 +Query 1/1: Action query time = 5.554 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6039 +t=170: Selected seed 195 with value = 0.6039 +Query 1/1: Action query time = 5.856 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6956 +t=186: Selected seed 195 with value = 0.6956 +Query 1/1: Action query time = 4.197 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8143 +t=202: Selected seed 195 with value = 0.8143 +Query 1/1: Action query time = 3.824 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9629 +t=218: Selected seed 195 with value = 0.9629 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=12--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 12 +# successes: 10 (83.3%) + +Task: push the plate to the front of the stove +Starting episode 13... +Query 1/1: Action query time = 4.836 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2259 +t=10: Selected seed 195 with value = 0.2259 +Query 1/1: Action query time = 4.770 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2895 +t=26: Selected seed 195 with value = 0.2895 +Query 1/1: Action query time = 4.116 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3570 +t=42: Selected seed 195 with value = 0.3570 +Query 1/1: Action query time = 5.411 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4391 +t=58: Selected seed 195 with value = 0.4391 +Query 1/1: Action query time = 5.263 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4667 +t=74: Selected seed 195 with value = 0.4667 +Query 1/1: Action query time = 4.096 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5650 +t=90: Selected seed 195 with value = 0.5650 +Query 1/1: Action query time = 4.105 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6418 +t=106: Selected seed 195 with value = 0.6418 +Query 1/1: Action query time = 4.921 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7457 +t=122: Selected seed 195 with value = 0.7457 +Query 1/1: Action query time = 4.872 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5010 +t=138: Selected seed 195 with value = 0.5010 +Query 1/1: Action query time = 2.984 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6184 +t=154: Selected seed 195 with value = 0.6184 +Query 1/1: Action query time = 5.139 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6973 +t=170: Selected seed 195 with value = 0.6973 +Query 1/1: Action query time = 4.853 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7889 +t=186: Selected seed 195 with value = 0.7889 +Query 1/1: Action query time = 4.012 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9314 +t=202: Selected seed 195 with value = 0.9314 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=13--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 13 +# successes: 11 (84.6%) + +Task: push the plate to the front of the stove +Starting episode 14... +Query 1/1: Action query time = 3.326 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1900 +t=10: Selected seed 195 with value = 0.1900 +Query 1/1: Action query time = 3.823 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3222 +t=26: Selected seed 195 with value = 0.3222 +Query 1/1: Action query time = 3.701 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3527 +t=42: Selected seed 195 with value = 0.3527 +Query 1/1: Action query time = 2.287 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4450 +t=58: Selected seed 195 with value = 0.4450 +Query 1/1: Action query time = 2.627 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6113 +t=74: Selected seed 195 with value = 0.6113 +Query 1/1: Action query time = 2.950 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7189 +t=90: Selected seed 195 with value = 0.7189 +Query 1/1: Action query time = 2.903 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8381 +t=106: Selected seed 195 with value = 0.8381 +Query 1/1: Action query time = 3.117 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=122: Selected seed 195 with value = 0.9924 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=14--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 14 +# successes: 12 (85.7%) + +Task: push the plate to the front of the stove +Starting episode 15... +Query 1/1: Action query time = 3.384 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3003 +t=10: Selected seed 195 with value = 0.3003 +Query 1/1: Action query time = 3.240 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3728 +t=26: Selected seed 195 with value = 0.3728 +Query 1/1: Action query time = 3.644 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4944 +t=42: Selected seed 195 with value = 0.4944 +Query 1/1: Action query time = 3.954 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5868 +t=58: Selected seed 195 with value = 0.5868 +Query 1/1: Action query time = 2.733 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5661 +t=74: Selected seed 195 with value = 0.5661 +Query 1/1: Action query time = 2.709 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6532 +t=90: Selected seed 195 with value = 0.6532 +Query 1/1: Action query time = 4.095 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7624 +t=106: Selected seed 195 with value = 0.7624 +Query 1/1: Action query time = 3.362 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8848 +t=122: Selected seed 195 with value = 0.8848 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=15--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 15 +# successes: 13 (86.7%) + +Task: push the plate to the front of the stove +Starting episode 16... +Query 1/1: Action query time = 3.452 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1971 +t=10: Selected seed 195 with value = 0.1971 +Query 1/1: Action query time = 3.943 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3016 +t=26: Selected seed 195 with value = 0.3016 +Query 1/1: Action query time = 4.862 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3917 +t=42: Selected seed 195 with value = 0.3917 +Query 1/1: Action query time = 2.920 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4886 +t=58: Selected seed 195 with value = 0.4886 +Query 1/1: Action query time = 3.058 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5280 +t=74: Selected seed 195 with value = 0.5280 +Query 1/1: Action query time = 3.507 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6476 +t=90: Selected seed 195 with value = 0.6476 +Query 1/1: Action query time = 2.637 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7647 +t=106: Selected seed 195 with value = 0.7647 +Query 1/1: Action query time = 1.903 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8790 +t=122: Selected seed 195 with value = 0.8790 +Query 1/1: Action query time = 2.022 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=138: Selected seed 195 with value = 0.9983 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=16--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 16 +# successes: 14 (87.5%) + +Task: push the plate to the front of the stove +Starting episode 17... +Query 1/1: Action query time = 3.026 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1830 +t=10: Selected seed 195 with value = 0.1830 +Query 1/1: Action query time = 1.933 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1959 +t=26: Selected seed 195 with value = 0.1959 +Query 1/1: Action query time = 2.301 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4617 +t=42: Selected seed 195 with value = 0.4617 +Query 1/1: Action query time = 2.422 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5172 +t=58: Selected seed 195 with value = 0.5172 +Query 1/1: Action query time = 2.375 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6061 +t=74: Selected seed 195 with value = 0.6061 +Query 1/1: Action query time = 2.066 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7182 +t=90: Selected seed 195 with value = 0.7182 +Query 1/1: Action query time = 2.542 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8390 +t=106: Selected seed 195 with value = 0.8390 +Query 1/1: Action query time = 2.060 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9868 +t=122: Selected seed 195 with value = 0.9868 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=17--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 17 +# successes: 15 (88.2%) + +Task: push the plate to the front of the stove +Starting episode 18... +Query 1/1: Action query time = 3.504 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3261 +t=10: Selected seed 195 with value = 0.3261 +Query 1/1: Action query time = 2.148 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3535 +t=26: Selected seed 195 with value = 0.3535 +Query 1/1: Action query time = 1.546 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2688 +t=42: Selected seed 195 with value = 0.2688 +Query 1/1: Action query time = 1.750 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3178 +t=58: Selected seed 195 with value = 0.3178 +Query 1/1: Action query time = 2.285 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3730 +t=74: Selected seed 195 with value = 0.3730 +Query 1/1: Action query time = 2.927 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4463 +t=90: Selected seed 195 with value = 0.4463 +Query 1/1: Action query time = 2.376 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5265 +t=106: Selected seed 195 with value = 0.5265 +Query 1/1: Action query time = 2.764 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6321 +t=122: Selected seed 195 with value = 0.6321 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=18--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 18 +# successes: 16 (88.9%) + +Task: push the plate to the front of the stove +Starting episode 19... +Query 1/1: Action query time = 1.436 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2216 +t=10: Selected seed 195 with value = 0.2216 +Query 1/1: Action query time = 1.015 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1820 +t=26: Selected seed 195 with value = 0.1820 +Query 1/1: Action query time = 1.606 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4133 +t=42: Selected seed 195 with value = 0.4133 +Query 1/1: Action query time = 1.812 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5024 +t=58: Selected seed 195 with value = 0.5024 +Query 1/1: Action query time = 2.330 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5389 +t=74: Selected seed 195 with value = 0.5389 +Query 1/1: Action query time = 2.507 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5984 +t=90: Selected seed 195 with value = 0.5984 +Query 1/1: Action query time = 2.420 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6894 +t=106: Selected seed 195 with value = 0.6894 +Query 1/1: Action query time = 1.646 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8407 +t=122: Selected seed 195 with value = 0.8407 +Query 1/1: Action query time = 1.420 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9756 +t=138: Selected seed 195 with value = 0.9756 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=19--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 19 +# successes: 17 (89.5%) + +Task: push the plate to the front of the stove +Starting episode 20... +Query 1/1: Action query time = 2.302 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3418 +t=10: Selected seed 195 with value = 0.3418 +Query 1/1: Action query time = 2.636 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3455 +t=26: Selected seed 195 with value = 0.3455 +Query 1/1: Action query time = 2.045 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4360 +t=42: Selected seed 195 with value = 0.4360 +Query 1/1: Action query time = 1.599 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4828 +t=58: Selected seed 195 with value = 0.4828 +Query 1/1: Action query time = 1.934 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5731 +t=74: Selected seed 195 with value = 0.5731 +Query 1/1: Action query time = 1.310 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6654 +t=90: Selected seed 195 with value = 0.6654 +Query 1/1: Action query time = 1.498 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7769 +t=106: Selected seed 195 with value = 0.7769 +Query 1/1: Action query time = 1.290 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9123 +t=122: Selected seed 195 with value = 0.9123 +Query 1/1: Action query time = 0.978 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=20--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 20 +# successes: 18 (90.0%) + +Task: push the plate to the front of the stove +Starting episode 21... +Query 1/1: Action query time = 1.680 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3275 +t=10: Selected seed 195 with value = 0.3275 +Query 1/1: Action query time = 1.817 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3309 +t=26: Selected seed 195 with value = 0.3309 +Query 1/1: Action query time = 1.730 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3742 +t=42: Selected seed 195 with value = 0.3742 +Query 1/1: Action query time = 1.835 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4714 +t=58: Selected seed 195 with value = 0.4714 +Query 1/1: Action query time = 1.816 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5012 +t=74: Selected seed 195 with value = 0.5012 +Query 1/1: Action query time = 1.949 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6575 +t=90: Selected seed 195 with value = 0.6575 +Query 1/1: Action query time = 1.957 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7287 +t=106: Selected seed 195 with value = 0.7287 +Query 1/1: Action query time = 1.948 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8821 +t=122: Selected seed 195 with value = 0.8821 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=21--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 21 +# successes: 19 (90.5%) + +Task: push the plate to the front of the stove +Starting episode 22... +Query 1/1: Action query time = 1.074 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1586 +t=10: Selected seed 195 with value = 0.1586 +Query 1/1: Action query time = 1.040 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2165 +t=26: Selected seed 195 with value = 0.2165 +Query 1/1: Action query time = 0.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2722 +t=42: Selected seed 195 with value = 0.2722 +Query 1/1: Action query time = 1.341 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3176 +t=58: Selected seed 195 with value = 0.3176 +Query 1/1: Action query time = 1.394 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3761 +t=74: Selected seed 195 with value = 0.3761 +Query 1/1: Action query time = 1.585 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7127 +t=90: Selected seed 195 with value = 0.7127 +Query 1/1: Action query time = 1.491 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5753 +t=106: Selected seed 195 with value = 0.5753 +Query 1/1: Action query time = 1.329 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6360 +t=122: Selected seed 195 with value = 0.6360 +Query 1/1: Action query time = 1.137 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7564 +t=138: Selected seed 195 with value = 0.7564 +Query 1/1: Action query time = 1.390 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8885 +t=154: Selected seed 195 with value = 0.8885 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=22--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 22 +# successes: 20 (90.9%) + +Task: push the plate to the front of the stove +Starting episode 23... +Query 1/1: Action query time = 0.981 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2405 +t=10: Selected seed 195 with value = 0.2405 +Query 1/1: Action query time = 1.270 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3657 +t=26: Selected seed 195 with value = 0.3657 +Query 1/1: Action query time = 0.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4373 +t=42: Selected seed 195 with value = 0.4373 +Query 1/1: Action query time = 1.446 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5035 +t=58: Selected seed 195 with value = 0.5035 +Query 1/1: Action query time = 1.304 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5919 +t=74: Selected seed 195 with value = 0.5919 +Query 1/1: Action query time = 1.546 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5643 +t=90: Selected seed 195 with value = 0.5643 +Query 1/1: Action query time = 1.254 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7472 +t=106: Selected seed 195 with value = 0.7472 +Query 1/1: Action query time = 1.571 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9941 +t=122: Selected seed 195 with value = 0.9941 +Query 1/1: Action query time = 1.481 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.359 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7739 +t=154: Selected seed 195 with value = 0.7739 +Query 1/1: Action query time = 1.572 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8741 +t=170: Selected seed 195 with value = 0.8741 +Query 1/1: Action query time = 1.373 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=186: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 1.531 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9940 +t=202: Selected seed 195 with value = 0.9940 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=23--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 23 +# successes: 21 (91.3%) + +Task: push the plate to the front of the stove +Starting episode 24... +Query 1/1: Action query time = 1.319 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2978 +t=10: Selected seed 195 with value = 0.2978 +Query 1/1: Action query time = 1.221 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3143 +t=26: Selected seed 195 with value = 0.3143 +Query 1/1: Action query time = 1.262 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3946 +t=42: Selected seed 195 with value = 0.3946 +Query 1/1: Action query time = 1.798 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5365 +t=58: Selected seed 195 with value = 0.5365 +Query 1/1: Action query time = 1.427 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4953 +t=74: Selected seed 195 with value = 0.4953 +Query 1/1: Action query time = 1.511 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5656 +t=90: Selected seed 195 with value = 0.5656 +Query 1/1: Action query time = 1.238 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6286 +t=106: Selected seed 195 with value = 0.6286 +Query 1/1: Action query time = 1.345 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7366 +t=122: Selected seed 195 with value = 0.7366 +Query 1/1: Action query time = 0.978 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8671 +t=138: Selected seed 195 with value = 0.8671 +Query 1/1: Action query time = 0.977 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9811 +t=154: Selected seed 195 with value = 0.9811 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=24--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 24 +# successes: 22 (91.7%) + +Task: push the plate to the front of the stove +Starting episode 25... +Query 1/1: Action query time = 1.714 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1707 +t=10: Selected seed 195 with value = 0.1707 +Query 1/1: Action query time = 1.752 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3280 +t=26: Selected seed 195 with value = 0.3280 +Query 1/1: Action query time = 1.346 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3787 +t=42: Selected seed 195 with value = 0.3787 +Query 1/1: Action query time = 1.697 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4678 +t=58: Selected seed 195 with value = 0.4678 +Query 1/1: Action query time = 1.341 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4605 +t=74: Selected seed 195 with value = 0.4605 +Query 1/1: Action query time = 1.438 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5640 +t=90: Selected seed 195 with value = 0.5640 +Query 1/1: Action query time = 1.452 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6539 +t=106: Selected seed 195 with value = 0.6539 +Query 1/1: Action query time = 1.105 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7622 +t=122: Selected seed 195 with value = 0.7622 +Query 1/1: Action query time = 0.977 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9153 +t=138: Selected seed 195 with value = 0.9153 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t5_s1/2026_08_03-10_52_39--with_future_img--episode=25--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 25 +# successes: 23 (92.0%) +Current task success rate: 0.92 +Current total success rate: 0.92 +Final results: +Total episodes: 25 +Total successes: 23 +Overall success rate: 0.9200 (92.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-10_52_45--t7realcl800x50_t7_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-10_52_45--t7realcl800x50_t7_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..236cc6e1924969e6b33be035b4e85dcc0a3fc626 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-10_52_45--t7realcl800x50_t7_s0.txt @@ -0,0 +1,711 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_real_cl_from_realcl800_2gpu/checkpoints/iter_000000800/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='t7realcl800x50_t7_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 5.175 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4767 +t=10: Selected seed 195 with value = 0.4767 +Query 1/1: Action query time = 4.497 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5370 +t=26: Selected seed 195 with value = 0.5370 +Query 1/1: Action query time = 5.083 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6479 +t=42: Selected seed 195 with value = 0.6479 +Query 1/1: Action query time = 5.157 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7632 +t=58: Selected seed 195 with value = 0.7632 +Query 1/1: Action query time = 5.569 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8761 +t=74: Selected seed 195 with value = 0.8761 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 5.337 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4808 +t=10: Selected seed 195 with value = 0.4808 +Query 1/1: Action query time = 3.846 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5438 +t=26: Selected seed 195 with value = 0.5438 +Query 1/1: Action query time = 4.170 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6540 +t=42: Selected seed 195 with value = 0.6540 +Query 1/1: Action query time = 5.277 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7689 +t=58: Selected seed 195 with value = 0.7689 +Query 1/1: Action query time = 5.088 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9161 +t=74: Selected seed 195 with value = 0.9161 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 5.102 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4865 +t=10: Selected seed 195 with value = 0.4865 +Query 1/1: Action query time = 4.058 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5686 +t=26: Selected seed 195 with value = 0.5686 +Query 1/1: Action query time = 3.626 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6588 +t=42: Selected seed 195 with value = 0.6588 +Query 1/1: Action query time = 4.858 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7302 +t=58: Selected seed 195 with value = 0.7302 +Query 1/1: Action query time = 3.840 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8675 +t=74: Selected seed 195 with value = 0.8675 +Query 1/1: Action query time = 3.492 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9926 +t=90: Selected seed 195 with value = 0.9926 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 3.393 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4732 +t=10: Selected seed 195 with value = 0.4732 +Query 1/1: Action query time = 4.911 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5575 +t=26: Selected seed 195 with value = 0.5575 +Query 1/1: Action query time = 4.043 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6292 +t=42: Selected seed 195 with value = 0.6292 +Query 1/1: Action query time = 6.354 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7238 +t=58: Selected seed 195 with value = 0.7238 +Query 1/1: Action query time = 5.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8485 +t=74: Selected seed 195 with value = 0.8485 +Query 1/1: Action query time = 5.036 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9915 +t=90: Selected seed 195 with value = 0.9915 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 4.203 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4858 +t=10: Selected seed 195 with value = 0.4858 +Query 1/1: Action query time = 4.158 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5857 +t=26: Selected seed 195 with value = 0.5857 +Query 1/1: Action query time = 4.367 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6686 +t=42: Selected seed 195 with value = 0.6686 +Query 1/1: Action query time = 4.586 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7792 +t=58: Selected seed 195 with value = 0.7792 +Query 1/1: Action query time = 4.589 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9271 +t=74: Selected seed 195 with value = 0.9271 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 4.513 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4761 +t=10: Selected seed 195 with value = 0.4761 +Query 1/1: Action query time = 4.163 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5402 +t=26: Selected seed 195 with value = 0.5402 +Query 1/1: Action query time = 4.028 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6334 +t=42: Selected seed 195 with value = 0.6334 +Query 1/1: Action query time = 3.312 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7520 +t=58: Selected seed 195 with value = 0.7520 +Query 1/1: Action query time = 4.763 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8990 +t=74: Selected seed 195 with value = 0.8990 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: turn on the stove +Starting episode 7... +Query 1/1: Action query time = 4.960 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4619 +t=10: Selected seed 195 with value = 0.4619 +Query 1/1: Action query time = 4.808 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5592 +t=26: Selected seed 195 with value = 0.5592 +Query 1/1: Action query time = 4.935 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6269 +t=42: Selected seed 195 with value = 0.6269 +Query 1/1: Action query time = 4.710 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7309 +t=58: Selected seed 195 with value = 0.7309 +Query 1/1: Action query time = 4.049 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8853 +t=74: Selected seed 195 with value = 0.8853 +Query 1/1: Action query time = 4.815 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=90: Selected seed 195 with value = 0.9986 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=7--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: turn on the stove +Starting episode 8... +Query 1/1: Action query time = 4.181 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4945 +t=10: Selected seed 195 with value = 0.4945 +Query 1/1: Action query time = 3.315 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5556 +t=26: Selected seed 195 with value = 0.5556 +Query 1/1: Action query time = 4.611 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6398 +t=42: Selected seed 195 with value = 0.6398 +Query 1/1: Action query time = 4.082 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7775 +t=58: Selected seed 195 with value = 0.7775 +Query 1/1: Action query time = 3.733 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9350 +t=74: Selected seed 195 with value = 0.9350 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=8--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: turn on the stove +Starting episode 9... +Query 1/1: Action query time = 5.640 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4517 +t=10: Selected seed 195 with value = 0.4517 +Query 1/1: Action query time = 4.545 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5381 +t=26: Selected seed 195 with value = 0.5381 +Query 1/1: Action query time = 3.627 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6214 +t=42: Selected seed 195 with value = 0.6214 +Query 1/1: Action query time = 5.081 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7231 +t=58: Selected seed 195 with value = 0.7231 +Query 1/1: Action query time = 3.500 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9030 +t=74: Selected seed 195 with value = 0.9030 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=9--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: turn on the stove +Starting episode 10... +Query 1/1: Action query time = 5.042 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4786 +t=10: Selected seed 195 with value = 0.4786 +Query 1/1: Action query time = 5.521 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5610 +t=26: Selected seed 195 with value = 0.5610 +Query 1/1: Action query time = 5.744 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6767 +t=42: Selected seed 195 with value = 0.6767 +Query 1/1: Action query time = 3.046 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7238 +t=58: Selected seed 195 with value = 0.7238 +Query 1/1: Action query time = 4.903 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9100 +t=74: Selected seed 195 with value = 0.9100 +Query 1/1: Action query time = 3.799 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=90: Selected seed 195 with value = 0.9980 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=10--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: turn on the stove +Starting episode 11... +Query 1/1: Action query time = 5.450 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5409 +t=10: Selected seed 195 with value = 0.5409 +Query 1/1: Action query time = 4.113 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6031 +t=26: Selected seed 195 with value = 0.6031 +Query 1/1: Action query time = 3.824 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6724 +t=42: Selected seed 195 with value = 0.6724 +Query 1/1: Action query time = 4.967 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7969 +t=58: Selected seed 195 with value = 0.7969 +Query 1/1: Action query time = 3.563 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9728 +t=74: Selected seed 195 with value = 0.9728 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=11--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: turn on the stove +Starting episode 12... +Query 1/1: Action query time = 4.130 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5139 +t=10: Selected seed 195 with value = 0.5139 +Query 1/1: Action query time = 3.410 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6652 +t=26: Selected seed 195 with value = 0.6652 +Query 1/1: Action query time = 5.827 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7630 +t=42: Selected seed 195 with value = 0.7630 +Query 1/1: Action query time = 4.005 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8962 +t=58: Selected seed 195 with value = 0.8962 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=12--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: turn on the stove +Starting episode 13... +Query 1/1: Action query time = 4.850 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4505 +t=10: Selected seed 195 with value = 0.4505 +Query 1/1: Action query time = 5.188 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5265 +t=26: Selected seed 195 with value = 0.5265 +Query 1/1: Action query time = 5.122 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6315 +t=42: Selected seed 195 with value = 0.6315 +Query 1/1: Action query time = 4.076 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7014 +t=58: Selected seed 195 with value = 0.7014 +Query 1/1: Action query time = 3.894 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8744 +t=74: Selected seed 195 with value = 0.8744 +Query 1/1: Action query time = 4.304 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9942 +t=90: Selected seed 195 with value = 0.9942 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=13--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) + +Task: turn on the stove +Starting episode 14... +Query 1/1: Action query time = 4.374 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5193 +t=10: Selected seed 195 with value = 0.5193 +Query 1/1: Action query time = 3.836 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6198 +t=26: Selected seed 195 with value = 0.6198 +Query 1/1: Action query time = 4.252 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7078 +t=42: Selected seed 195 with value = 0.7078 +Query 1/1: Action query time = 4.844 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8480 +t=58: Selected seed 195 with value = 0.8480 +Query 1/1: Action query time = 3.996 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9296 +t=74: Selected seed 195 with value = 0.9296 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=14--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 14 +# successes: 14 (100.0%) + +Task: turn on the stove +Starting episode 15... +Query 1/1: Action query time = 5.655 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5133 +t=10: Selected seed 195 with value = 0.5133 +Query 1/1: Action query time = 3.971 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6120 +t=26: Selected seed 195 with value = 0.6120 +Query 1/1: Action query time = 4.060 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6693 +t=42: Selected seed 195 with value = 0.6693 +Query 1/1: Action query time = 4.582 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8058 +t=58: Selected seed 195 with value = 0.8058 +Query 1/1: Action query time = 5.364 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9616 +t=74: Selected seed 195 with value = 0.9616 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=15--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 15 +# successes: 15 (100.0%) + +Task: turn on the stove +Starting episode 16... +Query 1/1: Action query time = 4.931 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4724 +t=10: Selected seed 195 with value = 0.4724 +Query 1/1: Action query time = 5.170 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5561 +t=26: Selected seed 195 with value = 0.5561 +Query 1/1: Action query time = 4.630 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6173 +t=42: Selected seed 195 with value = 0.6173 +Query 1/1: Action query time = 5.473 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7648 +t=58: Selected seed 195 with value = 0.7648 +Query 1/1: Action query time = 4.194 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9343 +t=74: Selected seed 195 with value = 0.9343 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=16--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 16 +# successes: 16 (100.0%) + +Task: turn on the stove +Starting episode 17... +Query 1/1: Action query time = 4.442 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4829 +t=10: Selected seed 195 with value = 0.4829 +Query 1/1: Action query time = 5.023 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5570 +t=26: Selected seed 195 with value = 0.5570 +Query 1/1: Action query time = 4.994 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6557 +t=42: Selected seed 195 with value = 0.6557 +Query 1/1: Action query time = 4.561 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7418 +t=58: Selected seed 195 with value = 0.7418 +Query 1/1: Action query time = 3.902 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9102 +t=74: Selected seed 195 with value = 0.9102 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=17--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 17 +# successes: 17 (100.0%) + +Task: turn on the stove +Starting episode 18... +Query 1/1: Action query time = 4.399 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5346 +t=10: Selected seed 195 with value = 0.5346 +Query 1/1: Action query time = 3.905 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6355 +t=26: Selected seed 195 with value = 0.6355 +Query 1/1: Action query time = 3.645 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7108 +t=42: Selected seed 195 with value = 0.7108 +Query 1/1: Action query time = 5.174 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7958 +t=58: Selected seed 195 with value = 0.7958 +Query 1/1: Action query time = 4.521 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9379 +t=74: Selected seed 195 with value = 0.9379 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=18--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 18 +# successes: 18 (100.0%) + +Task: turn on the stove +Starting episode 19... +Query 1/1: Action query time = 5.161 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4808 +t=10: Selected seed 195 with value = 0.4808 +Query 1/1: Action query time = 3.423 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5530 +t=26: Selected seed 195 with value = 0.5530 +Query 1/1: Action query time = 4.652 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6687 +t=42: Selected seed 195 with value = 0.6687 +Query 1/1: Action query time = 5.885 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7809 +t=58: Selected seed 195 with value = 0.7809 +Query 1/1: Action query time = 4.646 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9375 +t=74: Selected seed 195 with value = 0.9375 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=19--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 19 +# successes: 19 (100.0%) + +Task: turn on the stove +Starting episode 20... +Query 1/1: Action query time = 4.910 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5200 +t=10: Selected seed 195 with value = 0.5200 +Query 1/1: Action query time = 4.749 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5732 +t=26: Selected seed 195 with value = 0.5732 +Query 1/1: Action query time = 3.722 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6648 +t=42: Selected seed 195 with value = 0.6648 +Query 1/1: Action query time = 4.667 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7109 +t=58: Selected seed 195 with value = 0.7109 +Query 1/1: Action query time = 4.914 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8662 +t=74: Selected seed 195 with value = 0.8662 +Query 1/1: Action query time = 4.703 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9785 +t=90: Selected seed 195 with value = 0.9785 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=20--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 20 +# successes: 20 (100.0%) + +Task: turn on the stove +Starting episode 21... +Query 1/1: Action query time = 4.952 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4674 +t=10: Selected seed 195 with value = 0.4674 +Query 1/1: Action query time = 5.212 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5585 +t=26: Selected seed 195 with value = 0.5585 +Query 1/1: Action query time = 4.950 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6585 +t=42: Selected seed 195 with value = 0.6585 +Query 1/1: Action query time = 4.729 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7714 +t=58: Selected seed 195 with value = 0.7714 +Query 1/1: Action query time = 5.718 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9052 +t=74: Selected seed 195 with value = 0.9052 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=21--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 21 +# successes: 21 (100.0%) + +Task: turn on the stove +Starting episode 22... +Query 1/1: Action query time = 5.287 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5244 +t=10: Selected seed 195 with value = 0.5244 +Query 1/1: Action query time = 4.632 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6025 +t=26: Selected seed 195 with value = 0.6025 +Query 1/1: Action query time = 5.139 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6654 +t=42: Selected seed 195 with value = 0.6654 +Query 1/1: Action query time = 5.706 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7617 +t=58: Selected seed 195 with value = 0.7617 +Query 1/1: Action query time = 5.955 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8919 +t=74: Selected seed 195 with value = 0.8919 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=22--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 22 +# successes: 22 (100.0%) + +Task: turn on the stove +Starting episode 23... +Query 1/1: Action query time = 3.869 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5034 +t=10: Selected seed 195 with value = 0.5034 +Query 1/1: Action query time = 3.463 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5434 +t=26: Selected seed 195 with value = 0.5434 +Query 1/1: Action query time = 4.100 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6652 +t=42: Selected seed 195 with value = 0.6652 +Query 1/1: Action query time = 4.521 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7629 +t=58: Selected seed 195 with value = 0.7629 +Query 1/1: Action query time = 4.998 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9531 +t=74: Selected seed 195 with value = 0.9531 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=23--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 23 +# successes: 23 (100.0%) + +Task: turn on the stove +Starting episode 24... +Query 1/1: Action query time = 5.622 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4781 +t=10: Selected seed 195 with value = 0.4781 +Query 1/1: Action query time = 4.800 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5382 +t=26: Selected seed 195 with value = 0.5382 +Query 1/1: Action query time = 4.253 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6346 +t=42: Selected seed 195 with value = 0.6346 +Query 1/1: Action query time = 3.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7436 +t=58: Selected seed 195 with value = 0.7436 +Query 1/1: Action query time = 4.927 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8882 +t=74: Selected seed 195 with value = 0.8882 +Query 1/1: Action query time = 5.017 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=90: Selected seed 195 with value = 0.9988 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=24--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 24 +# successes: 24 (100.0%) + +Task: turn on the stove +Starting episode 25... +Query 1/1: Action query time = 4.884 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5131 +t=10: Selected seed 195 with value = 0.5131 +Query 1/1: Action query time = 4.831 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5930 +t=26: Selected seed 195 with value = 0.5930 +Query 1/1: Action query time = 4.044 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7014 +t=42: Selected seed 195 with value = 0.7014 +Query 1/1: Action query time = 2.367 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8635 +t=58: Selected seed 195 with value = 0.8635 +Query 1/1: Action query time = 3.551 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9929 +t=74: Selected seed 195 with value = 0.9929 +Saved rollout MP4 at path ./rollouts/t7realcl800x50_t7_s0/2026_08_03-10_52_45--with_future_img--episode=25--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 25 +# successes: 25 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 25 +Total successes: 25 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-12_46_21--forgetcl_i175_t3_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-12_46_21--forgetcl_i175_t3_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..49ec45352e0c94d9fbbd99670aaff91c361589d3 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-12_46_21--forgetcl_i175_t3_s1.txt @@ -0,0 +1,457 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='forgetcl_i175_t3_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,9,17,25,33,41,49', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 2.268 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2339 +t=10: Selected seed 195 with value = 0.2339 +Query 1/1: Action query time = 2.355 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2612 +t=26: Selected seed 195 with value = 0.2612 +Query 1/1: Action query time = 2.739 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3200 +t=42: Selected seed 195 with value = 0.3200 +Query 1/1: Action query time = 2.220 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3752 +t=58: Selected seed 195 with value = 0.3752 +Query 1/1: Action query time = 2.237 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4514 +t=74: Selected seed 195 with value = 0.4514 +Query 1/1: Action query time = 2.514 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5138 +t=90: Selected seed 195 with value = 0.5138 +Query 1/1: Action query time = 2.262 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6131 +t=106: Selected seed 195 with value = 0.6131 +Query 1/1: Action query time = 2.575 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7345 +t=122: Selected seed 195 with value = 0.7345 +Query 1/1: Action query time = 2.351 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8768 +t=138: Selected seed 195 with value = 0.8768 +Query 1/1: Action query time = 2.415 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.283 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s1/2026_08_03-12_46_21--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 0.986 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2007 +t=10: Selected seed 195 with value = 0.2007 +Query 1/1: Action query time = 1.016 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2897 +t=26: Selected seed 195 with value = 0.2897 +Query 1/1: Action query time = 1.089 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3117 +t=42: Selected seed 195 with value = 0.3117 +Query 1/1: Action query time = 2.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3615 +t=58: Selected seed 195 with value = 0.3615 +Query 1/1: Action query time = 2.747 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4737 +t=74: Selected seed 195 with value = 0.4737 +Query 1/1: Action query time = 3.077 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5118 +t=90: Selected seed 195 with value = 0.5118 +Query 1/1: Action query time = 2.826 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6000 +t=106: Selected seed 195 with value = 0.6000 +Query 1/1: Action query time = 2.639 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6873 +t=122: Selected seed 195 with value = 0.6873 +Query 1/1: Action query time = 2.370 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8157 +t=138: Selected seed 195 with value = 0.8157 +Query 1/1: Action query time = 2.463 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9467 +t=154: Selected seed 195 with value = 0.9467 +Query 1/1: Action query time = 2.418 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9845 +t=170: Selected seed 195 with value = 0.9845 +Query 1/1: Action query time = 2.455 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9637 +t=186: Selected seed 195 with value = 0.9637 +Query 1/1: Action query time = 2.326 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9662 +t=202: Selected seed 195 with value = 0.9662 +Query 1/1: Action query time = 2.102 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9710 +t=218: Selected seed 195 with value = 0.9710 +Query 1/1: Action query time = 1.010 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9646 +t=234: Selected seed 195 with value = 0.9646 +Query 1/1: Action query time = 1.229 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9701 +t=250: Selected seed 195 with value = 0.9701 +Query 1/1: Action query time = 1.750 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9619 +t=266: Selected seed 195 with value = 0.9619 +Query 1/1: Action query time = 2.481 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9653 +t=282: Selected seed 195 with value = 0.9653 +Query 1/1: Action query time = 2.091 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9594 +t=298: Selected seed 195 with value = 0.9594 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s1/2026_08_03-12_46_21--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 2.760 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2212 +t=10: Selected seed 195 with value = 0.2212 +Query 1/1: Action query time = 2.676 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2312 +t=26: Selected seed 195 with value = 0.2312 +Query 1/1: Action query time = 3.240 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3239 +t=42: Selected seed 195 with value = 0.3239 +Query 1/1: Action query time = 2.520 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3773 +t=58: Selected seed 195 with value = 0.3773 +Query 1/1: Action query time = 1.762 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4390 +t=74: Selected seed 195 with value = 0.4390 +Query 1/1: Action query time = 2.208 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5146 +t=90: Selected seed 195 with value = 0.5146 +Query 1/1: Action query time = 2.588 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5992 +t=106: Selected seed 195 with value = 0.5992 +Query 1/1: Action query time = 1.684 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6339 +t=122: Selected seed 195 with value = 0.6339 +Query 1/1: Action query time = 1.298 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6596 +t=138: Selected seed 195 with value = 0.6596 +Query 1/1: Action query time = 1.504 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7088 +t=154: Selected seed 195 with value = 0.7088 +Query 1/1: Action query time = 2.164 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8207 +t=170: Selected seed 195 with value = 0.8207 +Query 1/1: Action query time = 2.951 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9305 +t=186: Selected seed 195 with value = 0.9305 +Query 1/1: Action query time = 2.751 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.550 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s1/2026_08_03-12_46_21--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 2.591 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2213 +t=10: Selected seed 195 with value = 0.2213 +Query 1/1: Action query time = 2.759 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2713 +t=26: Selected seed 195 with value = 0.2713 +Query 1/1: Action query time = 2.282 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3051 +t=42: Selected seed 195 with value = 0.3051 +Query 1/1: Action query time = 2.168 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3317 +t=58: Selected seed 195 with value = 0.3317 +Query 1/1: Action query time = 1.787 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3949 +t=74: Selected seed 195 with value = 0.3949 +Query 1/1: Action query time = 1.528 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4568 +t=90: Selected seed 195 with value = 0.4568 +Query 1/1: Action query time = 2.410 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5579 +t=106: Selected seed 195 with value = 0.5579 +Query 1/1: Action query time = 1.872 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6185 +t=122: Selected seed 195 with value = 0.6185 +Query 1/1: Action query time = 2.152 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8074 +t=138: Selected seed 195 with value = 0.8074 +Query 1/1: Action query time = 2.158 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8763 +t=154: Selected seed 195 with value = 0.8763 +Query 1/1: Action query time = 2.740 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9944 +t=170: Selected seed 195 with value = 0.9944 +Query 1/1: Action query time = 2.114 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s1/2026_08_03-12_46_21--with_future_img--episode=4--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 4 +# successes: 3 (75.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 5... +Query 1/1: Action query time = 1.667 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2266 +t=10: Selected seed 195 with value = 0.2266 +Query 1/1: Action query time = 1.701 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2640 +t=26: Selected seed 195 with value = 0.2640 +Query 1/1: Action query time = 2.464 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2988 +t=42: Selected seed 195 with value = 0.2988 +Query 1/1: Action query time = 2.569 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3357 +t=58: Selected seed 195 with value = 0.3357 +Query 1/1: Action query time = 2.572 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3985 +t=74: Selected seed 195 with value = 0.3985 +Query 1/1: Action query time = 1.943 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4671 +t=90: Selected seed 195 with value = 0.4671 +Query 1/1: Action query time = 1.678 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5725 +t=106: Selected seed 195 with value = 0.5725 +Query 1/1: Action query time = 2.131 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6340 +t=122: Selected seed 195 with value = 0.6340 +Query 1/1: Action query time = 2.243 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7303 +t=138: Selected seed 195 with value = 0.7303 +Query 1/1: Action query time = 2.216 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8782 +t=154: Selected seed 195 with value = 0.8782 +Query 1/1: Action query time = 2.310 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.711 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s1/2026_08_03-12_46_21--with_future_img--episode=5--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 5 +# successes: 4 (80.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 6... +Query 1/1: Action query time = 2.018 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2168 +t=10: Selected seed 195 with value = 0.2168 +Query 1/1: Action query time = 1.940 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2505 +t=26: Selected seed 195 with value = 0.2505 +Query 1/1: Action query time = 2.403 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3198 +t=42: Selected seed 195 with value = 0.3198 +Query 1/1: Action query time = 2.498 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3714 +t=58: Selected seed 195 with value = 0.3714 +Query 1/1: Action query time = 2.643 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4380 +t=74: Selected seed 195 with value = 0.4380 +Query 1/1: Action query time = 1.934 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5263 +t=90: Selected seed 195 with value = 0.5263 +Query 1/1: Action query time = 1.627 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6134 +t=106: Selected seed 195 with value = 0.6134 +Query 1/1: Action query time = 1.742 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7662 +t=122: Selected seed 195 with value = 0.7662 +Query 1/1: Action query time = 1.622 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8501 +t=138: Selected seed 195 with value = 0.8501 +Query 1/1: Action query time = 1.571 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9756 +t=154: Selected seed 195 with value = 0.9756 +Query 1/1: Action query time = 1.403 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9835 +t=170: Selected seed 195 with value = 0.9835 +Query 1/1: Action query time = 1.908 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9721 +t=186: Selected seed 195 with value = 0.9721 +Query 1/1: Action query time = 2.090 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9728 +t=202: Selected seed 195 with value = 0.9728 +Query 1/1: Action query time = 1.567 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9636 +t=218: Selected seed 195 with value = 0.9636 +Query 1/1: Action query time = 1.635 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3223 +t=234: Selected seed 195 with value = 0.3223 +Query 1/1: Action query time = 1.585 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3253 +t=250: Selected seed 195 with value = 0.3253 +Query 1/1: Action query time = 1.515 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3216 +t=266: Selected seed 195 with value = 0.3216 +Query 1/1: Action query time = 1.896 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3387 +t=282: Selected seed 195 with value = 0.3387 +Query 1/1: Action query time = 1.800 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3464 +t=298: Selected seed 195 with value = 0.3464 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s1/2026_08_03-12_46_21--with_future_img--episode=6--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 6 +# successes: 4 (66.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 7... +Query 1/1: Action query time = 0.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2116 +t=10: Selected seed 195 with value = 0.2116 +Query 1/1: Action query time = 0.977 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2419 +t=26: Selected seed 195 with value = 0.2419 +Query 1/1: Action query time = 0.985 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3184 +t=42: Selected seed 195 with value = 0.3184 +Query 1/1: Action query time = 0.989 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3392 +t=58: Selected seed 195 with value = 0.3392 +Query 1/1: Action query time = 0.975 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3942 +t=74: Selected seed 195 with value = 0.3942 +Query 1/1: Action query time = 0.965 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4737 +t=90: Selected seed 195 with value = 0.4737 +Query 1/1: Action query time = 0.969 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5915 +t=106: Selected seed 195 with value = 0.5915 +Query 1/1: Action query time = 0.975 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6923 +t=122: Selected seed 195 with value = 0.6923 +Query 1/1: Action query time = 0.967 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7106 +t=138: Selected seed 195 with value = 0.7106 +Query 1/1: Action query time = 0.975 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8556 +t=154: Selected seed 195 with value = 0.8556 +Query 1/1: Action query time = 0.977 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9750 +t=170: Selected seed 195 with value = 0.9750 +Query 1/1: Action query time = 1.006 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s1/2026_08_03-12_46_21--with_future_img--episode=7--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 7 +# successes: 5 (71.4%) +Current task success rate: 0.7142857142857143 +Current total success rate: 0.7142857142857143 +Final results: +Total episodes: 7 +Total successes: 5 +Overall success rate: 0.7143 (71.4%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-12_46_21--forgetcl_i175_t3_s4.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-12_46_21--forgetcl_i175_t3_s4.txt new file mode 100644 index 0000000000000000000000000000000000000000..0750b50e39ac1907e343c5d78e7bbbf50542a842 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-12_46_21--forgetcl_i175_t3_s4.txt @@ -0,0 +1,370 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='forgetcl_i175_t3_s4', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='4,12,20,28,36,44', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 2.064 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2441 +t=10: Selected seed 195 with value = 0.2441 +Query 1/1: Action query time = 2.186 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2663 +t=26: Selected seed 195 with value = 0.2663 +Query 1/1: Action query time = 2.968 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2969 +t=42: Selected seed 195 with value = 0.2969 +Query 1/1: Action query time = 2.692 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3440 +t=58: Selected seed 195 with value = 0.3440 +Query 1/1: Action query time = 2.393 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4101 +t=74: Selected seed 195 with value = 0.4101 +Query 1/1: Action query time = 2.637 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4764 +t=90: Selected seed 195 with value = 0.4764 +Query 1/1: Action query time = 2.585 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5694 +t=106: Selected seed 195 with value = 0.5694 +Query 1/1: Action query time = 2.394 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6055 +t=122: Selected seed 195 with value = 0.6055 +Query 1/1: Action query time = 2.384 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6584 +t=138: Selected seed 195 with value = 0.6584 +Query 1/1: Action query time = 2.387 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8627 +t=154: Selected seed 195 with value = 0.8627 +Query 1/1: Action query time = 2.122 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9736 +t=170: Selected seed 195 with value = 0.9736 +Query 1/1: Action query time = 1.894 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s4/2026_08_03-12_46_21--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 1.561 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2281 +t=10: Selected seed 195 with value = 0.2281 +Query 1/1: Action query time = 2.600 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2545 +t=26: Selected seed 195 with value = 0.2545 +Query 1/1: Action query time = 2.676 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3024 +t=42: Selected seed 195 with value = 0.3024 +Query 1/1: Action query time = 2.624 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3324 +t=58: Selected seed 195 with value = 0.3324 +Query 1/1: Action query time = 2.561 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3947 +t=74: Selected seed 195 with value = 0.3947 +Query 1/1: Action query time = 2.049 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4886 +t=90: Selected seed 195 with value = 0.4886 +Query 1/1: Action query time = 2.232 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5742 +t=106: Selected seed 195 with value = 0.5742 +Query 1/1: Action query time = 2.544 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6532 +t=122: Selected seed 195 with value = 0.6532 +Query 1/1: Action query time = 2.499 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7361 +t=138: Selected seed 195 with value = 0.7361 +Query 1/1: Action query time = 2.338 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9059 +t=154: Selected seed 195 with value = 0.9059 +Query 1/1: Action query time = 2.479 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.387 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s4/2026_08_03-12_46_21--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 1.917 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2429 +t=10: Selected seed 195 with value = 0.2429 +Query 1/1: Action query time = 2.787 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2613 +t=26: Selected seed 195 with value = 0.2613 +Query 1/1: Action query time = 2.528 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3016 +t=42: Selected seed 195 with value = 0.3016 +Query 1/1: Action query time = 1.679 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3465 +t=58: Selected seed 195 with value = 0.3465 +Query 1/1: Action query time = 1.874 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4110 +t=74: Selected seed 195 with value = 0.4110 +Query 1/1: Action query time = 2.499 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4735 +t=90: Selected seed 195 with value = 0.4735 +Query 1/1: Action query time = 3.085 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5684 +t=106: Selected seed 195 with value = 0.5684 +Query 1/1: Action query time = 3.163 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5891 +t=122: Selected seed 195 with value = 0.5891 +Query 1/1: Action query time = 2.820 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6241 +t=138: Selected seed 195 with value = 0.6241 +Query 1/1: Action query time = 2.224 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8101 +t=154: Selected seed 195 with value = 0.8101 +Query 1/1: Action query time = 2.727 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9394 +t=170: Selected seed 195 with value = 0.9394 +Query 1/1: Action query time = 2.307 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=186: Selected seed 195 with value = 0.9986 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s4/2026_08_03-12_46_21--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 1.482 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2342 +t=10: Selected seed 195 with value = 0.2342 +Query 1/1: Action query time = 1.800 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2553 +t=26: Selected seed 195 with value = 0.2553 +Query 1/1: Action query time = 1.895 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3133 +t=42: Selected seed 195 with value = 0.3133 +Query 1/1: Action query time = 2.248 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3624 +t=58: Selected seed 195 with value = 0.3624 +Query 1/1: Action query time = 2.603 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4237 +t=74: Selected seed 195 with value = 0.4237 +Query 1/1: Action query time = 2.086 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4820 +t=90: Selected seed 195 with value = 0.4820 +Query 1/1: Action query time = 2.467 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5873 +t=106: Selected seed 195 with value = 0.5873 +Query 1/1: Action query time = 2.528 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6141 +t=122: Selected seed 195 with value = 0.6141 +Query 1/1: Action query time = 2.342 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7309 +t=138: Selected seed 195 with value = 0.7309 +Query 1/1: Action query time = 2.593 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8221 +t=154: Selected seed 195 with value = 0.8221 +Query 1/1: Action query time = 2.188 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9588 +t=170: Selected seed 195 with value = 0.9588 +Query 1/1: Action query time = 2.596 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9761 +t=186: Selected seed 195 with value = 0.9761 +Query 1/1: Action query time = 1.854 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9633 +t=202: Selected seed 195 with value = 0.9633 +Query 1/1: Action query time = 1.518 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9701 +t=218: Selected seed 195 with value = 0.9701 +Query 1/1: Action query time = 1.893 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9650 +t=234: Selected seed 195 with value = 0.9650 +Query 1/1: Action query time = 2.081 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9602 +t=250: Selected seed 195 with value = 0.9602 +Query 1/1: Action query time = 1.863 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9615 +t=266: Selected seed 195 with value = 0.9615 +Query 1/1: Action query time = 2.431 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9626 +t=282: Selected seed 195 with value = 0.9626 +Query 1/1: Action query time = 2.742 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9619 +t=298: Selected seed 195 with value = 0.9619 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s4/2026_08_03-12_46_21--with_future_img--episode=4--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 4 +# successes: 3 (75.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 5... +Query 1/1: Action query time = 2.588 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1977 +t=10: Selected seed 195 with value = 0.1977 +Query 1/1: Action query time = 2.204 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2477 +t=26: Selected seed 195 with value = 0.2477 +Query 1/1: Action query time = 2.288 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1581 +t=42: Selected seed 195 with value = 0.1581 +Query 1/1: Action query time = 2.232 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.1944 +t=58: Selected seed 195 with value = 0.1944 +Query 1/1: Action query time = 1.833 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2772 +t=74: Selected seed 195 with value = 0.2772 +Query 1/1: Action query time = 1.348 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4813 +t=90: Selected seed 195 with value = 0.4813 +Query 1/1: Action query time = 1.401 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5789 +t=106: Selected seed 195 with value = 0.5789 +Query 1/1: Action query time = 1.955 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6458 +t=122: Selected seed 195 with value = 0.6458 +Query 1/1: Action query time = 2.507 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6271 +t=138: Selected seed 195 with value = 0.6271 +Query 1/1: Action query time = 2.604 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8474 +t=154: Selected seed 195 with value = 0.8474 +Query 1/1: Action query time = 2.408 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9610 +t=170: Selected seed 195 with value = 0.9610 +Query 1/1: Action query time = 2.429 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s4/2026_08_03-12_46_21--with_future_img--episode=5--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 5 +# successes: 4 (80.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 6... +Query 1/1: Action query time = 1.921 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2270 +t=10: Selected seed 195 with value = 0.2270 +Query 1/1: Action query time = 1.585 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2463 +t=26: Selected seed 195 with value = 0.2463 +Query 1/1: Action query time = 1.781 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2995 +t=42: Selected seed 195 with value = 0.2995 +Query 1/1: Action query time = 1.855 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3255 +t=58: Selected seed 195 with value = 0.3255 +Query 1/1: Action query time = 1.850 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3771 +t=74: Selected seed 195 with value = 0.3771 +Query 1/1: Action query time = 1.731 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4848 +t=90: Selected seed 195 with value = 0.4848 +Query 1/1: Action query time = 1.670 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5952 +t=106: Selected seed 195 with value = 0.5952 +Query 1/1: Action query time = 1.464 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6831 +t=122: Selected seed 195 with value = 0.6831 +Query 1/1: Action query time = 1.493 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7526 +t=138: Selected seed 195 with value = 0.7526 +Query 1/1: Action query time = 1.474 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8883 +t=154: Selected seed 195 with value = 0.8883 +Query 1/1: Action query time = 1.355 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9913 +t=170: Selected seed 195 with value = 0.9913 +Query 1/1: Action query time = 1.521 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/forgetcl_i175_t3_s4/2026_08_03-12_46_21--with_future_img--episode=6--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 6 +# successes: 5 (83.3%) +Current task success rate: 0.8333333333333334 +Current total success rate: 0.8333333333333334 +Final results: +Total episodes: 6 +Total successes: 5 +Overall success rate: 0.8333 (83.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_33_20--fc350full_t0_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_33_20--fc350full_t0_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..1458b917dbf9e5e17008f029d6fc98a3ad6dae49 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_33_20--fc350full_t0_s0.txt @@ -0,0 +1,519 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fc350full_t0_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,4,8,12,16,20,24,28,32,36,40,44,48', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.094 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3931 +t=10: Selected seed 195 with value = 0.3931 +Query 1/1: Action query time = 4.134 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4728 +t=26: Selected seed 195 with value = 0.4728 +Query 1/1: Action query time = 4.843 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5644 +t=42: Selected seed 195 with value = 0.5644 +Query 1/1: Action query time = 5.171 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6396 +t=58: Selected seed 195 with value = 0.6396 +Query 1/1: Action query time = 4.917 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7767 +t=74: Selected seed 195 with value = 0.7767 +Query 1/1: Action query time = 4.582 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8878 +t=90: Selected seed 195 with value = 0.8878 +Query 1/1: Action query time = 3.703 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9809 +t=106: Selected seed 195 with value = 0.9809 +Query 1/1: Action query time = 4.527 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.800 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4096 +t=10: Selected seed 195 with value = 0.4096 +Query 1/1: Action query time = 5.175 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4778 +t=26: Selected seed 195 with value = 0.4778 +Query 1/1: Action query time = 5.302 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5508 +t=42: Selected seed 195 with value = 0.5508 +Query 1/1: Action query time = 5.565 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6509 +t=58: Selected seed 195 with value = 0.6509 +Query 1/1: Action query time = 4.717 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7470 +t=74: Selected seed 195 with value = 0.7470 +Query 1/1: Action query time = 5.512 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8709 +t=90: Selected seed 195 with value = 0.8709 +Query 1/1: Action query time = 4.614 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=106: Selected seed 195 with value = 0.9980 +Query 1/1: Action query time = 4.824 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=122: Selected seed 195 with value = 0.9988 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.162 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3900 +t=10: Selected seed 195 with value = 0.3900 +Query 1/1: Action query time = 4.284 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4702 +t=26: Selected seed 195 with value = 0.4702 +Query 1/1: Action query time = 5.442 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5510 +t=42: Selected seed 195 with value = 0.5510 +Query 1/1: Action query time = 4.599 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6698 +t=58: Selected seed 195 with value = 0.6698 +Query 1/1: Action query time = 4.677 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7664 +t=74: Selected seed 195 with value = 0.7664 +Query 1/1: Action query time = 4.477 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8994 +t=90: Selected seed 195 with value = 0.8994 +Query 1/1: Action query time = 4.388 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=106: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 2.435 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 4... +Query 1/1: Action query time = 4.702 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3858 +t=10: Selected seed 195 with value = 0.3858 +Query 1/1: Action query time = 5.229 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4455 +t=26: Selected seed 195 with value = 0.4455 +Query 1/1: Action query time = 4.879 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5412 +t=42: Selected seed 195 with value = 0.5412 +Query 1/1: Action query time = 4.681 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6491 +t=58: Selected seed 195 with value = 0.6491 +Query 1/1: Action query time = 4.802 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7594 +t=74: Selected seed 195 with value = 0.7594 +Query 1/1: Action query time = 4.527 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9049 +t=90: Selected seed 195 with value = 0.9049 +Query 1/1: Action query time = 3.339 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9870 +t=106: Selected seed 195 with value = 0.9870 +Query 1/1: Action query time = 3.944 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=122: Selected seed 195 with value = 0.9979 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=4--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 5... +Query 1/1: Action query time = 4.657 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4011 +t=10: Selected seed 195 with value = 0.4011 +Query 1/1: Action query time = 5.496 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4719 +t=26: Selected seed 195 with value = 0.4719 +Query 1/1: Action query time = 5.395 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5987 +t=42: Selected seed 195 with value = 0.5987 +Query 1/1: Action query time = 4.813 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6577 +t=58: Selected seed 195 with value = 0.6577 +Query 1/1: Action query time = 4.854 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7874 +t=74: Selected seed 195 with value = 0.7874 +Query 1/1: Action query time = 4.118 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8800 +t=90: Selected seed 195 with value = 0.8800 +Query 1/1: Action query time = 4.610 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9911 +t=106: Selected seed 195 with value = 0.9911 +Query 1/1: Action query time = 3.628 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=5--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 6... +Query 1/1: Action query time = 3.339 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3963 +t=10: Selected seed 195 with value = 0.3963 +Query 1/1: Action query time = 4.334 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3465 +t=26: Selected seed 195 with value = 0.3465 +Query 1/1: Action query time = 4.271 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5710 +t=42: Selected seed 195 with value = 0.5710 +Query 1/1: Action query time = 4.506 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6381 +t=58: Selected seed 195 with value = 0.6381 +Query 1/1: Action query time = 4.644 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7628 +t=74: Selected seed 195 with value = 0.7628 +Query 1/1: Action query time = 6.179 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8784 +t=90: Selected seed 195 with value = 0.8784 +Query 1/1: Action query time = 3.699 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9938 +t=106: Selected seed 195 with value = 0.9938 +Query 1/1: Action query time = 2.892 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=6--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 7... +Query 1/1: Action query time = 4.262 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4133 +t=10: Selected seed 195 with value = 0.4133 +Query 1/1: Action query time = 5.571 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4761 +t=26: Selected seed 195 with value = 0.4761 +Query 1/1: Action query time = 5.599 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5534 +t=42: Selected seed 195 with value = 0.5534 +Query 1/1: Action query time = 5.240 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6532 +t=58: Selected seed 195 with value = 0.6532 +Query 1/1: Action query time = 4.456 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7821 +t=74: Selected seed 195 with value = 0.7821 +Query 1/1: Action query time = 4.642 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8567 +t=90: Selected seed 195 with value = 0.8567 +Query 1/1: Action query time = 4.425 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=106: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 4.374 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=7--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 8... +Query 1/1: Action query time = 5.404 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3898 +t=10: Selected seed 195 with value = 0.3898 +Query 1/1: Action query time = 4.609 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4836 +t=26: Selected seed 195 with value = 0.4836 +Query 1/1: Action query time = 3.818 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5593 +t=42: Selected seed 195 with value = 0.5593 +Query 1/1: Action query time = 5.261 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6557 +t=58: Selected seed 195 with value = 0.6557 +Query 1/1: Action query time = 4.812 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7570 +t=74: Selected seed 195 with value = 0.7570 +Query 1/1: Action query time = 5.165 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8635 +t=90: Selected seed 195 with value = 0.8635 +Query 1/1: Action query time = 4.424 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=106: Selected seed 195 with value = 0.9963 +Query 1/1: Action query time = 5.329 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=8--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 9... +Query 1/1: Action query time = 3.570 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4018 +t=10: Selected seed 195 with value = 0.4018 +Query 1/1: Action query time = 4.928 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4650 +t=26: Selected seed 195 with value = 0.4650 +Query 1/1: Action query time = 4.773 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5614 +t=42: Selected seed 195 with value = 0.5614 +Query 1/1: Action query time = 4.641 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6570 +t=58: Selected seed 195 with value = 0.6570 +Query 1/1: Action query time = 3.714 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7554 +t=74: Selected seed 195 with value = 0.7554 +Query 1/1: Action query time = 5.170 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8511 +t=90: Selected seed 195 with value = 0.8511 +Query 1/1: Action query time = 3.350 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9882 +t=106: Selected seed 195 with value = 0.9882 +Query 1/1: Action query time = 5.104 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=9--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 10... +Query 1/1: Action query time = 5.597 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3468 +t=10: Selected seed 195 with value = 0.3468 +Query 1/1: Action query time = 5.024 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3416 +t=26: Selected seed 195 with value = 0.3416 +Query 1/1: Action query time = 4.046 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4260 +t=42: Selected seed 195 with value = 0.4260 +Query 1/1: Action query time = 4.778 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5360 +t=58: Selected seed 195 with value = 0.5360 +Query 1/1: Action query time = 5.378 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7202 +t=74: Selected seed 195 with value = 0.7202 +Query 1/1: Action query time = 5.250 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8400 +t=90: Selected seed 195 with value = 0.8400 +Query 1/1: Action query time = 4.179 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9970 +t=106: Selected seed 195 with value = 0.9970 +Query 1/1: Action query time = 4.661 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=122: Selected seed 195 with value = 0.9998 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=10--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 11... +Query 1/1: Action query time = 4.161 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4056 +t=10: Selected seed 195 with value = 0.4056 +Query 1/1: Action query time = 4.333 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4792 +t=26: Selected seed 195 with value = 0.4792 +Query 1/1: Action query time = 4.527 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5632 +t=42: Selected seed 195 with value = 0.5632 +Query 1/1: Action query time = 3.979 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6636 +t=58: Selected seed 195 with value = 0.6636 +Query 1/1: Action query time = 3.839 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7559 +t=74: Selected seed 195 with value = 0.7559 +Query 1/1: Action query time = 3.408 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9200 +t=90: Selected seed 195 with value = 0.9200 +Query 1/1: Action query time = 2.973 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=106: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 3.159 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=122: Selected seed 195 with value = 0.9979 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=11--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 12... +Query 1/1: Action query time = 4.196 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4007 +t=10: Selected seed 195 with value = 0.4007 +Query 1/1: Action query time = 3.809 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4710 +t=26: Selected seed 195 with value = 0.4710 +Query 1/1: Action query time = 3.152 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5830 +t=42: Selected seed 195 with value = 0.5830 +Query 1/1: Action query time = 3.477 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6198 +t=58: Selected seed 195 with value = 0.6198 +Query 1/1: Action query time = 4.338 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6988 +t=74: Selected seed 195 with value = 0.6988 +Query 1/1: Action query time = 2.302 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8584 +t=90: Selected seed 195 with value = 0.8584 +Query 1/1: Action query time = 1.837 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9822 +t=106: Selected seed 195 with value = 0.9822 +Query 1/1: Action query time = 3.713 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=12--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 13... +Query 1/1: Action query time = 3.095 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4149 +t=10: Selected seed 195 with value = 0.4149 +Query 1/1: Action query time = 3.070 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4863 +t=26: Selected seed 195 with value = 0.4863 +Query 1/1: Action query time = 2.669 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5590 +t=42: Selected seed 195 with value = 0.5590 +Query 1/1: Action query time = 1.658 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6762 +t=58: Selected seed 195 with value = 0.6762 +Query 1/1: Action query time = 1.435 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7772 +t=74: Selected seed 195 with value = 0.7772 +Query 1/1: Action query time = 2.608 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8930 +t=90: Selected seed 195 with value = 0.8930 +Query 1/1: Action query time = 2.329 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9952 +t=106: Selected seed 195 with value = 0.9952 +Query 1/1: Action query time = 2.344 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t0_s0/2026_08_03-14_33_20--with_future_img--episode=13--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 13 +Total successes: 13 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_33_21--fc350full_t1_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_33_21--fc350full_t1_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..a440e5a4d2ad085ba15b97d3e9249fec62d1df5e --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_33_21--fc350full_t1_s1.txt @@ -0,0 +1,627 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fc350full_t1_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 3.499 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5974 +t=10: Selected seed 195 with value = 0.5974 +Query 1/1: Action query time = 4.607 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6706 +t=26: Selected seed 195 with value = 0.6706 +Query 1/1: Action query time = 4.472 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7292 +t=42: Selected seed 195 with value = 0.7292 +Query 1/1: Action query time = 5.658 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8941 +t=58: Selected seed 195 with value = 0.8941 +Query 1/1: Action query time = 5.725 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9996 +t=74: Selected seed 195 with value = 0.9996 +Query 1/1: Action query time = 4.395 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.468 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.956 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8727 +t=122: Selected seed 195 with value = 0.8727 +Query 1/1: Action query time = 4.660 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8240 +t=138: Selected seed 195 with value = 0.8240 +Query 1/1: Action query time = 3.438 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8959 +t=154: Selected seed 195 with value = 0.8959 +Query 1/1: Action query time = 4.622 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9766 +t=170: Selected seed 195 with value = 0.9766 +Query 1/1: Action query time = 5.027 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9951 +t=186: Selected seed 195 with value = 0.9951 +Query 1/1: Action query time = 5.256 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=202: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 4.433 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.664 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.365 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=250: Selected seed 195 with value = 0.9903 +Query 1/1: Action query time = 4.793 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9593 +t=266: Selected seed 195 with value = 0.9593 +Query 1/1: Action query time = 4.479 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8542 +t=282: Selected seed 195 with value = 0.8542 +Query 1/1: Action query time = 4.830 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8289 +t=298: Selected seed 195 with value = 0.8289 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 4.582 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6163 +t=10: Selected seed 195 with value = 0.6163 +Query 1/1: Action query time = 5.203 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8181 +t=26: Selected seed 195 with value = 0.8181 +Query 1/1: Action query time = 4.754 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8124 +t=42: Selected seed 195 with value = 0.8124 +Query 1/1: Action query time = 4.761 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9020 +t=58: Selected seed 195 with value = 0.9020 +Query 1/1: Action query time = 5.224 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.525 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 2.375 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5690 +t=10: Selected seed 195 with value = 0.5690 +Query 1/1: Action query time = 4.172 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8285 +t=26: Selected seed 195 with value = 0.8285 +Query 1/1: Action query time = 4.823 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7960 +t=42: Selected seed 195 with value = 0.7960 +Query 1/1: Action query time = 5.117 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9251 +t=58: Selected seed 195 with value = 0.9251 +Query 1/1: Action query time = 4.783 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.471 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 3.254 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5881 +t=10: Selected seed 195 with value = 0.5881 +Query 1/1: Action query time = 3.606 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6939 +t=26: Selected seed 195 with value = 0.6939 +Query 1/1: Action query time = 3.835 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7734 +t=42: Selected seed 195 with value = 0.7734 +Query 1/1: Action query time = 4.297 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8936 +t=58: Selected seed 195 with value = 0.8936 +Query 1/1: Action query time = 4.704 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9793 +t=74: Selected seed 195 with value = 0.9793 +Query 1/1: Action query time = 5.416 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9501 +t=90: Selected seed 195 with value = 0.9501 +Query 1/1: Action query time = 4.772 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9586 +t=106: Selected seed 195 with value = 0.9586 +Query 1/1: Action query time = 3.836 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9077 +t=122: Selected seed 195 with value = 0.9077 +Query 1/1: Action query time = 3.447 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8957 +t=138: Selected seed 195 with value = 0.8957 +Query 1/1: Action query time = 4.729 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8906 +t=154: Selected seed 195 with value = 0.8906 +Query 1/1: Action query time = 5.010 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8964 +t=170: Selected seed 195 with value = 0.8964 +Query 1/1: Action query time = 4.455 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9114 +t=186: Selected seed 195 with value = 0.9114 +Query 1/1: Action query time = 4.225 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9672 +t=202: Selected seed 195 with value = 0.9672 +Query 1/1: Action query time = 5.393 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9680 +t=218: Selected seed 195 with value = 0.9680 +Query 1/1: Action query time = 4.137 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9784 +t=234: Selected seed 195 with value = 0.9784 +Query 1/1: Action query time = 4.197 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9823 +t=250: Selected seed 195 with value = 0.9823 +Query 1/1: Action query time = 4.470 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9939 +t=266: Selected seed 195 with value = 0.9939 +Query 1/1: Action query time = 3.618 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9640 +t=282: Selected seed 195 with value = 0.9640 +Query 1/1: Action query time = 6.147 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9842 +t=298: Selected seed 195 with value = 0.9842 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 4 +# successes: 2 (50.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 3.883 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6127 +t=10: Selected seed 195 with value = 0.6127 +Query 1/1: Action query time = 3.960 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8396 +t=26: Selected seed 195 with value = 0.8396 +Query 1/1: Action query time = 5.211 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7411 +t=42: Selected seed 195 with value = 0.7411 +Query 1/1: Action query time = 4.782 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8948 +t=58: Selected seed 195 with value = 0.8948 +Query 1/1: Action query time = 4.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.621 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=5--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 3 (60.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 4.382 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5876 +t=10: Selected seed 195 with value = 0.5876 +Query 1/1: Action query time = 4.461 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7715 +t=26: Selected seed 195 with value = 0.7715 +Query 1/1: Action query time = 4.412 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8004 +t=42: Selected seed 195 with value = 0.8004 +Query 1/1: Action query time = 4.444 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8635 +t=58: Selected seed 195 with value = 0.8635 +Query 1/1: Action query time = 5.631 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=74: Selected seed 195 with value = 0.9997 +Query 1/1: Action query time = 4.513 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=6--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 4 (66.7%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 5.601 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6271 +t=10: Selected seed 195 with value = 0.6271 +Query 1/1: Action query time = 5.025 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7581 +t=26: Selected seed 195 with value = 0.7581 +Query 1/1: Action query time = 2.437 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7720 +t=42: Selected seed 195 with value = 0.7720 +Query 1/1: Action query time = 3.959 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8457 +t=58: Selected seed 195 with value = 0.8457 +Query 1/1: Action query time = 3.728 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9047 +t=74: Selected seed 195 with value = 0.9047 +Query 1/1: Action query time = 5.478 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9947 +t=90: Selected seed 195 with value = 0.9947 +Query 1/1: Action query time = 3.632 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=7--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 5 (71.4%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 5.745 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6336 +t=10: Selected seed 195 with value = 0.6336 +Query 1/1: Action query time = 4.169 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6738 +t=26: Selected seed 195 with value = 0.6738 +Query 1/1: Action query time = 5.294 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8220 +t=42: Selected seed 195 with value = 0.8220 +Query 1/1: Action query time = 4.234 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9438 +t=58: Selected seed 195 with value = 0.9438 +Query 1/1: Action query time = 4.961 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=74: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 3.309 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=8--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 6 (75.0%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 5.098 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5481 +t=10: Selected seed 195 with value = 0.5481 +Query 1/1: Action query time = 3.112 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6466 +t=26: Selected seed 195 with value = 0.6466 +Query 1/1: Action query time = 4.824 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7397 +t=42: Selected seed 195 with value = 0.7397 +Query 1/1: Action query time = 4.996 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8633 +t=58: Selected seed 195 with value = 0.8633 +Query 1/1: Action query time = 5.889 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9632 +t=74: Selected seed 195 with value = 0.9632 +Query 1/1: Action query time = 3.971 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9127 +t=90: Selected seed 195 with value = 0.9127 +Query 1/1: Action query time = 4.171 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9636 +t=106: Selected seed 195 with value = 0.9636 +Query 1/1: Action query time = 4.951 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9253 +t=122: Selected seed 195 with value = 0.9253 +Query 1/1: Action query time = 3.554 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9052 +t=138: Selected seed 195 with value = 0.9052 +Query 1/1: Action query time = 5.158 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9062 +t=154: Selected seed 195 with value = 0.9062 +Query 1/1: Action query time = 3.942 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9110 +t=170: Selected seed 195 with value = 0.9110 +Query 1/1: Action query time = 4.103 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9391 +t=186: Selected seed 195 with value = 0.9391 +Query 1/1: Action query time = 1.901 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9833 +t=202: Selected seed 195 with value = 0.9833 +Query 1/1: Action query time = 4.212 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9745 +t=218: Selected seed 195 with value = 0.9745 +Query 1/1: Action query time = 4.353 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9842 +t=234: Selected seed 195 with value = 0.9842 +Query 1/1: Action query time = 3.351 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9906 +t=250: Selected seed 195 with value = 0.9906 +Query 1/1: Action query time = 4.104 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=266: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 3.194 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9585 +t=282: Selected seed 195 with value = 0.9585 +Query 1/1: Action query time = 2.682 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9783 +t=298: Selected seed 195 with value = 0.9783 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 9 +# successes: 6 (66.7%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 4.248 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5616 +t=10: Selected seed 195 with value = 0.5616 +Query 1/1: Action query time = 3.750 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6644 +t=26: Selected seed 195 with value = 0.6644 +Query 1/1: Action query time = 3.237 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7845 +t=42: Selected seed 195 with value = 0.7845 +Query 1/1: Action query time = 3.004 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9193 +t=58: Selected seed 195 with value = 0.9193 +Query 1/1: Action query time = 3.289 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.396 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=10--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 7 (70.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 2.871 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5688 +t=10: Selected seed 195 with value = 0.5688 +Query 1/1: Action query time = 1.750 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6546 +t=26: Selected seed 195 with value = 0.6546 +Query 1/1: Action query time = 2.239 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7570 +t=42: Selected seed 195 with value = 0.7570 +Query 1/1: Action query time = 3.001 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9120 +t=58: Selected seed 195 with value = 0.9120 +Query 1/1: Action query time = 2.528 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=74: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 2.654 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.071 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=106: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 2.067 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8851 +t=122: Selected seed 195 with value = 0.8851 +Query 1/1: Action query time = 2.347 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8048 +t=138: Selected seed 195 with value = 0.8048 +Query 1/1: Action query time = 2.452 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8962 +t=154: Selected seed 195 with value = 0.8962 +Query 1/1: Action query time = 1.942 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9790 +t=170: Selected seed 195 with value = 0.9790 +Query 1/1: Action query time = 1.237 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.263 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=202: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 1.101 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.068 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=234: Selected seed 195 with value = 0.9993 +Query 1/1: Action query time = 1.132 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9601 +t=250: Selected seed 195 with value = 0.9601 +Query 1/1: Action query time = 0.967 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9204 +t=266: Selected seed 195 with value = 0.9204 +Query 1/1: Action query time = 1.316 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9261 +t=282: Selected seed 195 with value = 0.9261 +Query 1/1: Action query time = 0.968 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9338 +t=298: Selected seed 195 with value = 0.9338 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=11--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 11 +# successes: 7 (63.6%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 0.969 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6597 +t=10: Selected seed 195 with value = 0.6597 +Query 1/1: Action query time = 1.524 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8260 +t=26: Selected seed 195 with value = 0.8260 +Query 1/1: Action query time = 0.949 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8054 +t=42: Selected seed 195 with value = 0.8054 +Query 1/1: Action query time = 0.949 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8663 +t=58: Selected seed 195 with value = 0.8663 +Query 1/1: Action query time = 0.948 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=74: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 0.948 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=12--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 8 (66.7%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 1.024 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6094 +t=10: Selected seed 195 with value = 0.6094 +Query 1/1: Action query time = 1.311 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7400 +t=26: Selected seed 195 with value = 0.7400 +Query 1/1: Action query time = 1.152 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8347 +t=42: Selected seed 195 with value = 0.8347 +Query 1/1: Action query time = 0.963 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9654 +t=58: Selected seed 195 with value = 0.9654 +Query 1/1: Action query time = 0.964 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.140 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t1_s1/2026_08_03-14_33_21--with_future_img--episode=13--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 13 +# successes: 9 (69.2%) +Current task success rate: 0.6923076923076923 +Current total success rate: 0.6923076923076923 +Final results: +Total episodes: 13 +Total successes: 9 +Overall success rate: 0.6923 (69.2%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_46_55--fc350full_t4_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_46_55--fc350full_t4_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..6d5ac9eb1ea67191170ef670dac4b0002f3ca450 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_46_55--fc350full_t4_s0.txt @@ -0,0 +1,415 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fc350full_t4_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,4,8,12,16,20,24,28,32,36,40,44,48', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 1.558 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5484 +t=10: Selected seed 195 with value = 0.5484 +Query 1/1: Action query time = 1.608 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6893 +t=26: Selected seed 195 with value = 0.6893 +Query 1/1: Action query time = 3.170 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7821 +t=42: Selected seed 195 with value = 0.7821 +Query 1/1: Action query time = 5.503 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9341 +t=58: Selected seed 195 with value = 0.9341 +Query 1/1: Action query time = 4.923 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.254 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.849 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5563 +t=10: Selected seed 195 with value = 0.5563 +Query 1/1: Action query time = 3.826 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6891 +t=26: Selected seed 195 with value = 0.6891 +Query 1/1: Action query time = 3.673 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7607 +t=42: Selected seed 195 with value = 0.7607 +Query 1/1: Action query time = 4.639 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9264 +t=58: Selected seed 195 with value = 0.9264 +Query 1/1: Action query time = 5.310 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.802 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.835 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5691 +t=10: Selected seed 195 with value = 0.5691 +Query 1/1: Action query time = 4.598 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6995 +t=26: Selected seed 195 with value = 0.6995 +Query 1/1: Action query time = 4.596 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7845 +t=42: Selected seed 195 with value = 0.7845 +Query 1/1: Action query time = 4.269 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9572 +t=58: Selected seed 195 with value = 0.9572 +Query 1/1: Action query time = 4.373 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.058 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 4.364 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5675 +t=10: Selected seed 195 with value = 0.5675 +Query 1/1: Action query time = 4.080 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6629 +t=26: Selected seed 195 with value = 0.6629 +Query 1/1: Action query time = 3.797 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7863 +t=42: Selected seed 195 with value = 0.7863 +Query 1/1: Action query time = 1.874 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9025 +t=58: Selected seed 195 with value = 0.9025 +Query 1/1: Action query time = 3.495 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9918 +t=74: Selected seed 195 with value = 0.9918 +Query 1/1: Action query time = 5.093 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=4--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 5... +Query 1/1: Action query time = 5.040 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5930 +t=10: Selected seed 195 with value = 0.5930 +Query 1/1: Action query time = 2.995 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6959 +t=26: Selected seed 195 with value = 0.6959 +Query 1/1: Action query time = 4.760 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8396 +t=42: Selected seed 195 with value = 0.8396 +Query 1/1: Action query time = 4.035 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9635 +t=58: Selected seed 195 with value = 0.9635 +Query 1/1: Action query time = 5.272 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.690 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=5--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 6... +Query 1/1: Action query time = 3.937 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5801 +t=10: Selected seed 195 with value = 0.5801 +Query 1/1: Action query time = 3.019 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6732 +t=26: Selected seed 195 with value = 0.6732 +Query 1/1: Action query time = 4.869 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8163 +t=42: Selected seed 195 with value = 0.8163 +Query 1/1: Action query time = 4.468 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9297 +t=58: Selected seed 195 with value = 0.9297 +Query 1/1: Action query time = 4.674 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=74: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 4.935 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=6--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 7... +Query 1/1: Action query time = 3.784 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5625 +t=10: Selected seed 195 with value = 0.5625 +Query 1/1: Action query time = 4.425 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6986 +t=26: Selected seed 195 with value = 0.6986 +Query 1/1: Action query time = 3.501 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7967 +t=42: Selected seed 195 with value = 0.7967 +Query 1/1: Action query time = 5.284 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9549 +t=58: Selected seed 195 with value = 0.9549 +Query 1/1: Action query time = 2.932 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.220 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=7--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 8... +Query 1/1: Action query time = 4.957 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5570 +t=10: Selected seed 195 with value = 0.5570 +Query 1/1: Action query time = 5.237 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6424 +t=26: Selected seed 195 with value = 0.6424 +Query 1/1: Action query time = 4.693 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7581 +t=42: Selected seed 195 with value = 0.7581 +Query 1/1: Action query time = 5.067 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9171 +t=58: Selected seed 195 with value = 0.9171 +Query 1/1: Action query time = 4.603 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=74: Selected seed 195 with value = 0.9986 +Query 1/1: Action query time = 4.593 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=8--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 9... +Query 1/1: Action query time = 5.260 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6016 +t=10: Selected seed 195 with value = 0.6016 +Query 1/1: Action query time = 3.813 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7148 +t=26: Selected seed 195 with value = 0.7148 +Query 1/1: Action query time = 3.513 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6769 +t=42: Selected seed 195 with value = 0.6769 +Query 1/1: Action query time = 4.930 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9218 +t=58: Selected seed 195 with value = 0.9218 +Query 1/1: Action query time = 4.819 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.210 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=9--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 10... +Query 1/1: Action query time = 5.831 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5631 +t=10: Selected seed 195 with value = 0.5631 +Query 1/1: Action query time = 5.177 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6845 +t=26: Selected seed 195 with value = 0.6845 +Query 1/1: Action query time = 4.875 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7779 +t=42: Selected seed 195 with value = 0.7779 +Query 1/1: Action query time = 5.008 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9504 +t=58: Selected seed 195 with value = 0.9504 +Query 1/1: Action query time = 2.814 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.153 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=10--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 11... +Query 1/1: Action query time = 3.257 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5742 +t=10: Selected seed 195 with value = 0.5742 +Query 1/1: Action query time = 5.019 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6886 +t=26: Selected seed 195 with value = 0.6886 +Query 1/1: Action query time = 4.388 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7670 +t=42: Selected seed 195 with value = 0.7670 +Query 1/1: Action query time = 4.652 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9098 +t=58: Selected seed 195 with value = 0.9098 +Query 1/1: Action query time = 4.737 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=74: Selected seed 195 with value = 0.9962 +Query 1/1: Action query time = 3.470 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=11--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 12... +Query 1/1: Action query time = 4.688 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=10: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 4.600 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7060 +t=26: Selected seed 195 with value = 0.7060 +Query 1/1: Action query time = 3.258 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7994 +t=42: Selected seed 195 with value = 0.7994 +Query 1/1: Action query time = 4.643 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9458 +t=58: Selected seed 195 with value = 0.9458 +Query 1/1: Action query time = 4.078 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.115 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=12--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 13... +Query 1/1: Action query time = 5.579 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5683 +t=10: Selected seed 195 with value = 0.5683 +Query 1/1: Action query time = 5.452 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6752 +t=26: Selected seed 195 with value = 0.6752 +Query 1/1: Action query time = 4.564 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7670 +t=42: Selected seed 195 with value = 0.7670 +Query 1/1: Action query time = 4.721 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9349 +t=58: Selected seed 195 with value = 0.9349 +Query 1/1: Action query time = 4.535 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.281 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t4_s0/2026_08_03-14_46_55--with_future_img--episode=13--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 13 +Total successes: 13 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_46_55--fc350full_t5_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_46_55--fc350full_t5_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..c67ceef8505aa7aadfd5cf0e1cfb07094d051df0 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_46_55--fc350full_t5_s0.txt @@ -0,0 +1,623 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fc350full_t5_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,4,8,12,16,20,24,28,32,36,40,44,48', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 5.576 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3784 +t=10: Selected seed 195 with value = 0.3784 +Query 1/1: Action query time = 5.324 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4421 +t=26: Selected seed 195 with value = 0.4421 +Query 1/1: Action query time = 5.332 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5224 +t=42: Selected seed 195 with value = 0.5224 +Query 1/1: Action query time = 4.223 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5796 +t=58: Selected seed 195 with value = 0.5796 +Query 1/1: Action query time = 4.695 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6398 +t=74: Selected seed 195 with value = 0.6398 +Query 1/1: Action query time = 4.920 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7325 +t=90: Selected seed 195 with value = 0.7325 +Query 1/1: Action query time = 3.789 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9231 +t=106: Selected seed 195 with value = 0.9231 +Query 1/1: Action query time = 4.423 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.468 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.839 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3465 +t=10: Selected seed 195 with value = 0.3465 +Query 1/1: Action query time = 4.599 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4311 +t=26: Selected seed 195 with value = 0.4311 +Query 1/1: Action query time = 4.874 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5526 +t=42: Selected seed 195 with value = 0.5526 +Query 1/1: Action query time = 5.097 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5914 +t=58: Selected seed 195 with value = 0.5914 +Query 1/1: Action query time = 4.546 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6801 +t=74: Selected seed 195 with value = 0.6801 +Query 1/1: Action query time = 5.583 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7916 +t=90: Selected seed 195 with value = 0.7916 +Query 1/1: Action query time = 4.812 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8902 +t=106: Selected seed 195 with value = 0.8902 +Query 1/1: Action query time = 4.533 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.229 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 4.306 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4095 +t=10: Selected seed 195 with value = 0.4095 +Query 1/1: Action query time = 4.544 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4263 +t=26: Selected seed 195 with value = 0.4263 +Query 1/1: Action query time = 4.328 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5194 +t=42: Selected seed 195 with value = 0.5194 +Query 1/1: Action query time = 4.160 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5365 +t=58: Selected seed 195 with value = 0.5365 +Query 1/1: Action query time = 5.340 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5617 +t=74: Selected seed 195 with value = 0.5617 +Query 1/1: Action query time = 4.179 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6462 +t=90: Selected seed 195 with value = 0.6462 +Query 1/1: Action query time = 3.525 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7921 +t=106: Selected seed 195 with value = 0.7921 +Query 1/1: Action query time = 5.507 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9163 +t=122: Selected seed 195 with value = 0.9163 +Query 1/1: Action query time = 4.640 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9778 +t=138: Selected seed 195 with value = 0.9778 +Query 1/1: Action query time = 4.619 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 4.787 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3841 +t=10: Selected seed 195 with value = 0.3841 +Query 1/1: Action query time = 3.724 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4423 +t=26: Selected seed 195 with value = 0.4423 +Query 1/1: Action query time = 5.130 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5296 +t=42: Selected seed 195 with value = 0.5296 +Query 1/1: Action query time = 4.757 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6426 +t=58: Selected seed 195 with value = 0.6426 +Query 1/1: Action query time = 3.364 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7503 +t=74: Selected seed 195 with value = 0.7503 +Query 1/1: Action query time = 5.567 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8345 +t=90: Selected seed 195 with value = 0.8345 +Query 1/1: Action query time = 4.579 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9428 +t=106: Selected seed 195 with value = 0.9428 +Query 1/1: Action query time = 5.102 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=122: Selected seed 195 with value = 0.9946 +Query 1/1: Action query time = 3.393 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=4--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 5... +Query 1/1: Action query time = 5.494 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3600 +t=10: Selected seed 195 with value = 0.3600 +Query 1/1: Action query time = 5.368 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4264 +t=26: Selected seed 195 with value = 0.4264 +Query 1/1: Action query time = 4.686 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5239 +t=42: Selected seed 195 with value = 0.5239 +Query 1/1: Action query time = 5.051 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6222 +t=58: Selected seed 195 with value = 0.6222 +Query 1/1: Action query time = 4.557 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6561 +t=74: Selected seed 195 with value = 0.6561 +Query 1/1: Action query time = 4.299 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7119 +t=90: Selected seed 195 with value = 0.7119 +Query 1/1: Action query time = 4.477 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8554 +t=106: Selected seed 195 with value = 0.8554 +Query 1/1: Action query time = 4.610 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9592 +t=122: Selected seed 195 with value = 0.9592 +Query 1/1: Action query time = 4.303 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.840 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9026 +t=154: Selected seed 195 with value = 0.9026 +Query 1/1: Action query time = 5.213 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8647 +t=170: Selected seed 195 with value = 0.8647 +Query 1/1: Action query time = 4.615 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9846 +t=186: Selected seed 195 with value = 0.9846 +Query 1/1: Action query time = 4.897 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9910 +t=202: Selected seed 195 with value = 0.9910 +Query 1/1: Action query time = 4.768 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9863 +t=218: Selected seed 195 with value = 0.9863 +Query 1/1: Action query time = 4.660 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9813 +t=234: Selected seed 195 with value = 0.9813 +Query 1/1: Action query time = 4.830 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=250: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 4.596 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9732 +t=266: Selected seed 195 with value = 0.9732 +Query 1/1: Action query time = 4.493 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9654 +t=282: Selected seed 195 with value = 0.9654 +Query 1/1: Action query time = 3.731 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9609 +t=298: Selected seed 195 with value = 0.9609 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=5--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 5 +# successes: 4 (80.0%) + +Task: push the plate to the front of the stove +Starting episode 6... +Query 1/1: Action query time = 5.146 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3780 +t=10: Selected seed 195 with value = 0.3780 +Query 1/1: Action query time = 4.440 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4381 +t=26: Selected seed 195 with value = 0.4381 +Query 1/1: Action query time = 4.943 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5166 +t=42: Selected seed 195 with value = 0.5166 +Query 1/1: Action query time = 5.130 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5845 +t=58: Selected seed 195 with value = 0.5845 +Query 1/1: Action query time = 4.161 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7084 +t=74: Selected seed 195 with value = 0.7084 +Query 1/1: Action query time = 3.516 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8072 +t=90: Selected seed 195 with value = 0.8072 +Query 1/1: Action query time = 5.078 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9585 +t=106: Selected seed 195 with value = 0.9585 +Query 1/1: Action query time = 4.575 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.153 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=6--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 6 +# successes: 5 (83.3%) + +Task: push the plate to the front of the stove +Starting episode 7... +Query 1/1: Action query time = 4.161 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3646 +t=10: Selected seed 195 with value = 0.3646 +Query 1/1: Action query time = 3.112 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4311 +t=26: Selected seed 195 with value = 0.4311 +Query 1/1: Action query time = 5.551 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5449 +t=42: Selected seed 195 with value = 0.5449 +Query 1/1: Action query time = 4.639 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6302 +t=58: Selected seed 195 with value = 0.6302 +Query 1/1: Action query time = 4.663 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7438 +t=74: Selected seed 195 with value = 0.7438 +Query 1/1: Action query time = 4.872 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8722 +t=90: Selected seed 195 with value = 0.8722 +Query 1/1: Action query time = 4.790 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=106: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 3.994 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.083 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=7--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 7 +# successes: 6 (85.7%) + +Task: push the plate to the front of the stove +Starting episode 8... +Query 1/1: Action query time = 4.453 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3806 +t=10: Selected seed 195 with value = 0.3806 +Query 1/1: Action query time = 4.031 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4409 +t=26: Selected seed 195 with value = 0.4409 +Query 1/1: Action query time = 3.802 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5207 +t=42: Selected seed 195 with value = 0.5207 +Query 1/1: Action query time = 4.018 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5543 +t=58: Selected seed 195 with value = 0.5543 +Query 1/1: Action query time = 3.960 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6127 +t=74: Selected seed 195 with value = 0.6127 +Query 1/1: Action query time = 3.443 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7743 +t=90: Selected seed 195 with value = 0.7743 +Query 1/1: Action query time = 2.460 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8935 +t=106: Selected seed 195 with value = 0.8935 +Query 1/1: Action query time = 3.678 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9459 +t=122: Selected seed 195 with value = 0.9459 +Query 1/1: Action query time = 3.953 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=138: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 3.645 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=8--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 8 +# successes: 7 (87.5%) + +Task: push the plate to the front of the stove +Starting episode 9... +Query 1/1: Action query time = 1.693 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4066 +t=10: Selected seed 195 with value = 0.4066 +Query 1/1: Action query time = 1.896 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4824 +t=26: Selected seed 195 with value = 0.4824 +Query 1/1: Action query time = 2.265 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5614 +t=42: Selected seed 195 with value = 0.5614 +Query 1/1: Action query time = 2.307 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5845 +t=58: Selected seed 195 with value = 0.5845 +Query 1/1: Action query time = 2.083 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7316 +t=74: Selected seed 195 with value = 0.7316 +Query 1/1: Action query time = 2.278 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8030 +t=90: Selected seed 195 with value = 0.8030 +Query 1/1: Action query time = 3.347 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8859 +t=106: Selected seed 195 with value = 0.8859 +Query 1/1: Action query time = 3.364 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.877 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=9--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 9 +# successes: 8 (88.9%) + +Task: push the plate to the front of the stove +Starting episode 10... +Query 1/1: Action query time = 1.924 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3895 +t=10: Selected seed 195 with value = 0.3895 +Query 1/1: Action query time = 2.153 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4166 +t=26: Selected seed 195 with value = 0.4166 +Query 1/1: Action query time = 2.433 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5066 +t=42: Selected seed 195 with value = 0.5066 +Query 1/1: Action query time = 2.331 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6597 +t=58: Selected seed 195 with value = 0.6597 +Query 1/1: Action query time = 1.797 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7215 +t=74: Selected seed 195 with value = 0.7215 +Query 1/1: Action query time = 1.604 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8336 +t=90: Selected seed 195 with value = 0.8336 +Query 1/1: Action query time = 2.427 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9865 +t=106: Selected seed 195 with value = 0.9865 +Query 1/1: Action query time = 2.406 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.354 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=10--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 10 +# successes: 9 (90.0%) + +Task: push the plate to the front of the stove +Starting episode 11... +Query 1/1: Action query time = 1.711 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3708 +t=10: Selected seed 195 with value = 0.3708 +Query 1/1: Action query time = 2.788 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4140 +t=26: Selected seed 195 with value = 0.4140 +Query 1/1: Action query time = 2.682 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5534 +t=42: Selected seed 195 with value = 0.5534 +Query 1/1: Action query time = 2.008 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6243 +t=58: Selected seed 195 with value = 0.6243 +Query 1/1: Action query time = 1.960 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7167 +t=74: Selected seed 195 with value = 0.7167 +Query 1/1: Action query time = 1.961 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7941 +t=90: Selected seed 195 with value = 0.7941 +Query 1/1: Action query time = 1.997 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9102 +t=106: Selected seed 195 with value = 0.9102 +Query 1/1: Action query time = 1.931 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.681 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.987 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=11--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 11 +# successes: 10 (90.9%) + +Task: push the plate to the front of the stove +Starting episode 12... +Query 1/1: Action query time = 1.341 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3697 +t=10: Selected seed 195 with value = 0.3697 +Query 1/1: Action query time = 1.339 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3104 +t=26: Selected seed 195 with value = 0.3104 +Query 1/1: Action query time = 1.297 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5154 +t=42: Selected seed 195 with value = 0.5154 +Query 1/1: Action query time = 1.242 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5421 +t=58: Selected seed 195 with value = 0.5421 +Query 1/1: Action query time = 1.275 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6824 +t=74: Selected seed 195 with value = 0.6824 +Query 1/1: Action query time = 1.264 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8030 +t=90: Selected seed 195 with value = 0.8030 +Query 1/1: Action query time = 1.238 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9579 +t=106: Selected seed 195 with value = 0.9579 +Query 1/1: Action query time = 1.258 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.259 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=12--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 12 +# successes: 11 (91.7%) + +Task: push the plate to the front of the stove +Starting episode 13... +Query 1/1: Action query time = 0.970 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3859 +t=10: Selected seed 195 with value = 0.3859 +Query 1/1: Action query time = 0.964 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4200 +t=26: Selected seed 195 with value = 0.4200 +Query 1/1: Action query time = 0.963 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5258 +t=42: Selected seed 195 with value = 0.5258 +Query 1/1: Action query time = 0.969 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6348 +t=58: Selected seed 195 with value = 0.6348 +Query 1/1: Action query time = 0.958 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7223 +t=74: Selected seed 195 with value = 0.7223 +Query 1/1: Action query time = 0.966 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8404 +t=90: Selected seed 195 with value = 0.8404 +Query 1/1: Action query time = 0.969 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9763 +t=106: Selected seed 195 with value = 0.9763 +Query 1/1: Action query time = 0.980 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.957 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t5_s0/2026_08_03-14_46_55--with_future_img--episode=13--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 13 +# successes: 12 (92.3%) +Current task success rate: 0.9230769230769231 +Current total success rate: 0.9230769230769231 +Final results: +Total episodes: 13 +Total successes: 12 +Overall success rate: 0.9231 (92.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_46_57--fc350full_t7_s2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_46_57--fc350full_t7_s2.txt new file mode 100644 index 0000000000000000000000000000000000000000..57bb1d27916b7fd5a1446592cc25ed5923f03b9e --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-14_46_57--fc350full_t7_s2.txt @@ -0,0 +1,348 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fc350full_t7_s2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,6,10,14,18,22,26,30,34,38,42,46', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 5.088 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4540 +t=10: Selected seed 195 with value = 0.4540 +Query 1/1: Action query time = 4.582 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5490 +t=26: Selected seed 195 with value = 0.5490 +Query 1/1: Action query time = 5.017 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6683 +t=42: Selected seed 195 with value = 0.6683 +Query 1/1: Action query time = 5.400 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7754 +t=58: Selected seed 195 with value = 0.7754 +Query 1/1: Action query time = 4.203 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8939 +t=74: Selected seed 195 with value = 0.8939 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 3.465 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4672 +t=10: Selected seed 195 with value = 0.4672 +Query 1/1: Action query time = 5.419 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5469 +t=26: Selected seed 195 with value = 0.5469 +Query 1/1: Action query time = 4.796 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6732 +t=42: Selected seed 195 with value = 0.6732 +Query 1/1: Action query time = 5.520 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7799 +t=58: Selected seed 195 with value = 0.7799 +Query 1/1: Action query time = 5.244 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8945 +t=74: Selected seed 195 with value = 0.8945 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 4.234 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4750 +t=10: Selected seed 195 with value = 0.4750 +Query 1/1: Action query time = 4.744 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5466 +t=26: Selected seed 195 with value = 0.5466 +Query 1/1: Action query time = 4.976 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6488 +t=42: Selected seed 195 with value = 0.6488 +Query 1/1: Action query time = 5.704 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7655 +t=58: Selected seed 195 with value = 0.7655 +Query 1/1: Action query time = 5.000 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9188 +t=74: Selected seed 195 with value = 0.9188 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 3.453 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4860 +t=10: Selected seed 195 with value = 0.4860 +Query 1/1: Action query time = 4.656 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5599 +t=26: Selected seed 195 with value = 0.5599 +Query 1/1: Action query time = 4.908 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6463 +t=42: Selected seed 195 with value = 0.6463 +Query 1/1: Action query time = 4.992 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7483 +t=58: Selected seed 195 with value = 0.7483 +Query 1/1: Action query time = 4.922 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8818 +t=74: Selected seed 195 with value = 0.8818 +Query 1/1: Action query time = 4.056 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9982 +t=90: Selected seed 195 with value = 0.9982 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 4.890 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4776 +t=10: Selected seed 195 with value = 0.4776 +Query 1/1: Action query time = 5.643 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5808 +t=26: Selected seed 195 with value = 0.5808 +Query 1/1: Action query time = 4.934 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7087 +t=42: Selected seed 195 with value = 0.7087 +Query 1/1: Action query time = 4.695 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7999 +t=58: Selected seed 195 with value = 0.7999 +Query 1/1: Action query time = 4.201 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9453 +t=74: Selected seed 195 with value = 0.9453 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 4.583 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5113 +t=10: Selected seed 195 with value = 0.5113 +Query 1/1: Action query time = 4.300 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6382 +t=26: Selected seed 195 with value = 0.6382 +Query 1/1: Action query time = 5.756 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6931 +t=42: Selected seed 195 with value = 0.6931 +Query 1/1: Action query time = 3.946 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8281 +t=58: Selected seed 195 with value = 0.8281 +Query 1/1: Action query time = 4.181 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9585 +t=74: Selected seed 195 with value = 0.9585 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: turn on the stove +Starting episode 7... +Query 1/1: Action query time = 4.957 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4779 +t=10: Selected seed 195 with value = 0.4779 +Query 1/1: Action query time = 4.221 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5533 +t=26: Selected seed 195 with value = 0.5533 +Query 1/1: Action query time = 4.521 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6531 +t=42: Selected seed 195 with value = 0.6531 +Query 1/1: Action query time = 5.417 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7962 +t=58: Selected seed 195 with value = 0.7962 +Query 1/1: Action query time = 4.876 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9061 +t=74: Selected seed 195 with value = 0.9061 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=7--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: turn on the stove +Starting episode 8... +Query 1/1: Action query time = 4.906 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4727 +t=10: Selected seed 195 with value = 0.4727 +Query 1/1: Action query time = 3.879 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5650 +t=26: Selected seed 195 with value = 0.5650 +Query 1/1: Action query time = 5.013 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6581 +t=42: Selected seed 195 with value = 0.6581 +Query 1/1: Action query time = 4.909 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7457 +t=58: Selected seed 195 with value = 0.7457 +Query 1/1: Action query time = 4.373 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8760 +t=74: Selected seed 195 with value = 0.8760 +Query 1/1: Action query time = 4.594 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=8--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: turn on the stove +Starting episode 9... +Query 1/1: Action query time = 4.030 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5033 +t=10: Selected seed 195 with value = 0.5033 +Query 1/1: Action query time = 3.866 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6156 +t=26: Selected seed 195 with value = 0.6156 +Query 1/1: Action query time = 4.923 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6826 +t=42: Selected seed 195 with value = 0.6826 +Query 1/1: Action query time = 4.874 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7906 +t=58: Selected seed 195 with value = 0.7906 +Query 1/1: Action query time = 4.214 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9219 +t=74: Selected seed 195 with value = 0.9219 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=9--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: turn on the stove +Starting episode 10... +Query 1/1: Action query time = 4.092 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5214 +t=10: Selected seed 195 with value = 0.5214 +Query 1/1: Action query time = 4.915 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5914 +t=26: Selected seed 195 with value = 0.5914 +Query 1/1: Action query time = 4.733 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7089 +t=42: Selected seed 195 with value = 0.7089 +Query 1/1: Action query time = 5.122 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8405 +t=58: Selected seed 195 with value = 0.8405 +Query 1/1: Action query time = 4.631 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9807 +t=74: Selected seed 195 with value = 0.9807 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=10--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: turn on the stove +Starting episode 11... +Query 1/1: Action query time = 4.701 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5136 +t=10: Selected seed 195 with value = 0.5136 +Query 1/1: Action query time = 4.015 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6086 +t=26: Selected seed 195 with value = 0.6086 +Query 1/1: Action query time = 4.845 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7100 +t=42: Selected seed 195 with value = 0.7100 +Query 1/1: Action query time = 5.108 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8171 +t=58: Selected seed 195 with value = 0.8171 +Query 1/1: Action query time = 3.846 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9474 +t=74: Selected seed 195 with value = 0.9474 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=11--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: turn on the stove +Starting episode 12... +Query 1/1: Action query time = 4.367 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4631 +t=10: Selected seed 195 with value = 0.4631 +Query 1/1: Action query time = 3.537 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5482 +t=26: Selected seed 195 with value = 0.5482 +Query 1/1: Action query time = 4.597 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6454 +t=42: Selected seed 195 with value = 0.6454 +Query 1/1: Action query time = 3.765 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7546 +t=58: Selected seed 195 with value = 0.7546 +Query 1/1: Action query time = 4.700 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9255 +t=74: Selected seed 195 with value = 0.9255 +Query 1/1: Action query time = 4.126 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc350full_t7_s2/2026_08_03-14_46_57--with_future_img--episode=12--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 12 +Total successes: 12 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_06_02--fc175full_t1_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_06_02--fc175full_t1_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..d33e578dc905bd88a45e2a657d2830cb7139d228 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_06_02--fc175full_t1_s0.txt @@ -0,0 +1,419 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fc175full_t1_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,4,8,12,16,20,24,28,32,36,40,44,48', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 3.632 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5373 +t=10: Selected seed 195 with value = 0.5373 +Query 1/1: Action query time = 4.757 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7033 +t=26: Selected seed 195 with value = 0.7033 +Query 1/1: Action query time = 4.836 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8012 +t=42: Selected seed 195 with value = 0.8012 +Query 1/1: Action query time = 5.534 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8860 +t=58: Selected seed 195 with value = 0.8860 +Query 1/1: Action query time = 5.266 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=74: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 4.104 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 5.055 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5947 +t=10: Selected seed 195 with value = 0.5947 +Query 1/1: Action query time = 4.846 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6932 +t=26: Selected seed 195 with value = 0.6932 +Query 1/1: Action query time = 3.745 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7794 +t=42: Selected seed 195 with value = 0.7794 +Query 1/1: Action query time = 4.659 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9020 +t=58: Selected seed 195 with value = 0.9020 +Query 1/1: Action query time = 4.917 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.992 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 4.435 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5798 +t=10: Selected seed 195 with value = 0.5798 +Query 1/1: Action query time = 4.211 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6763 +t=26: Selected seed 195 with value = 0.6763 +Query 1/1: Action query time = 4.750 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7880 +t=42: Selected seed 195 with value = 0.7880 +Query 1/1: Action query time = 5.688 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8969 +t=58: Selected seed 195 with value = 0.8969 +Query 1/1: Action query time = 4.203 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.598 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 5.159 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5966 +t=10: Selected seed 195 with value = 0.5966 +Query 1/1: Action query time = 5.018 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8278 +t=26: Selected seed 195 with value = 0.8278 +Query 1/1: Action query time = 4.644 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8251 +t=42: Selected seed 195 with value = 0.8251 +Query 1/1: Action query time = 3.994 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9014 +t=58: Selected seed 195 with value = 0.9014 +Query 1/1: Action query time = 4.467 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9936 +t=74: Selected seed 195 with value = 0.9936 +Query 1/1: Action query time = 5.265 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=4--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 3.120 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5725 +t=10: Selected seed 195 with value = 0.5725 +Query 1/1: Action query time = 5.044 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7160 +t=26: Selected seed 195 with value = 0.7160 +Query 1/1: Action query time = 5.400 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8256 +t=42: Selected seed 195 with value = 0.8256 +Query 1/1: Action query time = 5.346 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9377 +t=58: Selected seed 195 with value = 0.9377 +Query 1/1: Action query time = 4.642 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.440 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=5--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 4.822 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5686 +t=10: Selected seed 195 with value = 0.5686 +Query 1/1: Action query time = 5.216 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7355 +t=26: Selected seed 195 with value = 0.7355 +Query 1/1: Action query time = 5.184 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8038 +t=42: Selected seed 195 with value = 0.8038 +Query 1/1: Action query time = 5.169 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8999 +t=58: Selected seed 195 with value = 0.8999 +Query 1/1: Action query time = 3.122 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9996 +t=74: Selected seed 195 with value = 0.9996 +Query 1/1: Action query time = 5.793 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=6--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 4.195 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6012 +t=10: Selected seed 195 with value = 0.6012 +Query 1/1: Action query time = 4.829 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7250 +t=26: Selected seed 195 with value = 0.7250 +Query 1/1: Action query time = 4.181 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7652 +t=42: Selected seed 195 with value = 0.7652 +Query 1/1: Action query time = 5.482 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8665 +t=58: Selected seed 195 with value = 0.8665 +Query 1/1: Action query time = 5.174 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9463 +t=74: Selected seed 195 with value = 0.9463 +Query 1/1: Action query time = 2.779 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=7--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 5.441 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6019 +t=10: Selected seed 195 with value = 0.6019 +Query 1/1: Action query time = 4.612 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8464 +t=26: Selected seed 195 with value = 0.8464 +Query 1/1: Action query time = 5.040 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8402 +t=42: Selected seed 195 with value = 0.8402 +Query 1/1: Action query time = 4.253 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9049 +t=58: Selected seed 195 with value = 0.9049 +Query 1/1: Action query time = 3.470 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.585 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=8--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 4.812 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6086 +t=10: Selected seed 195 with value = 0.6086 +Query 1/1: Action query time = 4.438 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6947 +t=26: Selected seed 195 with value = 0.6947 +Query 1/1: Action query time = 5.087 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8059 +t=42: Selected seed 195 with value = 0.8059 +Query 1/1: Action query time = 4.551 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9071 +t=58: Selected seed 195 with value = 0.9071 +Query 1/1: Action query time = 4.449 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9978 +t=74: Selected seed 195 with value = 0.9978 +Query 1/1: Action query time = 4.757 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=9--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 5.246 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5567 +t=10: Selected seed 195 with value = 0.5567 +Query 1/1: Action query time = 4.463 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7115 +t=26: Selected seed 195 with value = 0.7115 +Query 1/1: Action query time = 4.563 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8384 +t=42: Selected seed 195 with value = 0.8384 +Query 1/1: Action query time = 4.448 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9599 +t=58: Selected seed 195 with value = 0.9599 +Query 1/1: Action query time = 4.722 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.028 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=10--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 4.097 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6693 +t=10: Selected seed 195 with value = 0.6693 +Query 1/1: Action query time = 4.138 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7736 +t=26: Selected seed 195 with value = 0.7736 +Query 1/1: Action query time = 4.603 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7633 +t=42: Selected seed 195 with value = 0.7633 +Query 1/1: Action query time = 4.275 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8429 +t=58: Selected seed 195 with value = 0.8429 +Query 1/1: Action query time = 4.966 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9795 +t=74: Selected seed 195 with value = 0.9795 +Query 1/1: Action query time = 4.959 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.197 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=11--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 5.433 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5617 +t=10: Selected seed 195 with value = 0.5617 +Query 1/1: Action query time = 5.465 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7646 +t=26: Selected seed 195 with value = 0.7646 +Query 1/1: Action query time = 3.959 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7865 +t=42: Selected seed 195 with value = 0.7865 +Query 1/1: Action query time = 4.332 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8614 +t=58: Selected seed 195 with value = 0.8614 +Query 1/1: Action query time = 4.276 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.069 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=12--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 5.224 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5540 +t=10: Selected seed 195 with value = 0.5540 +Query 1/1: Action query time = 4.016 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6709 +t=26: Selected seed 195 with value = 0.6709 +Query 1/1: Action query time = 4.747 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7865 +t=42: Selected seed 195 with value = 0.7865 +Query 1/1: Action query time = 4.905 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8847 +t=58: Selected seed 195 with value = 0.8847 +Query 1/1: Action query time = 4.830 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.288 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t1_s0/2026_08_03-15_06_02--with_future_img--episode=13--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 13 +Total successes: 13 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_06_03--fc175full_t3_s2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_06_03--fc175full_t3_s2.txt new file mode 100644 index 0000000000000000000000000000000000000000..654697a69be4ee096cdca38f4443de229d319906 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_06_03--fc175full_t3_s2.txt @@ -0,0 +1,796 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fc175full_t3_s2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,6,10,14,18,22,26,30,34,38,42,46', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 5.654 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2299 +t=10: Selected seed 195 with value = 0.2299 +Query 1/1: Action query time = 4.740 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2671 +t=26: Selected seed 195 with value = 0.2671 +Query 1/1: Action query time = 4.095 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3071 +t=42: Selected seed 195 with value = 0.3071 +Query 1/1: Action query time = 4.982 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3457 +t=58: Selected seed 195 with value = 0.3457 +Query 1/1: Action query time = 5.232 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4093 +t=74: Selected seed 195 with value = 0.4093 +Query 1/1: Action query time = 4.479 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4757 +t=90: Selected seed 195 with value = 0.4757 +Query 1/1: Action query time = 3.656 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5823 +t=106: Selected seed 195 with value = 0.5823 +Query 1/1: Action query time = 4.345 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6406 +t=122: Selected seed 195 with value = 0.6406 +Query 1/1: Action query time = 5.916 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7072 +t=138: Selected seed 195 with value = 0.7072 +Query 1/1: Action query time = 5.189 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8697 +t=154: Selected seed 195 with value = 0.8697 +Query 1/1: Action query time = 5.371 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9861 +t=170: Selected seed 195 with value = 0.9861 +Query 1/1: Action query time = 3.944 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 3.761 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2170 +t=10: Selected seed 195 with value = 0.2170 +Query 1/1: Action query time = 4.777 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2496 +t=26: Selected seed 195 with value = 0.2496 +Query 1/1: Action query time = 4.675 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3150 +t=42: Selected seed 195 with value = 0.3150 +Query 1/1: Action query time = 4.991 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3452 +t=58: Selected seed 195 with value = 0.3452 +Query 1/1: Action query time = 4.304 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4503 +t=74: Selected seed 195 with value = 0.4503 +Query 1/1: Action query time = 4.298 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4892 +t=90: Selected seed 195 with value = 0.4892 +Query 1/1: Action query time = 4.376 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5781 +t=106: Selected seed 195 with value = 0.5781 +Query 1/1: Action query time = 4.760 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6465 +t=122: Selected seed 195 with value = 0.6465 +Query 1/1: Action query time = 4.930 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7517 +t=138: Selected seed 195 with value = 0.7517 +Query 1/1: Action query time = 4.936 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8947 +t=154: Selected seed 195 with value = 0.8947 +Query 1/1: Action query time = 4.621 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=170: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 4.423 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9638 +t=186: Selected seed 195 with value = 0.9638 +Query 1/1: Action query time = 4.929 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9703 +t=202: Selected seed 195 with value = 0.9703 +Query 1/1: Action query time = 4.179 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9759 +t=218: Selected seed 195 with value = 0.9759 +Query 1/1: Action query time = 5.159 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3361 +t=234: Selected seed 195 with value = 0.3361 +Query 1/1: Action query time = 5.535 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3492 +t=250: Selected seed 195 with value = 0.3492 +Query 1/1: Action query time = 5.528 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9606 +t=266: Selected seed 195 with value = 0.9606 +Query 1/1: Action query time = 4.267 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9322 +t=282: Selected seed 195 with value = 0.9322 +Query 1/1: Action query time = 4.835 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9705 +t=298: Selected seed 195 with value = 0.9705 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=2--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.120 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2277 +t=10: Selected seed 195 with value = 0.2277 +Query 1/1: Action query time = 4.978 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2550 +t=26: Selected seed 195 with value = 0.2550 +Query 1/1: Action query time = 4.788 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3206 +t=42: Selected seed 195 with value = 0.3206 +Query 1/1: Action query time = 3.282 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3377 +t=58: Selected seed 195 with value = 0.3377 +Query 1/1: Action query time = 3.873 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4396 +t=74: Selected seed 195 with value = 0.4396 +Query 1/1: Action query time = 5.252 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5114 +t=90: Selected seed 195 with value = 0.5114 +Query 1/1: Action query time = 4.517 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5966 +t=106: Selected seed 195 with value = 0.5966 +Query 1/1: Action query time = 3.445 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7067 +t=122: Selected seed 195 with value = 0.7067 +Query 1/1: Action query time = 4.795 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7896 +t=138: Selected seed 195 with value = 0.7896 +Query 1/1: Action query time = 5.212 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9466 +t=154: Selected seed 195 with value = 0.9466 +Query 1/1: Action query time = 4.424 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9915 +t=170: Selected seed 195 with value = 0.9915 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 5.378 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2171 +t=10: Selected seed 195 with value = 0.2171 +Query 1/1: Action query time = 4.933 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2542 +t=26: Selected seed 195 with value = 0.2542 +Query 1/1: Action query time = 4.663 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3259 +t=42: Selected seed 195 with value = 0.3259 +Query 1/1: Action query time = 4.391 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3738 +t=58: Selected seed 195 with value = 0.3738 +Query 1/1: Action query time = 4.549 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4395 +t=74: Selected seed 195 with value = 0.4395 +Query 1/1: Action query time = 4.274 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5011 +t=90: Selected seed 195 with value = 0.5011 +Query 1/1: Action query time = 5.271 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6140 +t=106: Selected seed 195 with value = 0.6140 +Query 1/1: Action query time = 4.352 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7651 +t=122: Selected seed 195 with value = 0.7651 +Query 1/1: Action query time = 5.082 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8620 +t=138: Selected seed 195 with value = 0.8620 +Query 1/1: Action query time = 4.848 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=154: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 5.068 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9721 +t=170: Selected seed 195 with value = 0.9721 +Query 1/1: Action query time = 5.219 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9680 +t=186: Selected seed 195 with value = 0.9680 +Query 1/1: Action query time = 4.576 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4734 +t=202: Selected seed 195 with value = 0.4734 +Query 1/1: Action query time = 4.281 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3550 +t=218: Selected seed 195 with value = 0.3550 +Query 1/1: Action query time = 3.793 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9671 +t=234: Selected seed 195 with value = 0.9671 +Query 1/1: Action query time = 4.868 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9012 +t=250: Selected seed 195 with value = 0.9012 +Query 1/1: Action query time = 4.509 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3848 +t=266: Selected seed 195 with value = 0.3848 +Query 1/1: Action query time = 3.637 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9509 +t=282: Selected seed 195 with value = 0.9509 +Query 1/1: Action query time = 4.526 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9198 +t=298: Selected seed 195 with value = 0.9198 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=4--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 4 +# successes: 2 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 5... +Query 1/1: Action query time = 5.085 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2209 +t=10: Selected seed 195 with value = 0.2209 +Query 1/1: Action query time = 3.317 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2679 +t=26: Selected seed 195 with value = 0.2679 +Query 1/1: Action query time = 3.593 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3018 +t=42: Selected seed 195 with value = 0.3018 +Query 1/1: Action query time = 5.432 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3347 +t=58: Selected seed 195 with value = 0.3347 +Query 1/1: Action query time = 4.798 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4050 +t=74: Selected seed 195 with value = 0.4050 +Query 1/1: Action query time = 4.176 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4821 +t=90: Selected seed 195 with value = 0.4821 +Query 1/1: Action query time = 4.656 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5669 +t=106: Selected seed 195 with value = 0.5669 +Query 1/1: Action query time = 4.966 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6897 +t=122: Selected seed 195 with value = 0.6897 +Query 1/1: Action query time = 4.946 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7757 +t=138: Selected seed 195 with value = 0.7757 +Query 1/1: Action query time = 4.212 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8997 +t=154: Selected seed 195 with value = 0.8997 +Query 1/1: Action query time = 4.879 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.845 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=5--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 5 +# successes: 3 (60.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 6... +Query 1/1: Action query time = 4.318 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2284 +t=10: Selected seed 195 with value = 0.2284 +Query 1/1: Action query time = 5.545 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2559 +t=26: Selected seed 195 with value = 0.2559 +Query 1/1: Action query time = 4.102 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3202 +t=42: Selected seed 195 with value = 0.3202 +Query 1/1: Action query time = 3.943 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3653 +t=58: Selected seed 195 with value = 0.3653 +Query 1/1: Action query time = 4.627 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4394 +t=74: Selected seed 195 with value = 0.4394 +Query 1/1: Action query time = 5.973 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4958 +t=90: Selected seed 195 with value = 0.4958 +Query 1/1: Action query time = 4.745 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5957 +t=106: Selected seed 195 with value = 0.5957 +Query 1/1: Action query time = 5.577 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7786 +t=122: Selected seed 195 with value = 0.7786 +Query 1/1: Action query time = 5.380 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8169 +t=138: Selected seed 195 with value = 0.8169 +Query 1/1: Action query time = 5.395 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9536 +t=154: Selected seed 195 with value = 0.9536 +Query 1/1: Action query time = 4.158 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=170: Selected seed 195 with value = 0.9928 +Query 1/1: Action query time = 2.303 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=6--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 6 +# successes: 4 (66.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 7... +Query 1/1: Action query time = 4.280 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2147 +t=10: Selected seed 195 with value = 0.2147 +Query 1/1: Action query time = 5.159 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2451 +t=26: Selected seed 195 with value = 0.2451 +Query 1/1: Action query time = 5.311 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3121 +t=42: Selected seed 195 with value = 0.3121 +Query 1/1: Action query time = 5.045 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3317 +t=58: Selected seed 195 with value = 0.3317 +Query 1/1: Action query time = 3.747 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4098 +t=74: Selected seed 195 with value = 0.4098 +Query 1/1: Action query time = 3.234 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5015 +t=90: Selected seed 195 with value = 0.5015 +Query 1/1: Action query time = 4.187 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5803 +t=106: Selected seed 195 with value = 0.5803 +Query 1/1: Action query time = 3.040 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6934 +t=122: Selected seed 195 with value = 0.6934 +Query 1/1: Action query time = 3.216 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7766 +t=138: Selected seed 195 with value = 0.7766 +Query 1/1: Action query time = 4.215 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8947 +t=154: Selected seed 195 with value = 0.8947 +Query 1/1: Action query time = 4.249 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9274 +t=170: Selected seed 195 with value = 0.9274 +Query 1/1: Action query time = 4.070 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9802 +t=186: Selected seed 195 with value = 0.9802 +Query 1/1: Action query time = 4.286 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9715 +t=202: Selected seed 195 with value = 0.9715 +Query 1/1: Action query time = 3.407 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9782 +t=218: Selected seed 195 with value = 0.9782 +Query 1/1: Action query time = 2.197 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9689 +t=234: Selected seed 195 with value = 0.9689 +Query 1/1: Action query time = 2.488 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9710 +t=250: Selected seed 195 with value = 0.9710 +Query 1/1: Action query time = 2.619 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9718 +t=266: Selected seed 195 with value = 0.9718 +Query 1/1: Action query time = 2.946 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9875 +t=282: Selected seed 195 with value = 0.9875 +Query 1/1: Action query time = 3.651 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5085 +t=298: Selected seed 195 with value = 0.5085 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=7--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 7 +# successes: 4 (57.1%) + +Task: open the top drawer and put the bowl inside +Starting episode 8... +Query 1/1: Action query time = 2.125 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2213 +t=10: Selected seed 195 with value = 0.2213 +Query 1/1: Action query time = 1.562 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2545 +t=26: Selected seed 195 with value = 0.2545 +Query 1/1: Action query time = 2.311 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3249 +t=42: Selected seed 195 with value = 0.3249 +Query 1/1: Action query time = 2.083 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3699 +t=58: Selected seed 195 with value = 0.3699 +Query 1/1: Action query time = 2.295 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4419 +t=74: Selected seed 195 with value = 0.4419 +Query 1/1: Action query time = 2.326 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5181 +t=90: Selected seed 195 with value = 0.5181 +Query 1/1: Action query time = 1.978 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6074 +t=106: Selected seed 195 with value = 0.6074 +Query 1/1: Action query time = 1.863 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6747 +t=122: Selected seed 195 with value = 0.6747 +Query 1/1: Action query time = 1.848 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7718 +t=138: Selected seed 195 with value = 0.7718 +Query 1/1: Action query time = 1.881 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9396 +t=154: Selected seed 195 with value = 0.9396 +Query 1/1: Action query time = 2.345 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=8--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 8 +# successes: 5 (62.5%) + +Task: open the top drawer and put the bowl inside +Starting episode 9... +Query 1/1: Action query time = 2.703 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2164 +t=10: Selected seed 195 with value = 0.2164 +Query 1/1: Action query time = 2.039 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2547 +t=26: Selected seed 195 with value = 0.2547 +Query 1/1: Action query time = 2.025 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3217 +t=42: Selected seed 195 with value = 0.3217 +Query 1/1: Action query time = 2.729 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3668 +t=58: Selected seed 195 with value = 0.3668 +Query 1/1: Action query time = 1.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4513 +t=74: Selected seed 195 with value = 0.4513 +Query 1/1: Action query time = 1.690 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5425 +t=90: Selected seed 195 with value = 0.5425 +Query 1/1: Action query time = 2.406 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6020 +t=106: Selected seed 195 with value = 0.6020 +Query 1/1: Action query time = 2.128 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8154 +t=122: Selected seed 195 with value = 0.8154 +Query 1/1: Action query time = 2.369 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8096 +t=138: Selected seed 195 with value = 0.8096 +Query 1/1: Action query time = 1.949 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9608 +t=154: Selected seed 195 with value = 0.9608 +Query 1/1: Action query time = 1.674 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=9--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 9 +# successes: 6 (66.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 10... +Query 1/1: Action query time = 3.045 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2264 +t=10: Selected seed 195 with value = 0.2264 +Query 1/1: Action query time = 2.887 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2505 +t=26: Selected seed 195 with value = 0.2505 +Query 1/1: Action query time = 2.687 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3058 +t=42: Selected seed 195 with value = 0.3058 +Query 1/1: Action query time = 1.451 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3221 +t=58: Selected seed 195 with value = 0.3221 +Query 1/1: Action query time = 1.583 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4231 +t=74: Selected seed 195 with value = 0.4231 +Query 1/1: Action query time = 2.211 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4826 +t=90: Selected seed 195 with value = 0.4826 +Query 1/1: Action query time = 1.935 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5879 +t=106: Selected seed 195 with value = 0.5879 +Query 1/1: Action query time = 1.948 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6390 +t=122: Selected seed 195 with value = 0.6390 +Query 1/1: Action query time = 2.534 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6677 +t=138: Selected seed 195 with value = 0.6677 +Query 1/1: Action query time = 1.812 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8371 +t=154: Selected seed 195 with value = 0.8371 +Query 1/1: Action query time = 1.905 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9501 +t=170: Selected seed 195 with value = 0.9501 +Query 1/1: Action query time = 2.640 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9819 +t=186: Selected seed 195 with value = 0.9819 +Query 1/1: Action query time = 2.591 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9664 +t=202: Selected seed 195 with value = 0.9664 +Query 1/1: Action query time = 2.464 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9783 +t=218: Selected seed 195 with value = 0.9783 +Query 1/1: Action query time = 2.348 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9749 +t=234: Selected seed 195 with value = 0.9749 +Query 1/1: Action query time = 2.310 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9733 +t=250: Selected seed 195 with value = 0.9733 +Query 1/1: Action query time = 2.417 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9635 +t=266: Selected seed 195 with value = 0.9635 +Query 1/1: Action query time = 1.953 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9640 +t=282: Selected seed 195 with value = 0.9640 +Query 1/1: Action query time = 1.982 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9643 +t=298: Selected seed 195 with value = 0.9643 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=10--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 10 +# successes: 6 (60.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 11... +Query 1/1: Action query time = 2.105 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2297 +t=10: Selected seed 195 with value = 0.2297 +Query 1/1: Action query time = 2.168 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2611 +t=26: Selected seed 195 with value = 0.2611 +Query 1/1: Action query time = 2.341 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3171 +t=42: Selected seed 195 with value = 0.3171 +Query 1/1: Action query time = 2.211 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3685 +t=58: Selected seed 195 with value = 0.3685 +Query 1/1: Action query time = 1.917 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4354 +t=74: Selected seed 195 with value = 0.4354 +Query 1/1: Action query time = 1.941 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5034 +t=90: Selected seed 195 with value = 0.5034 +Query 1/1: Action query time = 1.710 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6037 +t=106: Selected seed 195 with value = 0.6037 +Query 1/1: Action query time = 0.982 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7380 +t=122: Selected seed 195 with value = 0.7380 +Query 1/1: Action query time = 0.965 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8609 +t=138: Selected seed 195 with value = 0.8609 +Query 1/1: Action query time = 0.981 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.055 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=11--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 11 +# successes: 7 (63.6%) + +Task: open the top drawer and put the bowl inside +Starting episode 12... +Query 1/1: Action query time = 1.799 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2237 +t=10: Selected seed 195 with value = 0.2237 +Query 1/1: Action query time = 1.787 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2478 +t=26: Selected seed 195 with value = 0.2478 +Query 1/1: Action query time = 1.748 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3117 +t=42: Selected seed 195 with value = 0.3117 +Query 1/1: Action query time = 1.757 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3478 +t=58: Selected seed 195 with value = 0.3478 +Query 1/1: Action query time = 1.758 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4175 +t=74: Selected seed 195 with value = 0.4175 +Query 1/1: Action query time = 1.789 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4871 +t=90: Selected seed 195 with value = 0.4871 +Query 1/1: Action query time = 1.912 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5829 +t=106: Selected seed 195 with value = 0.5829 +Query 1/1: Action query time = 1.844 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6600 +t=122: Selected seed 195 with value = 0.6600 +Query 1/1: Action query time = 0.961 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7818 +t=138: Selected seed 195 with value = 0.7818 +Query 1/1: Action query time = 0.964 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8944 +t=154: Selected seed 195 with value = 0.8944 +Query 1/1: Action query time = 0.974 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=170: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 1.006 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9632 +t=186: Selected seed 195 with value = 0.9632 +Query 1/1: Action query time = 1.868 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9714 +t=202: Selected seed 195 with value = 0.9714 +Query 1/1: Action query time = 1.869 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9626 +t=218: Selected seed 195 with value = 0.9626 +Query 1/1: Action query time = 1.836 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9764 +t=234: Selected seed 195 with value = 0.9764 +Query 1/1: Action query time = 1.794 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3323 +t=250: Selected seed 195 with value = 0.3323 +Query 1/1: Action query time = 1.759 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3480 +t=266: Selected seed 195 with value = 0.3480 +Query 1/1: Action query time = 1.667 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9557 +t=282: Selected seed 195 with value = 0.9557 +Query 1/1: Action query time = 1.584 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9480 +t=298: Selected seed 195 with value = 0.9480 +Saved rollout MP4 at path ./rollouts/fc175full_t3_s2/2026_08_03-15_06_03--with_future_img--episode=12--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 12 +# successes: 7 (58.3%) +Current task success rate: 0.5833333333333334 +Current total success rate: 0.5833333333333334 +Final results: +Total episodes: 12 +Total successes: 7 +Overall success rate: 0.5833 (58.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_19_31--fc175full_t5_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_19_31--fc175full_t5_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..d2bfa269e84bacbe853ae5a9625f6c7be2087d09 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_19_31--fc175full_t5_s3.txt @@ -0,0 +1,584 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fc175full_t5_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,7,11,15,19,23,27,31,35,39,43,47', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 2.661 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3793 +t=10: Selected seed 195 with value = 0.3793 +Query 1/1: Action query time = 4.140 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4187 +t=26: Selected seed 195 with value = 0.4187 +Query 1/1: Action query time = 4.492 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5649 +t=42: Selected seed 195 with value = 0.5649 +Query 1/1: Action query time = 5.162 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6459 +t=58: Selected seed 195 with value = 0.6459 +Query 1/1: Action query time = 5.066 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7184 +t=74: Selected seed 195 with value = 0.7184 +Query 1/1: Action query time = 4.215 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8499 +t=90: Selected seed 195 with value = 0.8499 +Query 1/1: Action query time = 4.726 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9849 +t=106: Selected seed 195 with value = 0.9849 +Query 1/1: Action query time = 4.758 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.944 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.290 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3958 +t=10: Selected seed 195 with value = 0.3958 +Query 1/1: Action query time = 4.442 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4406 +t=26: Selected seed 195 with value = 0.4406 +Query 1/1: Action query time = 3.401 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5457 +t=42: Selected seed 195 with value = 0.5457 +Query 1/1: Action query time = 4.957 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5661 +t=58: Selected seed 195 with value = 0.5661 +Query 1/1: Action query time = 4.501 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6766 +t=74: Selected seed 195 with value = 0.6766 +Query 1/1: Action query time = 4.579 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7574 +t=90: Selected seed 195 with value = 0.7574 +Query 1/1: Action query time = 4.672 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8044 +t=106: Selected seed 195 with value = 0.8044 +Query 1/1: Action query time = 5.082 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8465 +t=122: Selected seed 195 with value = 0.8465 +Query 1/1: Action query time = 5.233 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9573 +t=138: Selected seed 195 with value = 0.9573 +Query 1/1: Action query time = 4.686 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 5.318 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4129 +t=10: Selected seed 195 with value = 0.4129 +Query 1/1: Action query time = 3.992 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4600 +t=26: Selected seed 195 with value = 0.4600 +Query 1/1: Action query time = 4.717 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5721 +t=42: Selected seed 195 with value = 0.5721 +Query 1/1: Action query time = 5.080 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6739 +t=58: Selected seed 195 with value = 0.6739 +Query 1/1: Action query time = 4.050 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7483 +t=74: Selected seed 195 with value = 0.7483 +Query 1/1: Action query time = 4.604 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8559 +t=90: Selected seed 195 with value = 0.8559 +Query 1/1: Action query time = 4.056 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=106: Selected seed 195 with value = 0.9965 +Query 1/1: Action query time = 4.655 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.454 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 4.731 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4199 +t=10: Selected seed 195 with value = 0.4199 +Query 1/1: Action query time = 4.869 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4871 +t=26: Selected seed 195 with value = 0.4871 +Query 1/1: Action query time = 5.103 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5759 +t=42: Selected seed 195 with value = 0.5759 +Query 1/1: Action query time = 3.617 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6026 +t=58: Selected seed 195 with value = 0.6026 +Query 1/1: Action query time = 4.396 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7327 +t=74: Selected seed 195 with value = 0.7327 +Query 1/1: Action query time = 3.674 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8409 +t=90: Selected seed 195 with value = 0.8409 +Query 1/1: Action query time = 5.117 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9422 +t=106: Selected seed 195 with value = 0.9422 +Query 1/1: Action query time = 5.426 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.980 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=4--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 5... +Query 1/1: Action query time = 4.057 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3949 +t=10: Selected seed 195 with value = 0.3949 +Query 1/1: Action query time = 4.000 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4449 +t=26: Selected seed 195 with value = 0.4449 +Query 1/1: Action query time = 5.394 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5332 +t=42: Selected seed 195 with value = 0.5332 +Query 1/1: Action query time = 4.675 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5893 +t=58: Selected seed 195 with value = 0.5893 +Query 1/1: Action query time = 4.250 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6811 +t=74: Selected seed 195 with value = 0.6811 +Query 1/1: Action query time = 5.023 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7942 +t=90: Selected seed 195 with value = 0.7942 +Query 1/1: Action query time = 3.967 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8949 +t=106: Selected seed 195 with value = 0.8949 +Query 1/1: Action query time = 5.351 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.740 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=5--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 6... +Query 1/1: Action query time = 3.805 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4008 +t=10: Selected seed 195 with value = 0.4008 +Query 1/1: Action query time = 4.932 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4866 +t=26: Selected seed 195 with value = 0.4866 +Query 1/1: Action query time = 3.631 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5598 +t=42: Selected seed 195 with value = 0.5598 +Query 1/1: Action query time = 4.833 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6388 +t=58: Selected seed 195 with value = 0.6388 +Query 1/1: Action query time = 4.661 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7429 +t=74: Selected seed 195 with value = 0.7429 +Query 1/1: Action query time = 4.862 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8637 +t=90: Selected seed 195 with value = 0.8637 +Query 1/1: Action query time = 4.831 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.284 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.202 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=6--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 7... +Query 1/1: Action query time = 4.853 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3561 +t=10: Selected seed 195 with value = 0.3561 +Query 1/1: Action query time = 3.648 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4205 +t=26: Selected seed 195 with value = 0.4205 +Query 1/1: Action query time = 5.151 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5028 +t=42: Selected seed 195 with value = 0.5028 +Query 1/1: Action query time = 4.784 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6066 +t=58: Selected seed 195 with value = 0.6066 +Query 1/1: Action query time = 3.533 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6779 +t=74: Selected seed 195 with value = 0.6779 +Query 1/1: Action query time = 5.552 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7328 +t=90: Selected seed 195 with value = 0.7328 +Query 1/1: Action query time = 4.523 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7993 +t=106: Selected seed 195 with value = 0.7993 +Query 1/1: Action query time = 5.538 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9372 +t=122: Selected seed 195 with value = 0.9372 +Query 1/1: Action query time = 4.475 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.955 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=7--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 8... +Query 1/1: Action query time = 3.906 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3883 +t=10: Selected seed 195 with value = 0.3883 +Query 1/1: Action query time = 4.557 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4154 +t=26: Selected seed 195 with value = 0.4154 +Query 1/1: Action query time = 4.170 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5278 +t=42: Selected seed 195 with value = 0.5278 +Query 1/1: Action query time = 5.054 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6077 +t=58: Selected seed 195 with value = 0.6077 +Query 1/1: Action query time = 5.600 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6417 +t=74: Selected seed 195 with value = 0.6417 +Query 1/1: Action query time = 5.801 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7180 +t=90: Selected seed 195 with value = 0.7180 +Query 1/1: Action query time = 3.864 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8450 +t=106: Selected seed 195 with value = 0.8450 +Query 1/1: Action query time = 4.760 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9638 +t=122: Selected seed 195 with value = 0.9638 +Query 1/1: Action query time = 4.518 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=138: Selected seed 195 with value = 0.9903 +Query 1/1: Action query time = 4.159 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7651 +t=154: Selected seed 195 with value = 0.7651 +Query 1/1: Action query time = 4.732 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8308 +t=170: Selected seed 195 with value = 0.8308 +Query 1/1: Action query time = 3.153 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9683 +t=186: Selected seed 195 with value = 0.9683 +Query 1/1: Action query time = 3.775 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9575 +t=202: Selected seed 195 with value = 0.9575 +Query 1/1: Action query time = 3.436 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9355 +t=218: Selected seed 195 with value = 0.9355 +Query 1/1: Action query time = 3.099 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9282 +t=234: Selected seed 195 with value = 0.9282 +Query 1/1: Action query time = 3.520 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9411 +t=250: Selected seed 195 with value = 0.9411 +Query 1/1: Action query time = 3.178 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8578 +t=266: Selected seed 195 with value = 0.8578 +Query 1/1: Action query time = 3.126 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8659 +t=282: Selected seed 195 with value = 0.8659 +Query 1/1: Action query time = 3.164 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8375 +t=298: Selected seed 195 with value = 0.8375 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=8--success=False--task=push_the_plate_to_the_front_of_the_.mp4 +Success: False +# episodes completed so far: 8 +# successes: 7 (87.5%) + +Task: push the plate to the front of the stove +Starting episode 9... +Query 1/1: Action query time = 1.914 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3752 +t=10: Selected seed 195 with value = 0.3752 +Query 1/1: Action query time = 2.127 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4337 +t=26: Selected seed 195 with value = 0.4337 +Query 1/1: Action query time = 3.237 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5214 +t=42: Selected seed 195 with value = 0.5214 +Query 1/1: Action query time = 3.193 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5706 +t=58: Selected seed 195 with value = 0.5706 +Query 1/1: Action query time = 3.037 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6646 +t=74: Selected seed 195 with value = 0.6646 +Query 1/1: Action query time = 2.728 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7356 +t=90: Selected seed 195 with value = 0.7356 +Query 1/1: Action query time = 2.908 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8453 +t=106: Selected seed 195 with value = 0.8453 +Query 1/1: Action query time = 2.948 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9724 +t=122: Selected seed 195 with value = 0.9724 +Query 1/1: Action query time = 2.789 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.466 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=9--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 9 +# successes: 8 (88.9%) + +Task: push the plate to the front of the stove +Starting episode 10... +Query 1/1: Action query time = 1.845 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4055 +t=10: Selected seed 195 with value = 0.4055 +Query 1/1: Action query time = 3.642 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4383 +t=26: Selected seed 195 with value = 0.4383 +Query 1/1: Action query time = 3.268 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5490 +t=42: Selected seed 195 with value = 0.5490 +Query 1/1: Action query time = 3.077 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5582 +t=58: Selected seed 195 with value = 0.5582 +Query 1/1: Action query time = 3.079 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6699 +t=74: Selected seed 195 with value = 0.6699 +Query 1/1: Action query time = 2.968 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8082 +t=90: Selected seed 195 with value = 0.8082 +Query 1/1: Action query time = 2.639 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9539 +t=106: Selected seed 195 with value = 0.9539 +Query 1/1: Action query time = 2.487 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=122: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 2.365 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=10--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 10 +# successes: 9 (90.0%) + +Task: push the plate to the front of the stove +Starting episode 11... +Query 1/1: Action query time = 0.987 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3987 +t=10: Selected seed 195 with value = 0.3987 +Query 1/1: Action query time = 0.985 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4726 +t=26: Selected seed 195 with value = 0.4726 +Query 1/1: Action query time = 1.927 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5769 +t=42: Selected seed 195 with value = 0.5769 +Query 1/1: Action query time = 2.384 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5677 +t=58: Selected seed 195 with value = 0.5677 +Query 1/1: Action query time = 2.466 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7250 +t=74: Selected seed 195 with value = 0.7250 +Query 1/1: Action query time = 2.433 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7894 +t=90: Selected seed 195 with value = 0.7894 +Query 1/1: Action query time = 2.368 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8774 +t=106: Selected seed 195 with value = 0.8774 +Query 1/1: Action query time = 2.413 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.392 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=11--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 11 +# successes: 10 (90.9%) + +Task: push the plate to the front of the stove +Starting episode 12... +Query 1/1: Action query time = 2.009 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3565 +t=10: Selected seed 195 with value = 0.3565 +Query 1/1: Action query time = 0.982 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4568 +t=26: Selected seed 195 with value = 0.4568 +Query 1/1: Action query time = 1.056 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5471 +t=42: Selected seed 195 with value = 0.5471 +Query 1/1: Action query time = 1.570 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6038 +t=58: Selected seed 195 with value = 0.6038 +Query 1/1: Action query time = 1.548 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6481 +t=74: Selected seed 195 with value = 0.6481 +Query 1/1: Action query time = 1.854 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7444 +t=90: Selected seed 195 with value = 0.7444 +Query 1/1: Action query time = 2.094 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8303 +t=106: Selected seed 195 with value = 0.8303 +Query 1/1: Action query time = 2.358 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9221 +t=122: Selected seed 195 with value = 0.9221 +Query 1/1: Action query time = 2.265 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9851 +t=138: Selected seed 195 with value = 0.9851 +Query 1/1: Action query time = 2.170 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t5_s3/2026_08_03-15_19_31--with_future_img--episode=12--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 12 +# successes: 11 (91.7%) +Current task success rate: 0.9166666666666666 +Current total success rate: 0.9166666666666666 +Final results: +Total episodes: 12 +Total successes: 11 +Overall success rate: 0.9167 (91.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_19_32--fc175full_t7_s2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_19_32--fc175full_t7_s2.txt new file mode 100644 index 0000000000000000000000000000000000000000..18e728844106b5f81331917dcf3c52aab026612c --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_19_32--fc175full_t7_s2.txt @@ -0,0 +1,360 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fc175full_t7_s2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,6,10,14,18,22,26,30,34,38,42,46', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 5.816 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4550 +t=10: Selected seed 195 with value = 0.4550 +Query 1/1: Action query time = 5.452 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5562 +t=26: Selected seed 195 with value = 0.5562 +Query 1/1: Action query time = 4.699 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6522 +t=42: Selected seed 195 with value = 0.6522 +Query 1/1: Action query time = 4.858 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7601 +t=58: Selected seed 195 with value = 0.7601 +Query 1/1: Action query time = 4.684 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8057 +t=74: Selected seed 195 with value = 0.8057 +Query 1/1: Action query time = 4.074 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9365 +t=90: Selected seed 195 with value = 0.9365 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 5.255 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4541 +t=10: Selected seed 195 with value = 0.4541 +Query 1/1: Action query time = 5.199 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5544 +t=26: Selected seed 195 with value = 0.5544 +Query 1/1: Action query time = 4.890 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6384 +t=42: Selected seed 195 with value = 0.6384 +Query 1/1: Action query time = 3.828 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7663 +t=58: Selected seed 195 with value = 0.7663 +Query 1/1: Action query time = 4.712 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7780 +t=74: Selected seed 195 with value = 0.7780 +Query 1/1: Action query time = 4.017 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9295 +t=90: Selected seed 195 with value = 0.9295 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 4.806 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4700 +t=10: Selected seed 195 with value = 0.4700 +Query 1/1: Action query time = 5.045 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5438 +t=26: Selected seed 195 with value = 0.5438 +Query 1/1: Action query time = 4.757 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6398 +t=42: Selected seed 195 with value = 0.6398 +Query 1/1: Action query time = 5.100 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7597 +t=58: Selected seed 195 with value = 0.7597 +Query 1/1: Action query time = 3.384 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8712 +t=74: Selected seed 195 with value = 0.8712 +Query 1/1: Action query time = 3.601 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 3.885 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4679 +t=10: Selected seed 195 with value = 0.4679 +Query 1/1: Action query time = 4.216 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5647 +t=26: Selected seed 195 with value = 0.5647 +Query 1/1: Action query time = 5.163 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6467 +t=42: Selected seed 195 with value = 0.6467 +Query 1/1: Action query time = 3.993 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7348 +t=58: Selected seed 195 with value = 0.7348 +Query 1/1: Action query time = 5.158 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8070 +t=74: Selected seed 195 with value = 0.8070 +Query 1/1: Action query time = 4.400 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9319 +t=90: Selected seed 195 with value = 0.9319 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 3.793 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4875 +t=10: Selected seed 195 with value = 0.4875 +Query 1/1: Action query time = 4.889 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5549 +t=26: Selected seed 195 with value = 0.5549 +Query 1/1: Action query time = 4.434 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6657 +t=42: Selected seed 195 with value = 0.6657 +Query 1/1: Action query time = 5.329 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7727 +t=58: Selected seed 195 with value = 0.7727 +Query 1/1: Action query time = 5.790 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8635 +t=74: Selected seed 195 with value = 0.8635 +Query 1/1: Action query time = 3.958 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 2.043 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4655 +t=10: Selected seed 195 with value = 0.4655 +Query 1/1: Action query time = 4.391 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5653 +t=26: Selected seed 195 with value = 0.5653 +Query 1/1: Action query time = 4.992 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6823 +t=42: Selected seed 195 with value = 0.6823 +Query 1/1: Action query time = 5.278 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7866 +t=58: Selected seed 195 with value = 0.7866 +Query 1/1: Action query time = 5.353 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9153 +t=74: Selected seed 195 with value = 0.9153 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: turn on the stove +Starting episode 7... +Query 1/1: Action query time = 4.615 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4904 +t=10: Selected seed 195 with value = 0.4904 +Query 1/1: Action query time = 2.245 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5457 +t=26: Selected seed 195 with value = 0.5457 +Query 1/1: Action query time = 4.145 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6303 +t=42: Selected seed 195 with value = 0.6303 +Query 1/1: Action query time = 4.542 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7484 +t=58: Selected seed 195 with value = 0.7484 +Query 1/1: Action query time = 4.739 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8429 +t=74: Selected seed 195 with value = 0.8429 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=7--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: turn on the stove +Starting episode 8... +Query 1/1: Action query time = 4.515 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4844 +t=10: Selected seed 195 with value = 0.4844 +Query 1/1: Action query time = 5.085 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5494 +t=26: Selected seed 195 with value = 0.5494 +Query 1/1: Action query time = 4.724 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6656 +t=42: Selected seed 195 with value = 0.6656 +Query 1/1: Action query time = 2.894 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7876 +t=58: Selected seed 195 with value = 0.7876 +Query 1/1: Action query time = 4.819 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9633 +t=74: Selected seed 195 with value = 0.9633 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=8--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: turn on the stove +Starting episode 9... +Query 1/1: Action query time = 3.835 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5142 +t=10: Selected seed 195 with value = 0.5142 +Query 1/1: Action query time = 3.165 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5647 +t=26: Selected seed 195 with value = 0.5647 +Query 1/1: Action query time = 4.388 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6534 +t=42: Selected seed 195 with value = 0.6534 +Query 1/1: Action query time = 4.978 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7565 +t=58: Selected seed 195 with value = 0.7565 +Query 1/1: Action query time = 4.824 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8825 +t=74: Selected seed 195 with value = 0.8825 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=9--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: turn on the stove +Starting episode 10... +Query 1/1: Action query time = 3.629 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5100 +t=10: Selected seed 195 with value = 0.5100 +Query 1/1: Action query time = 4.495 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5997 +t=26: Selected seed 195 with value = 0.5997 +Query 1/1: Action query time = 4.732 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7292 +t=42: Selected seed 195 with value = 0.7292 +Query 1/1: Action query time = 4.108 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8648 +t=58: Selected seed 195 with value = 0.8648 +Query 1/1: Action query time = 4.893 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9970 +t=74: Selected seed 195 with value = 0.9970 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=10--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: turn on the stove +Starting episode 11... +Query 1/1: Action query time = 5.169 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5096 +t=10: Selected seed 195 with value = 0.5096 +Query 1/1: Action query time = 3.424 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5811 +t=26: Selected seed 195 with value = 0.5811 +Query 1/1: Action query time = 4.771 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6834 +t=42: Selected seed 195 with value = 0.6834 +Query 1/1: Action query time = 5.269 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8020 +t=58: Selected seed 195 with value = 0.8020 +Query 1/1: Action query time = 5.857 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9624 +t=74: Selected seed 195 with value = 0.9624 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=11--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: turn on the stove +Starting episode 12... +Query 1/1: Action query time = 4.294 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4565 +t=10: Selected seed 195 with value = 0.4565 +Query 1/1: Action query time = 4.392 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5579 +t=26: Selected seed 195 with value = 0.5579 +Query 1/1: Action query time = 4.754 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6430 +t=42: Selected seed 195 with value = 0.6430 +Query 1/1: Action query time = 4.347 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7677 +t=58: Selected seed 195 with value = 0.7677 +Query 1/1: Action query time = 4.572 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8601 +t=74: Selected seed 195 with value = 0.8601 +Query 1/1: Action query time = 4.350 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=90: Selected seed 195 with value = 0.9980 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s2/2026_08_03-15_19_32--with_future_img--episode=12--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 12 +Total successes: 12 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_19_32--fc175full_t7_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_19_32--fc175full_t7_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..ac20334cf9ab09bd80cb394fd768e497eff178a8 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_19_32--fc175full_t7_s3.txt @@ -0,0 +1,348 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_from100_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='fc175full_t7_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,7,11,15,19,23,27,31,35,39,43,47', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 4.445 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4577 +t=10: Selected seed 195 with value = 0.4577 +Query 1/1: Action query time = 4.601 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5555 +t=26: Selected seed 195 with value = 0.5555 +Query 1/1: Action query time = 5.034 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6678 +t=42: Selected seed 195 with value = 0.6678 +Query 1/1: Action query time = 4.870 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7563 +t=58: Selected seed 195 with value = 0.7563 +Query 1/1: Action query time = 4.779 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9020 +t=74: Selected seed 195 with value = 0.9020 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 4.334 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4769 +t=10: Selected seed 195 with value = 0.4769 +Query 1/1: Action query time = 2.034 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5680 +t=26: Selected seed 195 with value = 0.5680 +Query 1/1: Action query time = 5.599 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6436 +t=42: Selected seed 195 with value = 0.6436 +Query 1/1: Action query time = 4.412 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7566 +t=58: Selected seed 195 with value = 0.7566 +Query 1/1: Action query time = 4.780 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8817 +t=74: Selected seed 195 with value = 0.8817 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 4.824 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4945 +t=10: Selected seed 195 with value = 0.4945 +Query 1/1: Action query time = 4.414 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5760 +t=26: Selected seed 195 with value = 0.5760 +Query 1/1: Action query time = 3.392 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6540 +t=42: Selected seed 195 with value = 0.6540 +Query 1/1: Action query time = 4.319 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7643 +t=58: Selected seed 195 with value = 0.7643 +Query 1/1: Action query time = 5.300 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9325 +t=74: Selected seed 195 with value = 0.9325 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 4.535 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5106 +t=10: Selected seed 195 with value = 0.5106 +Query 1/1: Action query time = 5.074 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5614 +t=26: Selected seed 195 with value = 0.5614 +Query 1/1: Action query time = 4.854 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6547 +t=42: Selected seed 195 with value = 0.6547 +Query 1/1: Action query time = 4.365 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7792 +t=58: Selected seed 195 with value = 0.7792 +Query 1/1: Action query time = 3.602 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9143 +t=74: Selected seed 195 with value = 0.9143 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 3.679 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4839 +t=10: Selected seed 195 with value = 0.4839 +Query 1/1: Action query time = 5.024 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5690 +t=26: Selected seed 195 with value = 0.5690 +Query 1/1: Action query time = 4.814 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6385 +t=42: Selected seed 195 with value = 0.6385 +Query 1/1: Action query time = 4.624 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7121 +t=58: Selected seed 195 with value = 0.7121 +Query 1/1: Action query time = 4.734 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8080 +t=74: Selected seed 195 with value = 0.8080 +Query 1/1: Action query time = 4.076 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9655 +t=90: Selected seed 195 with value = 0.9655 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 5.037 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4645 +t=10: Selected seed 195 with value = 0.4645 +Query 1/1: Action query time = 4.422 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5573 +t=26: Selected seed 195 with value = 0.5573 +Query 1/1: Action query time = 4.612 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6375 +t=42: Selected seed 195 with value = 0.6375 +Query 1/1: Action query time = 5.264 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7526 +t=58: Selected seed 195 with value = 0.7526 +Query 1/1: Action query time = 3.811 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8580 +t=74: Selected seed 195 with value = 0.8580 +Query 1/1: Action query time = 2.068 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9879 +t=90: Selected seed 195 with value = 0.9879 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: turn on the stove +Starting episode 7... +Query 1/1: Action query time = 4.572 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4816 +t=10: Selected seed 195 with value = 0.4816 +Query 1/1: Action query time = 4.356 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5742 +t=26: Selected seed 195 with value = 0.5742 +Query 1/1: Action query time = 4.484 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6484 +t=42: Selected seed 195 with value = 0.6484 +Query 1/1: Action query time = 4.593 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7617 +t=58: Selected seed 195 with value = 0.7617 +Query 1/1: Action query time = 4.237 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9419 +t=74: Selected seed 195 with value = 0.9419 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=7--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: turn on the stove +Starting episode 8... +Query 1/1: Action query time = 2.415 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4955 +t=10: Selected seed 195 with value = 0.4955 +Query 1/1: Action query time = 4.719 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5641 +t=26: Selected seed 195 with value = 0.5641 +Query 1/1: Action query time = 5.113 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7052 +t=42: Selected seed 195 with value = 0.7052 +Query 1/1: Action query time = 5.145 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8419 +t=58: Selected seed 195 with value = 0.8419 +Query 1/1: Action query time = 3.731 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=8--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: turn on the stove +Starting episode 9... +Query 1/1: Action query time = 4.518 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4950 +t=10: Selected seed 195 with value = 0.4950 +Query 1/1: Action query time = 4.372 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5790 +t=26: Selected seed 195 with value = 0.5790 +Query 1/1: Action query time = 3.441 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6841 +t=42: Selected seed 195 with value = 0.6841 +Query 1/1: Action query time = 5.198 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7966 +t=58: Selected seed 195 with value = 0.7966 +Query 1/1: Action query time = 4.575 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9310 +t=74: Selected seed 195 with value = 0.9310 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=9--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: turn on the stove +Starting episode 10... +Query 1/1: Action query time = 3.309 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4857 +t=10: Selected seed 195 with value = 0.4857 +Query 1/1: Action query time = 4.141 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5755 +t=26: Selected seed 195 with value = 0.5755 +Query 1/1: Action query time = 5.150 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6715 +t=42: Selected seed 195 with value = 0.6715 +Query 1/1: Action query time = 4.916 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7866 +t=58: Selected seed 195 with value = 0.7866 +Query 1/1: Action query time = 3.967 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9231 +t=74: Selected seed 195 with value = 0.9231 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=10--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: turn on the stove +Starting episode 11... +Query 1/1: Action query time = 4.374 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4708 +t=10: Selected seed 195 with value = 0.4708 +Query 1/1: Action query time = 4.415 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5480 +t=26: Selected seed 195 with value = 0.5480 +Query 1/1: Action query time = 4.347 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6230 +t=42: Selected seed 195 with value = 0.6230 +Query 1/1: Action query time = 5.218 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7267 +t=58: Selected seed 195 with value = 0.7267 +Query 1/1: Action query time = 4.500 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8645 +t=74: Selected seed 195 with value = 0.8645 +Query 1/1: Action query time = 5.119 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9876 +t=90: Selected seed 195 with value = 0.9876 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=11--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: turn on the stove +Starting episode 12... +Query 1/1: Action query time = 4.846 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4982 +t=10: Selected seed 195 with value = 0.4982 +Query 1/1: Action query time = 5.367 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5896 +t=26: Selected seed 195 with value = 0.5896 +Query 1/1: Action query time = 3.960 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6560 +t=42: Selected seed 195 with value = 0.6560 +Query 1/1: Action query time = 3.727 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7652 +t=58: Selected seed 195 with value = 0.7652 +Query 1/1: Action query time = 4.363 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9395 +t=74: Selected seed 195 with value = 0.9395 +Saved rollout MP4 at path ./rollouts/fc175full_t7_s3/2026_08_03-15_19_32--with_future_img--episode=12--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 12 +Total successes: 12 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_54_03--v2i175_t0_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_54_03--v2i175_t0_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..f3b7e9fab7ee52083a1f004f7813a207c0d75a63 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_54_03--v2i175_t0_s3.txt @@ -0,0 +1,480 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i175_t0_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,7,11,15,19,23,27,31,35,39,43,47', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 6.574 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4004 +t=10: Selected seed 195 with value = 0.4004 +Query 1/1: Action query time = 4.404 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4591 +t=26: Selected seed 195 with value = 0.4591 +Query 1/1: Action query time = 5.299 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5657 +t=42: Selected seed 195 with value = 0.5657 +Query 1/1: Action query time = 5.641 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6375 +t=58: Selected seed 195 with value = 0.6375 +Query 1/1: Action query time = 5.080 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7594 +t=74: Selected seed 195 with value = 0.7594 +Query 1/1: Action query time = 3.958 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9494 +t=90: Selected seed 195 with value = 0.9494 +Query 1/1: Action query time = 2.726 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.593 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.460 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3920 +t=10: Selected seed 195 with value = 0.3920 +Query 1/1: Action query time = 4.895 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4735 +t=26: Selected seed 195 with value = 0.4735 +Query 1/1: Action query time = 5.057 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5530 +t=42: Selected seed 195 with value = 0.5530 +Query 1/1: Action query time = 4.652 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6505 +t=58: Selected seed 195 with value = 0.6505 +Query 1/1: Action query time = 4.675 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7453 +t=74: Selected seed 195 with value = 0.7453 +Query 1/1: Action query time = 4.466 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8810 +t=90: Selected seed 195 with value = 0.8810 +Query 1/1: Action query time = 5.135 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=106: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 5.314 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 3.994 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3820 +t=10: Selected seed 195 with value = 0.3820 +Query 1/1: Action query time = 4.304 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4551 +t=26: Selected seed 195 with value = 0.4551 +Query 1/1: Action query time = 5.347 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5518 +t=42: Selected seed 195 with value = 0.5518 +Query 1/1: Action query time = 5.019 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6463 +t=58: Selected seed 195 with value = 0.6463 +Query 1/1: Action query time = 4.293 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7779 +t=74: Selected seed 195 with value = 0.7779 +Query 1/1: Action query time = 4.831 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9218 +t=90: Selected seed 195 with value = 0.9218 +Query 1/1: Action query time = 4.589 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=106: Selected seed 195 with value = 0.9963 +Query 1/1: Action query time = 4.151 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 4... +Query 1/1: Action query time = 4.000 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3798 +t=10: Selected seed 195 with value = 0.3798 +Query 1/1: Action query time = 5.508 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4521 +t=26: Selected seed 195 with value = 0.4521 +Query 1/1: Action query time = 4.173 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5346 +t=42: Selected seed 195 with value = 0.5346 +Query 1/1: Action query time = 4.793 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6300 +t=58: Selected seed 195 with value = 0.6300 +Query 1/1: Action query time = 4.121 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7586 +t=74: Selected seed 195 with value = 0.7586 +Query 1/1: Action query time = 5.028 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9084 +t=90: Selected seed 195 with value = 0.9084 +Query 1/1: Action query time = 3.196 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9902 +t=106: Selected seed 195 with value = 0.9902 +Query 1/1: Action query time = 4.257 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=4--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 5... +Query 1/1: Action query time = 4.303 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4397 +t=10: Selected seed 195 with value = 0.4397 +Query 1/1: Action query time = 4.776 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5078 +t=26: Selected seed 195 with value = 0.5078 +Query 1/1: Action query time = 5.904 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5688 +t=42: Selected seed 195 with value = 0.5688 +Query 1/1: Action query time = 4.444 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6481 +t=58: Selected seed 195 with value = 0.6481 +Query 1/1: Action query time = 5.454 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7749 +t=74: Selected seed 195 with value = 0.7749 +Query 1/1: Action query time = 5.577 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9002 +t=90: Selected seed 195 with value = 0.9002 +Query 1/1: Action query time = 4.922 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=106: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 5.747 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=5--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 6... +Query 1/1: Action query time = 5.005 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4012 +t=10: Selected seed 195 with value = 0.4012 +Query 1/1: Action query time = 3.384 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4367 +t=26: Selected seed 195 with value = 0.4367 +Query 1/1: Action query time = 5.881 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5651 +t=42: Selected seed 195 with value = 0.5651 +Query 1/1: Action query time = 5.303 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6609 +t=58: Selected seed 195 with value = 0.6609 +Query 1/1: Action query time = 4.639 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7610 +t=74: Selected seed 195 with value = 0.7610 +Query 1/1: Action query time = 4.382 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8932 +t=90: Selected seed 195 with value = 0.8932 +Query 1/1: Action query time = 4.730 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=106: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 4.022 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=6--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 7... +Query 1/1: Action query time = 5.189 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4226 +t=10: Selected seed 195 with value = 0.4226 +Query 1/1: Action query time = 3.811 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4862 +t=26: Selected seed 195 with value = 0.4862 +Query 1/1: Action query time = 4.135 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5607 +t=42: Selected seed 195 with value = 0.5607 +Query 1/1: Action query time = 5.051 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6536 +t=58: Selected seed 195 with value = 0.6536 +Query 1/1: Action query time = 4.656 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7460 +t=74: Selected seed 195 with value = 0.7460 +Query 1/1: Action query time = 3.399 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8962 +t=90: Selected seed 195 with value = 0.8962 +Query 1/1: Action query time = 5.632 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=106: Selected seed 195 with value = 0.9980 +Query 1/1: Action query time = 5.761 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=7--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 8... +Query 1/1: Action query time = 5.102 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3427 +t=10: Selected seed 195 with value = 0.3427 +Query 1/1: Action query time = 3.438 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4553 +t=26: Selected seed 195 with value = 0.4553 +Query 1/1: Action query time = 5.058 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5418 +t=42: Selected seed 195 with value = 0.5418 +Query 1/1: Action query time = 5.618 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6240 +t=58: Selected seed 195 with value = 0.6240 +Query 1/1: Action query time = 5.201 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7350 +t=74: Selected seed 195 with value = 0.7350 +Query 1/1: Action query time = 4.192 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9033 +t=90: Selected seed 195 with value = 0.9033 +Query 1/1: Action query time = 4.970 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.570 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=8--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 9... +Query 1/1: Action query time = 4.799 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3999 +t=10: Selected seed 195 with value = 0.3999 +Query 1/1: Action query time = 4.003 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4756 +t=26: Selected seed 195 with value = 0.4756 +Query 1/1: Action query time = 4.220 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5507 +t=42: Selected seed 195 with value = 0.5507 +Query 1/1: Action query time = 5.681 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6580 +t=58: Selected seed 195 with value = 0.6580 +Query 1/1: Action query time = 5.356 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7548 +t=74: Selected seed 195 with value = 0.7548 +Query 1/1: Action query time = 3.279 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8855 +t=90: Selected seed 195 with value = 0.8855 +Query 1/1: Action query time = 3.392 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.509 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=9--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 10... +Query 1/1: Action query time = 3.456 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4287 +t=10: Selected seed 195 with value = 0.4287 +Query 1/1: Action query time = 5.538 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4574 +t=26: Selected seed 195 with value = 0.4574 +Query 1/1: Action query time = 4.145 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5417 +t=42: Selected seed 195 with value = 0.5417 +Query 1/1: Action query time = 3.036 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6233 +t=58: Selected seed 195 with value = 0.6233 +Query 1/1: Action query time = 4.513 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7257 +t=74: Selected seed 195 with value = 0.7257 +Query 1/1: Action query time = 3.242 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8828 +t=90: Selected seed 195 with value = 0.8828 +Query 1/1: Action query time = 4.849 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.993 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=10--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 11... +Query 1/1: Action query time = 3.921 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3977 +t=10: Selected seed 195 with value = 0.3977 +Query 1/1: Action query time = 4.054 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4355 +t=26: Selected seed 195 with value = 0.4355 +Query 1/1: Action query time = 4.128 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5707 +t=42: Selected seed 195 with value = 0.5707 +Query 1/1: Action query time = 5.156 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6509 +t=58: Selected seed 195 with value = 0.6509 +Query 1/1: Action query time = 3.204 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7592 +t=74: Selected seed 195 with value = 0.7592 +Query 1/1: Action query time = 3.785 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8864 +t=90: Selected seed 195 with value = 0.8864 +Query 1/1: Action query time = 4.466 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=106: Selected seed 195 with value = 0.9924 +Query 1/1: Action query time = 3.219 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=11--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 12... +Query 1/1: Action query time = 2.832 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4358 +t=10: Selected seed 195 with value = 0.4358 +Query 1/1: Action query time = 3.477 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4977 +t=26: Selected seed 195 with value = 0.4977 +Query 1/1: Action query time = 3.785 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5579 +t=42: Selected seed 195 with value = 0.5579 +Query 1/1: Action query time = 4.007 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6710 +t=58: Selected seed 195 with value = 0.6710 +Query 1/1: Action query time = 3.525 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8009 +t=74: Selected seed 195 with value = 0.8009 +Query 1/1: Action query time = 2.926 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9328 +t=90: Selected seed 195 with value = 0.9328 +Query 1/1: Action query time = 3.517 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.260 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t0_s3/2026_08_03-15_54_03--with_future_img--episode=12--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 12 +Total successes: 12 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_54_04--v2i175_t3_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_54_04--v2i175_t3_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..c77710c9f218fe75c961aea2524e1696d65b6daa --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-15_54_04--v2i175_t3_s1.txt @@ -0,0 +1,807 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i175_t3_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 6.243 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2220 +t=10: Selected seed 195 with value = 0.2220 +Query 1/1: Action query time = 5.373 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2504 +t=26: Selected seed 195 with value = 0.2504 +Query 1/1: Action query time = 4.662 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3081 +t=42: Selected seed 195 with value = 0.3081 +Query 1/1: Action query time = 5.066 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3580 +t=58: Selected seed 195 with value = 0.3580 +Query 1/1: Action query time = 4.814 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4159 +t=74: Selected seed 195 with value = 0.4159 +Query 1/1: Action query time = 4.228 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4879 +t=90: Selected seed 195 with value = 0.4879 +Query 1/1: Action query time = 2.909 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6001 +t=106: Selected seed 195 with value = 0.6001 +Query 1/1: Action query time = 4.714 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7249 +t=122: Selected seed 195 with value = 0.7249 +Query 1/1: Action query time = 5.213 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8813 +t=138: Selected seed 195 with value = 0.8813 +Query 1/1: Action query time = 5.372 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.161 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2060 +t=10: Selected seed 195 with value = 0.2060 +Query 1/1: Action query time = 5.220 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2570 +t=26: Selected seed 195 with value = 0.2570 +Query 1/1: Action query time = 3.453 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3024 +t=42: Selected seed 195 with value = 0.3024 +Query 1/1: Action query time = 4.536 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3363 +t=58: Selected seed 195 with value = 0.3363 +Query 1/1: Action query time = 4.922 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4055 +t=74: Selected seed 195 with value = 0.4055 +Query 1/1: Action query time = 5.523 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4432 +t=90: Selected seed 195 with value = 0.4432 +Query 1/1: Action query time = 5.417 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5653 +t=106: Selected seed 195 with value = 0.5653 +Query 1/1: Action query time = 4.030 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6477 +t=122: Selected seed 195 with value = 0.6477 +Query 1/1: Action query time = 3.607 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7573 +t=138: Selected seed 195 with value = 0.7573 +Query 1/1: Action query time = 3.752 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8945 +t=154: Selected seed 195 with value = 0.8945 +Query 1/1: Action query time = 5.502 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.223 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2327 +t=10: Selected seed 195 with value = 0.2327 +Query 1/1: Action query time = 4.826 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2695 +t=26: Selected seed 195 with value = 0.2695 +Query 1/1: Action query time = 4.544 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3032 +t=42: Selected seed 195 with value = 0.3032 +Query 1/1: Action query time = 5.195 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3458 +t=58: Selected seed 195 with value = 0.3458 +Query 1/1: Action query time = 4.753 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4028 +t=74: Selected seed 195 with value = 0.4028 +Query 1/1: Action query time = 5.143 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4627 +t=90: Selected seed 195 with value = 0.4627 +Query 1/1: Action query time = 5.954 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5713 +t=106: Selected seed 195 with value = 0.5713 +Query 1/1: Action query time = 4.425 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6973 +t=122: Selected seed 195 with value = 0.6973 +Query 1/1: Action query time = 3.851 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8061 +t=138: Selected seed 195 with value = 0.8061 +Query 1/1: Action query time = 4.403 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8743 +t=154: Selected seed 195 with value = 0.8743 +Query 1/1: Action query time = 4.951 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=170: Selected seed 195 with value = 0.9986 +Query 1/1: Action query time = 3.787 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9464 +t=186: Selected seed 195 with value = 0.9464 +Query 1/1: Action query time = 4.258 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9572 +t=202: Selected seed 195 with value = 0.9572 +Query 1/1: Action query time = 5.415 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9138 +t=218: Selected seed 195 with value = 0.9138 +Query 1/1: Action query time = 5.573 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3481 +t=234: Selected seed 195 with value = 0.3481 +Query 1/1: Action query time = 4.430 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9304 +t=250: Selected seed 195 with value = 0.9304 +Query 1/1: Action query time = 3.537 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8865 +t=266: Selected seed 195 with value = 0.8865 +Query 1/1: Action query time = 3.890 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6212 +t=282: Selected seed 195 with value = 0.6212 +Query 1/1: Action query time = 4.470 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5424 +t=298: Selected seed 195 with value = 0.5424 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 4.636 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2222 +t=10: Selected seed 195 with value = 0.2222 +Query 1/1: Action query time = 4.473 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2615 +t=26: Selected seed 195 with value = 0.2615 +Query 1/1: Action query time = 4.324 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3009 +t=42: Selected seed 195 with value = 0.3009 +Query 1/1: Action query time = 5.518 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3379 +t=58: Selected seed 195 with value = 0.3379 +Query 1/1: Action query time = 4.753 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4219 +t=74: Selected seed 195 with value = 0.4219 +Query 1/1: Action query time = 3.187 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4493 +t=90: Selected seed 195 with value = 0.4493 +Query 1/1: Action query time = 6.145 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5541 +t=106: Selected seed 195 with value = 0.5541 +Query 1/1: Action query time = 5.765 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6240 +t=122: Selected seed 195 with value = 0.6240 +Query 1/1: Action query time = 5.268 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8037 +t=138: Selected seed 195 with value = 0.8037 +Query 1/1: Action query time = 4.990 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8836 +t=154: Selected seed 195 with value = 0.8836 +Query 1/1: Action query time = 4.588 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9967 +t=170: Selected seed 195 with value = 0.9967 +Query 1/1: Action query time = 4.091 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9509 +t=186: Selected seed 195 with value = 0.9509 +Query 1/1: Action query time = 3.129 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9493 +t=202: Selected seed 195 with value = 0.9493 +Query 1/1: Action query time = 4.349 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9455 +t=218: Selected seed 195 with value = 0.9455 +Query 1/1: Action query time = 5.066 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9443 +t=234: Selected seed 195 with value = 0.9443 +Query 1/1: Action query time = 4.410 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3576 +t=250: Selected seed 195 with value = 0.3576 +Query 1/1: Action query time = 4.822 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3400 +t=266: Selected seed 195 with value = 0.3400 +Query 1/1: Action query time = 4.996 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3846 +t=282: Selected seed 195 with value = 0.3846 +Query 1/1: Action query time = 4.071 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4208 +t=298: Selected seed 195 with value = 0.4208 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=4--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 4 +# successes: 2 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 5... +Query 1/1: Action query time = 5.585 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2352 +t=10: Selected seed 195 with value = 0.2352 +Query 1/1: Action query time = 3.306 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2458 +t=26: Selected seed 195 with value = 0.2458 +Query 1/1: Action query time = 5.006 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3165 +t=42: Selected seed 195 with value = 0.3165 +Query 1/1: Action query time = 4.771 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3579 +t=58: Selected seed 195 with value = 0.3579 +Query 1/1: Action query time = 4.086 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4073 +t=74: Selected seed 195 with value = 0.4073 +Query 1/1: Action query time = 3.610 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4751 +t=90: Selected seed 195 with value = 0.4751 +Query 1/1: Action query time = 5.503 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5799 +t=106: Selected seed 195 with value = 0.5799 +Query 1/1: Action query time = 4.854 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6277 +t=122: Selected seed 195 with value = 0.6277 +Query 1/1: Action query time = 4.192 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7053 +t=138: Selected seed 195 with value = 0.7053 +Query 1/1: Action query time = 4.457 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8516 +t=154: Selected seed 195 with value = 0.8516 +Query 1/1: Action query time = 5.921 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9851 +t=170: Selected seed 195 with value = 0.9851 +Query 1/1: Action query time = 5.087 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9530 +t=186: Selected seed 195 with value = 0.9530 +Query 1/1: Action query time = 4.799 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9467 +t=202: Selected seed 195 with value = 0.9467 +Query 1/1: Action query time = 3.986 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9510 +t=218: Selected seed 195 with value = 0.9510 +Query 1/1: Action query time = 5.078 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9539 +t=234: Selected seed 195 with value = 0.9539 +Query 1/1: Action query time = 6.361 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9629 +t=250: Selected seed 195 with value = 0.9629 +Query 1/1: Action query time = 5.604 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3140 +t=266: Selected seed 195 with value = 0.3140 +Query 1/1: Action query time = 3.664 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3726 +t=282: Selected seed 195 with value = 0.3726 +Query 1/1: Action query time = 3.660 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8845 +t=298: Selected seed 195 with value = 0.8845 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=5--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 5 +# successes: 2 (40.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 6... +Query 1/1: Action query time = 5.571 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2200 +t=10: Selected seed 195 with value = 0.2200 +Query 1/1: Action query time = 4.472 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2361 +t=26: Selected seed 195 with value = 0.2361 +Query 1/1: Action query time = 5.993 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3038 +t=42: Selected seed 195 with value = 0.3038 +Query 1/1: Action query time = 2.741 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3461 +t=58: Selected seed 195 with value = 0.3461 +Query 1/1: Action query time = 4.623 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4010 +t=74: Selected seed 195 with value = 0.4010 +Query 1/1: Action query time = 4.713 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4551 +t=90: Selected seed 195 with value = 0.4551 +Query 1/1: Action query time = 5.067 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5657 +t=106: Selected seed 195 with value = 0.5657 +Query 1/1: Action query time = 5.463 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5918 +t=122: Selected seed 195 with value = 0.5918 +Query 1/1: Action query time = 4.383 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7567 +t=138: Selected seed 195 with value = 0.7567 +Query 1/1: Action query time = 2.978 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8150 +t=154: Selected seed 195 with value = 0.8150 +Query 1/1: Action query time = 5.312 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9376 +t=170: Selected seed 195 with value = 0.9376 +Query 1/1: Action query time = 4.048 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=6--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 6 +# successes: 3 (50.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 7... +Query 1/1: Action query time = 4.484 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2180 +t=10: Selected seed 195 with value = 0.2180 +Query 1/1: Action query time = 5.315 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2562 +t=26: Selected seed 195 with value = 0.2562 +Query 1/1: Action query time = 4.636 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3006 +t=42: Selected seed 195 with value = 0.3006 +Query 1/1: Action query time = 5.026 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3312 +t=58: Selected seed 195 with value = 0.3312 +Query 1/1: Action query time = 4.348 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3925 +t=74: Selected seed 195 with value = 0.3925 +Query 1/1: Action query time = 4.215 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4430 +t=90: Selected seed 195 with value = 0.4430 +Query 1/1: Action query time = 4.310 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5557 +t=106: Selected seed 195 with value = 0.5557 +Query 1/1: Action query time = 3.362 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6303 +t=122: Selected seed 195 with value = 0.6303 +Query 1/1: Action query time = 2.403 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7727 +t=138: Selected seed 195 with value = 0.7727 +Query 1/1: Action query time = 1.520 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8842 +t=154: Selected seed 195 with value = 0.8842 +Query 1/1: Action query time = 3.316 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9974 +t=170: Selected seed 195 with value = 0.9974 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=7--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 7 +# successes: 4 (57.1%) + +Task: open the top drawer and put the bowl inside +Starting episode 8... +Query 1/1: Action query time = 2.818 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2171 +t=10: Selected seed 195 with value = 0.2171 +Query 1/1: Action query time = 2.970 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2658 +t=26: Selected seed 195 with value = 0.2658 +Query 1/1: Action query time = 3.088 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3019 +t=42: Selected seed 195 with value = 0.3019 +Query 1/1: Action query time = 2.543 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3416 +t=58: Selected seed 195 with value = 0.3416 +Query 1/1: Action query time = 2.014 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4048 +t=74: Selected seed 195 with value = 0.4048 +Query 1/1: Action query time = 2.817 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4486 +t=90: Selected seed 195 with value = 0.4486 +Query 1/1: Action query time = 2.773 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5650 +t=106: Selected seed 195 with value = 0.5650 +Query 1/1: Action query time = 3.152 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6803 +t=122: Selected seed 195 with value = 0.6803 +Query 1/1: Action query time = 2.991 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7911 +t=138: Selected seed 195 with value = 0.7911 +Query 1/1: Action query time = 2.189 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9168 +t=154: Selected seed 195 with value = 0.9168 +Query 1/1: Action query time = 2.593 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=8--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 8 +# successes: 5 (62.5%) + +Task: open the top drawer and put the bowl inside +Starting episode 9... +Query 1/1: Action query time = 3.646 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2160 +t=10: Selected seed 195 with value = 0.2160 +Query 1/1: Action query time = 2.911 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2565 +t=26: Selected seed 195 with value = 0.2565 +Query 1/1: Action query time = 2.719 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2953 +t=42: Selected seed 195 with value = 0.2953 +Query 1/1: Action query time = 2.743 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3292 +t=58: Selected seed 195 with value = 0.3292 +Query 1/1: Action query time = 2.040 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3834 +t=74: Selected seed 195 with value = 0.3834 +Query 1/1: Action query time = 2.079 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4402 +t=90: Selected seed 195 with value = 0.4402 +Query 1/1: Action query time = 2.535 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5437 +t=106: Selected seed 195 with value = 0.5437 +Query 1/1: Action query time = 2.893 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6095 +t=122: Selected seed 195 with value = 0.6095 +Query 1/1: Action query time = 3.031 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6882 +t=138: Selected seed 195 with value = 0.6882 +Query 1/1: Action query time = 3.066 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8462 +t=154: Selected seed 195 with value = 0.8462 +Query 1/1: Action query time = 2.383 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9806 +t=170: Selected seed 195 with value = 0.9806 +Query 1/1: Action query time = 2.880 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=186: Selected seed 195 with value = 0.9966 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=9--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 9 +# successes: 6 (66.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 10... +Query 1/1: Action query time = 2.481 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2004 +t=10: Selected seed 195 with value = 0.2004 +Query 1/1: Action query time = 2.721 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2625 +t=26: Selected seed 195 with value = 0.2625 +Query 1/1: Action query time = 2.591 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3040 +t=42: Selected seed 195 with value = 0.3040 +Query 1/1: Action query time = 1.608 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3376 +t=58: Selected seed 195 with value = 0.3376 +Query 1/1: Action query time = 1.291 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4033 +t=74: Selected seed 195 with value = 0.4033 +Query 1/1: Action query time = 1.594 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4549 +t=90: Selected seed 195 with value = 0.4549 +Query 1/1: Action query time = 2.403 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5680 +t=106: Selected seed 195 with value = 0.5680 +Query 1/1: Action query time = 2.477 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6616 +t=122: Selected seed 195 with value = 0.6616 +Query 1/1: Action query time = 2.526 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7921 +t=138: Selected seed 195 with value = 0.7921 +Query 1/1: Action query time = 2.391 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9274 +t=154: Selected seed 195 with value = 0.9274 +Query 1/1: Action query time = 1.853 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9944 +t=170: Selected seed 195 with value = 0.9944 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=10--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 10 +# successes: 7 (70.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 11... +Query 1/1: Action query time = 2.179 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2086 +t=10: Selected seed 195 with value = 0.2086 +Query 1/1: Action query time = 2.179 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2418 +t=26: Selected seed 195 with value = 0.2418 +Query 1/1: Action query time = 2.310 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3102 +t=42: Selected seed 195 with value = 0.3102 +Query 1/1: Action query time = 2.448 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3565 +t=58: Selected seed 195 with value = 0.3565 +Query 1/1: Action query time = 1.834 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4139 +t=74: Selected seed 195 with value = 0.4139 +Query 1/1: Action query time = 1.448 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4985 +t=90: Selected seed 195 with value = 0.4985 +Query 1/1: Action query time = 1.490 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6094 +t=106: Selected seed 195 with value = 0.6094 +Query 1/1: Action query time = 1.560 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7809 +t=122: Selected seed 195 with value = 0.7809 +Query 1/1: Action query time = 2.041 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8519 +t=138: Selected seed 195 with value = 0.8519 +Query 1/1: Action query time = 2.079 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=154: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 1.191 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9524 +t=170: Selected seed 195 with value = 0.9524 +Query 1/1: Action query time = 1.229 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9498 +t=186: Selected seed 195 with value = 0.9498 +Query 1/1: Action query time = 1.243 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9528 +t=202: Selected seed 195 with value = 0.9528 +Query 1/1: Action query time = 1.278 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3385 +t=218: Selected seed 195 with value = 0.3385 +Query 1/1: Action query time = 1.214 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3456 +t=234: Selected seed 195 with value = 0.3456 +Query 1/1: Action query time = 1.143 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3870 +t=250: Selected seed 195 with value = 0.3870 +Query 1/1: Action query time = 1.238 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4176 +t=266: Selected seed 195 with value = 0.4176 +Query 1/1: Action query time = 1.178 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4328 +t=282: Selected seed 195 with value = 0.4328 +Query 1/1: Action query time = 0.983 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4392 +t=298: Selected seed 195 with value = 0.4392 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=11--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 11 +# successes: 7 (63.6%) + +Task: open the top drawer and put the bowl inside +Starting episode 12... +Query 1/1: Action query time = 1.390 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2161 +t=10: Selected seed 195 with value = 0.2161 +Query 1/1: Action query time = 1.393 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2628 +t=26: Selected seed 195 with value = 0.2628 +Query 1/1: Action query time = 1.421 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2994 +t=42: Selected seed 195 with value = 0.2994 +Query 1/1: Action query time = 1.411 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3367 +t=58: Selected seed 195 with value = 0.3367 +Query 1/1: Action query time = 1.364 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4026 +t=74: Selected seed 195 with value = 0.4026 +Query 1/1: Action query time = 1.482 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4444 +t=90: Selected seed 195 with value = 0.4444 +Query 1/1: Action query time = 1.694 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5531 +t=106: Selected seed 195 with value = 0.5531 +Query 1/1: Action query time = 1.670 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6137 +t=122: Selected seed 195 with value = 0.6137 +Query 1/1: Action query time = 1.430 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7522 +t=138: Selected seed 195 with value = 0.7522 +Query 1/1: Action query time = 0.964 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8751 +t=154: Selected seed 195 with value = 0.8751 +Query 1/1: Action query time = 0.998 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=12--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 12 +# successes: 8 (66.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 13... +Query 1/1: Action query time = 1.028 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2029 +t=10: Selected seed 195 with value = 0.2029 +Query 1/1: Action query time = 1.014 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2370 +t=26: Selected seed 195 with value = 0.2370 +Query 1/1: Action query time = 0.974 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3103 +t=42: Selected seed 195 with value = 0.3103 +Query 1/1: Action query time = 0.975 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3632 +t=58: Selected seed 195 with value = 0.3632 +Query 1/1: Action query time = 0.979 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4190 +t=74: Selected seed 195 with value = 0.4190 +Query 1/1: Action query time = 0.973 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5081 +t=90: Selected seed 195 with value = 0.5081 +Query 1/1: Action query time = 1.003 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6042 +t=106: Selected seed 195 with value = 0.6042 +Query 1/1: Action query time = 0.994 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6734 +t=122: Selected seed 195 with value = 0.6734 +Query 1/1: Action query time = 0.988 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8012 +t=138: Selected seed 195 with value = 0.8012 +Query 1/1: Action query time = 0.975 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9450 +t=154: Selected seed 195 with value = 0.9450 +Query 1/1: Action query time = 0.965 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9943 +t=170: Selected seed 195 with value = 0.9943 +Saved rollout MP4 at path ./rollouts/v2i175_t3_s1/2026_08_03-15_54_04--with_future_img--episode=13--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 13 +# successes: 9 (69.2%) +Current task success rate: 0.6923076923076923 +Current total success rate: 0.6923076923076923 +Final results: +Total episodes: 13 +Total successes: 9 +Overall success rate: 0.6923 (69.2%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_07_38--v2i175_t4_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_07_38--v2i175_t4_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..4068fe0f6e71d752b63976ca672b2522d6512949 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_07_38--v2i175_t4_s0.txt @@ -0,0 +1,415 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i175_t4_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,4,8,12,16,20,24,28,32,36,40,44,48', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 1.693 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5512 +t=10: Selected seed 195 with value = 0.5512 +Query 1/1: Action query time = 1.774 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6825 +t=26: Selected seed 195 with value = 0.6825 +Query 1/1: Action query time = 4.037 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7793 +t=42: Selected seed 195 with value = 0.7793 +Query 1/1: Action query time = 4.115 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9587 +t=58: Selected seed 195 with value = 0.9587 +Query 1/1: Action query time = 4.145 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.038 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.564 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5660 +t=10: Selected seed 195 with value = 0.5660 +Query 1/1: Action query time = 3.331 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6798 +t=26: Selected seed 195 with value = 0.6798 +Query 1/1: Action query time = 3.479 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7633 +t=42: Selected seed 195 with value = 0.7633 +Query 1/1: Action query time = 4.485 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9431 +t=58: Selected seed 195 with value = 0.9431 +Query 1/1: Action query time = 4.646 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.800 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.426 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5790 +t=10: Selected seed 195 with value = 0.5790 +Query 1/1: Action query time = 4.291 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6969 +t=26: Selected seed 195 with value = 0.6969 +Query 1/1: Action query time = 3.716 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7746 +t=42: Selected seed 195 with value = 0.7746 +Query 1/1: Action query time = 5.390 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9349 +t=58: Selected seed 195 with value = 0.9349 +Query 1/1: Action query time = 4.864 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.973 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 4.750 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5732 +t=10: Selected seed 195 with value = 0.5732 +Query 1/1: Action query time = 4.157 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6639 +t=26: Selected seed 195 with value = 0.6639 +Query 1/1: Action query time = 2.905 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7942 +t=42: Selected seed 195 with value = 0.7942 +Query 1/1: Action query time = 5.511 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9264 +t=58: Selected seed 195 with value = 0.9264 +Query 1/1: Action query time = 4.967 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.750 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=4--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 5... +Query 1/1: Action query time = 5.564 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5937 +t=10: Selected seed 195 with value = 0.5937 +Query 1/1: Action query time = 4.554 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6914 +t=26: Selected seed 195 with value = 0.6914 +Query 1/1: Action query time = 2.962 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8312 +t=42: Selected seed 195 with value = 0.8312 +Query 1/1: Action query time = 2.970 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9684 +t=58: Selected seed 195 with value = 0.9684 +Query 1/1: Action query time = 4.484 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.216 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=5--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 6... +Query 1/1: Action query time = 5.914 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5865 +t=10: Selected seed 195 with value = 0.5865 +Query 1/1: Action query time = 5.343 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6805 +t=26: Selected seed 195 with value = 0.6805 +Query 1/1: Action query time = 4.871 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8112 +t=42: Selected seed 195 with value = 0.8112 +Query 1/1: Action query time = 2.742 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9317 +t=58: Selected seed 195 with value = 0.9317 +Query 1/1: Action query time = 4.717 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.037 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=6--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 7... +Query 1/1: Action query time = 5.503 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5643 +t=10: Selected seed 195 with value = 0.5643 +Query 1/1: Action query time = 5.398 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6971 +t=26: Selected seed 195 with value = 0.6971 +Query 1/1: Action query time = 3.591 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8014 +t=42: Selected seed 195 with value = 0.8014 +Query 1/1: Action query time = 4.081 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9804 +t=58: Selected seed 195 with value = 0.9804 +Query 1/1: Action query time = 5.500 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.533 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=7--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 8... +Query 1/1: Action query time = 5.352 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5605 +t=10: Selected seed 195 with value = 0.5605 +Query 1/1: Action query time = 3.959 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6467 +t=26: Selected seed 195 with value = 0.6467 +Query 1/1: Action query time = 3.822 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7441 +t=42: Selected seed 195 with value = 0.7441 +Query 1/1: Action query time = 4.583 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9276 +t=58: Selected seed 195 with value = 0.9276 +Query 1/1: Action query time = 4.489 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.009 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=8--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 9... +Query 1/1: Action query time = 5.092 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6032 +t=10: Selected seed 195 with value = 0.6032 +Query 1/1: Action query time = 5.166 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7220 +t=26: Selected seed 195 with value = 0.7220 +Query 1/1: Action query time = 5.222 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7612 +t=42: Selected seed 195 with value = 0.7612 +Query 1/1: Action query time = 3.995 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9199 +t=58: Selected seed 195 with value = 0.9199 +Query 1/1: Action query time = 3.592 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.738 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=9--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 10... +Query 1/1: Action query time = 5.149 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5686 +t=10: Selected seed 195 with value = 0.5686 +Query 1/1: Action query time = 5.077 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6775 +t=26: Selected seed 195 with value = 0.6775 +Query 1/1: Action query time = 4.701 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7779 +t=42: Selected seed 195 with value = 0.7779 +Query 1/1: Action query time = 4.435 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9348 +t=58: Selected seed 195 with value = 0.9348 +Query 1/1: Action query time = 5.167 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.140 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=10--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 11... +Query 1/1: Action query time = 3.990 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5764 +t=10: Selected seed 195 with value = 0.5764 +Query 1/1: Action query time = 4.062 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6982 +t=26: Selected seed 195 with value = 0.6982 +Query 1/1: Action query time = 4.132 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7594 +t=42: Selected seed 195 with value = 0.7594 +Query 1/1: Action query time = 3.195 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9168 +t=58: Selected seed 195 with value = 0.9168 +Query 1/1: Action query time = 5.021 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=74: Selected seed 195 with value = 0.9966 +Query 1/1: Action query time = 5.068 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=11--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 12... +Query 1/1: Action query time = 5.294 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5955 +t=10: Selected seed 195 with value = 0.5955 +Query 1/1: Action query time = 4.979 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7116 +t=26: Selected seed 195 with value = 0.7116 +Query 1/1: Action query time = 3.538 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7967 +t=42: Selected seed 195 with value = 0.7967 +Query 1/1: Action query time = 3.960 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9104 +t=58: Selected seed 195 with value = 0.9104 +Query 1/1: Action query time = 3.585 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.813 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=12--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 13... +Query 1/1: Action query time = 4.524 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5733 +t=10: Selected seed 195 with value = 0.5733 +Query 1/1: Action query time = 4.988 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6787 +t=26: Selected seed 195 with value = 0.6787 +Query 1/1: Action query time = 4.226 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7555 +t=42: Selected seed 195 with value = 0.7555 +Query 1/1: Action query time = 4.928 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9085 +t=58: Selected seed 195 with value = 0.9085 +Query 1/1: Action query time = 4.459 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=74: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 4.421 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s0/2026_08_03-16_07_38--with_future_img--episode=13--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 13 +Total successes: 13 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_07_38--v2i175_t4_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_07_38--v2i175_t4_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..4525cc4403325ca7b057e41e86dbb325ae9b72e1 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_07_38--v2i175_t4_s1.txt @@ -0,0 +1,419 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i175_t4_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 3.316 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5930 +t=10: Selected seed 195 with value = 0.5930 +Query 1/1: Action query time = 5.150 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6655 +t=26: Selected seed 195 with value = 0.6655 +Query 1/1: Action query time = 4.757 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7554 +t=42: Selected seed 195 with value = 0.7554 +Query 1/1: Action query time = 4.752 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8927 +t=58: Selected seed 195 with value = 0.8927 +Query 1/1: Action query time = 4.485 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9901 +t=74: Selected seed 195 with value = 0.9901 +Query 1/1: Action query time = 4.608 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 3.929 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5923 +t=10: Selected seed 195 with value = 0.5923 +Query 1/1: Action query time = 4.775 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7108 +t=26: Selected seed 195 with value = 0.7108 +Query 1/1: Action query time = 4.922 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7972 +t=42: Selected seed 195 with value = 0.7972 +Query 1/1: Action query time = 5.358 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9542 +t=58: Selected seed 195 with value = 0.9542 +Query 1/1: Action query time = 4.344 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.447 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.545 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5674 +t=10: Selected seed 195 with value = 0.5674 +Query 1/1: Action query time = 4.952 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6832 +t=26: Selected seed 195 with value = 0.6832 +Query 1/1: Action query time = 4.755 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8267 +t=42: Selected seed 195 with value = 0.8267 +Query 1/1: Action query time = 4.939 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9430 +t=58: Selected seed 195 with value = 0.9430 +Query 1/1: Action query time = 4.837 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.559 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 2.955 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5708 +t=10: Selected seed 195 with value = 0.5708 +Query 1/1: Action query time = 5.478 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6642 +t=26: Selected seed 195 with value = 0.6642 +Query 1/1: Action query time = 5.462 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7856 +t=42: Selected seed 195 with value = 0.7856 +Query 1/1: Action query time = 5.134 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8769 +t=58: Selected seed 195 with value = 0.8769 +Query 1/1: Action query time = 5.458 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9710 +t=74: Selected seed 195 with value = 0.9710 +Query 1/1: Action query time = 4.626 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.318 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=4--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 5... +Query 1/1: Action query time = 4.589 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5711 +t=10: Selected seed 195 with value = 0.5711 +Query 1/1: Action query time = 4.998 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6767 +t=26: Selected seed 195 with value = 0.6767 +Query 1/1: Action query time = 3.515 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7930 +t=42: Selected seed 195 with value = 0.7930 +Query 1/1: Action query time = 5.979 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9201 +t=58: Selected seed 195 with value = 0.9201 +Query 1/1: Action query time = 5.386 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.822 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=5--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 6... +Query 1/1: Action query time = 4.548 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5871 +t=10: Selected seed 195 with value = 0.5871 +Query 1/1: Action query time = 3.664 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6711 +t=26: Selected seed 195 with value = 0.6711 +Query 1/1: Action query time = 3.584 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7944 +t=42: Selected seed 195 with value = 0.7944 +Query 1/1: Action query time = 4.857 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9279 +t=58: Selected seed 195 with value = 0.9279 +Query 1/1: Action query time = 5.329 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.650 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=6--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 7... +Query 1/1: Action query time = 4.662 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5523 +t=10: Selected seed 195 with value = 0.5523 +Query 1/1: Action query time = 5.175 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6302 +t=26: Selected seed 195 with value = 0.6302 +Query 1/1: Action query time = 4.690 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7476 +t=42: Selected seed 195 with value = 0.7476 +Query 1/1: Action query time = 4.995 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9091 +t=58: Selected seed 195 with value = 0.9091 +Query 1/1: Action query time = 3.822 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.730 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=7--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 8... +Query 1/1: Action query time = 4.403 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5758 +t=10: Selected seed 195 with value = 0.5758 +Query 1/1: Action query time = 4.550 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6774 +t=26: Selected seed 195 with value = 0.6774 +Query 1/1: Action query time = 5.041 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7673 +t=42: Selected seed 195 with value = 0.7673 +Query 1/1: Action query time = 5.038 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9187 +t=58: Selected seed 195 with value = 0.9187 +Query 1/1: Action query time = 5.463 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.913 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=8--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 9... +Query 1/1: Action query time = 4.063 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5845 +t=10: Selected seed 195 with value = 0.5845 +Query 1/1: Action query time = 3.412 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6866 +t=26: Selected seed 195 with value = 0.6866 +Query 1/1: Action query time = 4.984 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7889 +t=42: Selected seed 195 with value = 0.7889 +Query 1/1: Action query time = 4.954 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9320 +t=58: Selected seed 195 with value = 0.9320 +Query 1/1: Action query time = 4.211 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.092 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=9--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 10... +Query 1/1: Action query time = 4.458 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5738 +t=10: Selected seed 195 with value = 0.5738 +Query 1/1: Action query time = 4.962 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6723 +t=26: Selected seed 195 with value = 0.6723 +Query 1/1: Action query time = 4.786 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7701 +t=42: Selected seed 195 with value = 0.7701 +Query 1/1: Action query time = 3.979 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9115 +t=58: Selected seed 195 with value = 0.9115 +Query 1/1: Action query time = 3.179 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.240 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=90: Selected seed 195 with value = 0.9976 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=10--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 11... +Query 1/1: Action query time = 3.951 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5979 +t=10: Selected seed 195 with value = 0.5979 +Query 1/1: Action query time = 5.316 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7048 +t=26: Selected seed 195 with value = 0.7048 +Query 1/1: Action query time = 5.085 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7413 +t=42: Selected seed 195 with value = 0.7413 +Query 1/1: Action query time = 4.898 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9152 +t=58: Selected seed 195 with value = 0.9152 +Query 1/1: Action query time = 5.149 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.283 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=11--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 12... +Query 1/1: Action query time = 3.695 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5812 +t=10: Selected seed 195 with value = 0.5812 +Query 1/1: Action query time = 4.525 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6866 +t=26: Selected seed 195 with value = 0.6866 +Query 1/1: Action query time = 4.888 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7741 +t=42: Selected seed 195 with value = 0.7741 +Query 1/1: Action query time = 4.214 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9395 +t=58: Selected seed 195 with value = 0.9395 +Query 1/1: Action query time = 5.101 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.656 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=12--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 13... +Query 1/1: Action query time = 4.461 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5761 +t=10: Selected seed 195 with value = 0.5761 +Query 1/1: Action query time = 4.387 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6767 +t=26: Selected seed 195 with value = 0.6767 +Query 1/1: Action query time = 3.528 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7521 +t=42: Selected seed 195 with value = 0.7521 +Query 1/1: Action query time = 3.335 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9172 +t=58: Selected seed 195 with value = 0.9172 +Query 1/1: Action query time = 2.487 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.320 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t4_s1/2026_08_03-16_07_38--with_future_img--episode=13--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 13 +Total successes: 13 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_07_39--v2i175_t5_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_07_39--v2i175_t5_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..19204db365e6cb05550aacebb3222f30669fe03a --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_07_39--v2i175_t5_s3.txt @@ -0,0 +1,512 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i175_t5_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,7,11,15,19,23,27,31,35,39,43,47', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 2.982 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3763 +t=10: Selected seed 195 with value = 0.3763 +Query 1/1: Action query time = 4.697 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4220 +t=26: Selected seed 195 with value = 0.4220 +Query 1/1: Action query time = 4.918 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5402 +t=42: Selected seed 195 with value = 0.5402 +Query 1/1: Action query time = 5.618 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7031 +t=58: Selected seed 195 with value = 0.7031 +Query 1/1: Action query time = 4.825 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8258 +t=74: Selected seed 195 with value = 0.8258 +Query 1/1: Action query time = 4.941 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9307 +t=90: Selected seed 195 with value = 0.9307 +Query 1/1: Action query time = 3.483 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.814 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=122: Selected seed 195 with value = 0.9973 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.984 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3984 +t=10: Selected seed 195 with value = 0.3984 +Query 1/1: Action query time = 5.240 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4416 +t=26: Selected seed 195 with value = 0.4416 +Query 1/1: Action query time = 3.722 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5360 +t=42: Selected seed 195 with value = 0.5360 +Query 1/1: Action query time = 4.605 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5400 +t=58: Selected seed 195 with value = 0.5400 +Query 1/1: Action query time = 3.869 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5992 +t=74: Selected seed 195 with value = 0.5992 +Query 1/1: Action query time = 4.550 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7004 +t=90: Selected seed 195 with value = 0.7004 +Query 1/1: Action query time = 5.103 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7937 +t=106: Selected seed 195 with value = 0.7937 +Query 1/1: Action query time = 4.882 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9131 +t=122: Selected seed 195 with value = 0.9131 +Query 1/1: Action query time = 4.970 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9929 +t=138: Selected seed 195 with value = 0.9929 +Query 1/1: Action query time = 4.233 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9956 +t=154: Selected seed 195 with value = 0.9956 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 3.839 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4067 +t=10: Selected seed 195 with value = 0.4067 +Query 1/1: Action query time = 3.484 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4612 +t=26: Selected seed 195 with value = 0.4612 +Query 1/1: Action query time = 5.090 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6024 +t=42: Selected seed 195 with value = 0.6024 +Query 1/1: Action query time = 5.061 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6520 +t=58: Selected seed 195 with value = 0.6520 +Query 1/1: Action query time = 4.774 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7661 +t=74: Selected seed 195 with value = 0.7661 +Query 1/1: Action query time = 3.660 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8808 +t=90: Selected seed 195 with value = 0.8808 +Query 1/1: Action query time = 5.643 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.720 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 4.253 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4159 +t=10: Selected seed 195 with value = 0.4159 +Query 1/1: Action query time = 4.992 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4687 +t=26: Selected seed 195 with value = 0.4687 +Query 1/1: Action query time = 4.334 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5597 +t=42: Selected seed 195 with value = 0.5597 +Query 1/1: Action query time = 3.213 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5829 +t=58: Selected seed 195 with value = 0.5829 +Query 1/1: Action query time = 4.978 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7167 +t=74: Selected seed 195 with value = 0.7167 +Query 1/1: Action query time = 5.288 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7847 +t=90: Selected seed 195 with value = 0.7847 +Query 1/1: Action query time = 5.100 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9754 +t=106: Selected seed 195 with value = 0.9754 +Query 1/1: Action query time = 4.425 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.492 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=138: Selected seed 195 with value = 0.9986 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=4--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 5... +Query 1/1: Action query time = 5.378 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3927 +t=10: Selected seed 195 with value = 0.3927 +Query 1/1: Action query time = 5.126 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4491 +t=26: Selected seed 195 with value = 0.4491 +Query 1/1: Action query time = 3.627 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5355 +t=42: Selected seed 195 with value = 0.5355 +Query 1/1: Action query time = 4.523 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6097 +t=58: Selected seed 195 with value = 0.6097 +Query 1/1: Action query time = 4.995 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6822 +t=74: Selected seed 195 with value = 0.6822 +Query 1/1: Action query time = 4.825 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8376 +t=90: Selected seed 195 with value = 0.8376 +Query 1/1: Action query time = 4.746 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=106: Selected seed 195 with value = 0.9867 +Query 1/1: Action query time = 5.288 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=5--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 6... +Query 1/1: Action query time = 3.673 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3843 +t=10: Selected seed 195 with value = 0.3843 +Query 1/1: Action query time = 4.329 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4778 +t=26: Selected seed 195 with value = 0.4778 +Query 1/1: Action query time = 4.912 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5726 +t=42: Selected seed 195 with value = 0.5726 +Query 1/1: Action query time = 4.258 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6464 +t=58: Selected seed 195 with value = 0.6464 +Query 1/1: Action query time = 4.084 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7350 +t=74: Selected seed 195 with value = 0.7350 +Query 1/1: Action query time = 5.193 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8849 +t=90: Selected seed 195 with value = 0.8849 +Query 1/1: Action query time = 5.062 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9811 +t=106: Selected seed 195 with value = 0.9811 +Query 1/1: Action query time = 5.276 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.171 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=6--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 7... +Query 1/1: Action query time = 3.671 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3651 +t=10: Selected seed 195 with value = 0.3651 +Query 1/1: Action query time = 4.820 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4172 +t=26: Selected seed 195 with value = 0.4172 +Query 1/1: Action query time = 5.211 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4848 +t=42: Selected seed 195 with value = 0.4848 +Query 1/1: Action query time = 4.367 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6340 +t=58: Selected seed 195 with value = 0.6340 +Query 1/1: Action query time = 4.933 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6880 +t=74: Selected seed 195 with value = 0.6880 +Query 1/1: Action query time = 3.993 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8159 +t=90: Selected seed 195 with value = 0.8159 +Query 1/1: Action query time = 5.269 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.959 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.679 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9993 +t=138: Selected seed 195 with value = 0.9993 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=7--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 8... +Query 1/1: Action query time = 3.328 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3894 +t=10: Selected seed 195 with value = 0.3894 +Query 1/1: Action query time = 4.324 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4431 +t=26: Selected seed 195 with value = 0.4431 +Query 1/1: Action query time = 3.748 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5061 +t=42: Selected seed 195 with value = 0.5061 +Query 1/1: Action query time = 5.027 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6396 +t=58: Selected seed 195 with value = 0.6396 +Query 1/1: Action query time = 5.133 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7436 +t=74: Selected seed 195 with value = 0.7436 +Query 1/1: Action query time = 4.633 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8912 +t=90: Selected seed 195 with value = 0.8912 +Query 1/1: Action query time = 5.382 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.359 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=8--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 9... +Query 1/1: Action query time = 4.110 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3817 +t=10: Selected seed 195 with value = 0.3817 +Query 1/1: Action query time = 4.491 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4135 +t=26: Selected seed 195 with value = 0.4135 +Query 1/1: Action query time = 5.248 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5157 +t=42: Selected seed 195 with value = 0.5157 +Query 1/1: Action query time = 4.179 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5624 +t=58: Selected seed 195 with value = 0.5624 +Query 1/1: Action query time = 5.512 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6780 +t=74: Selected seed 195 with value = 0.6780 +Query 1/1: Action query time = 4.935 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7524 +t=90: Selected seed 195 with value = 0.7524 +Query 1/1: Action query time = 4.970 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8389 +t=106: Selected seed 195 with value = 0.8389 +Query 1/1: Action query time = 4.269 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9862 +t=122: Selected seed 195 with value = 0.9862 +Query 1/1: Action query time = 4.358 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=9--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 10... +Query 1/1: Action query time = 3.226 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3997 +t=10: Selected seed 195 with value = 0.3997 +Query 1/1: Action query time = 3.080 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4371 +t=26: Selected seed 195 with value = 0.4371 +Query 1/1: Action query time = 2.764 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5694 +t=42: Selected seed 195 with value = 0.5694 +Query 1/1: Action query time = 2.112 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5888 +t=58: Selected seed 195 with value = 0.5888 +Query 1/1: Action query time = 2.307 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6962 +t=74: Selected seed 195 with value = 0.6962 +Query 1/1: Action query time = 2.886 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8885 +t=90: Selected seed 195 with value = 0.8885 +Query 1/1: Action query time = 3.346 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=106: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 3.582 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=10--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 11... +Query 1/1: Action query time = 2.491 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3987 +t=10: Selected seed 195 with value = 0.3987 +Query 1/1: Action query time = 1.847 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4483 +t=26: Selected seed 195 with value = 0.4483 +Query 1/1: Action query time = 1.650 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5576 +t=42: Selected seed 195 with value = 0.5576 +Query 1/1: Action query time = 1.679 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6446 +t=58: Selected seed 195 with value = 0.6446 +Query 1/1: Action query time = 1.869 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7223 +t=74: Selected seed 195 with value = 0.7223 +Query 1/1: Action query time = 2.574 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8313 +t=90: Selected seed 195 with value = 0.8313 +Query 1/1: Action query time = 2.766 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9763 +t=106: Selected seed 195 with value = 0.9763 +Query 1/1: Action query time = 2.679 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.532 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=11--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 12... +Query 1/1: Action query time = 2.732 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3611 +t=10: Selected seed 195 with value = 0.3611 +Query 1/1: Action query time = 2.363 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4330 +t=26: Selected seed 195 with value = 0.4330 +Query 1/1: Action query time = 2.532 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5075 +t=42: Selected seed 195 with value = 0.5075 +Query 1/1: Action query time = 1.442 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5296 +t=58: Selected seed 195 with value = 0.5296 +Query 1/1: Action query time = 1.350 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6704 +t=74: Selected seed 195 with value = 0.6704 +Query 1/1: Action query time = 2.423 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7471 +t=90: Selected seed 195 with value = 0.7471 +Query 1/1: Action query time = 2.267 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8567 +t=106: Selected seed 195 with value = 0.8567 +Query 1/1: Action query time = 2.272 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9654 +t=122: Selected seed 195 with value = 0.9654 +Query 1/1: Action query time = 2.232 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9947 +t=138: Selected seed 195 with value = 0.9947 +Saved rollout MP4 at path ./rollouts/v2i175_t5_s3/2026_08_03-16_07_39--with_future_img--episode=12--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 12 +Total successes: 12 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_03--v2i175_t7_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_03--v2i175_t7_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..b48aa43a878c0e5b460c8fd8e31329d27e64d5de --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_03--v2i175_t7_s0.txt @@ -0,0 +1,213 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i175_t7_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,8,16,24,32,40,48', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 2.283 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4778 +t=10: Selected seed 195 with value = 0.4778 +Query 1/1: Action query time = 2.269 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5362 +t=26: Selected seed 195 with value = 0.5362 +Query 1/1: Action query time = 2.789 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6369 +t=42: Selected seed 195 with value = 0.6369 +Query 1/1: Action query time = 1.683 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7261 +t=58: Selected seed 195 with value = 0.7261 +Query 1/1: Action query time = 2.350 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8289 +t=74: Selected seed 195 with value = 0.8289 +Query 1/1: Action query time = 1.841 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9703 +t=90: Selected seed 195 with value = 0.9703 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s0/2026_08_03-16_22_03--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 1.599 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5122 +t=10: Selected seed 195 with value = 0.5122 +Query 1/1: Action query time = 2.743 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6403 +t=26: Selected seed 195 with value = 0.6403 +Query 1/1: Action query time = 2.834 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6684 +t=42: Selected seed 195 with value = 0.6684 +Query 1/1: Action query time = 3.168 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7944 +t=58: Selected seed 195 with value = 0.7944 +Query 1/1: Action query time = 1.723 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9493 +t=74: Selected seed 195 with value = 0.9493 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s0/2026_08_03-16_22_03--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 1.705 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4592 +t=10: Selected seed 195 with value = 0.4592 +Query 1/1: Action query time = 2.682 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5471 +t=26: Selected seed 195 with value = 0.5471 +Query 1/1: Action query time = 2.759 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6184 +t=42: Selected seed 195 with value = 0.6184 +Query 1/1: Action query time = 2.899 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7109 +t=58: Selected seed 195 with value = 0.7109 +Query 1/1: Action query time = 2.255 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8678 +t=74: Selected seed 195 with value = 0.8678 +Query 1/1: Action query time = 1.653 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9939 +t=90: Selected seed 195 with value = 0.9939 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s0/2026_08_03-16_22_03--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 2.889 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4602 +t=10: Selected seed 195 with value = 0.4602 +Query 1/1: Action query time = 3.060 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5425 +t=26: Selected seed 195 with value = 0.5425 +Query 1/1: Action query time = 2.278 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6273 +t=42: Selected seed 195 with value = 0.6273 +Query 1/1: Action query time = 2.304 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6899 +t=58: Selected seed 195 with value = 0.6899 +Query 1/1: Action query time = 1.341 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8594 +t=74: Selected seed 195 with value = 0.8594 +Query 1/1: Action query time = 1.012 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9909 +t=90: Selected seed 195 with value = 0.9909 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s0/2026_08_03-16_22_03--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 2.324 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4917 +t=10: Selected seed 195 with value = 0.4917 +Query 1/1: Action query time = 2.575 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5684 +t=26: Selected seed 195 with value = 0.5684 +Query 1/1: Action query time = 2.337 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6615 +t=42: Selected seed 195 with value = 0.6615 +Query 1/1: Action query time = 1.866 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7926 +t=58: Selected seed 195 with value = 0.7926 +Query 1/1: Action query time = 1.786 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9473 +t=74: Selected seed 195 with value = 0.9473 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s0/2026_08_03-16_22_03--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 1.849 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4770 +t=10: Selected seed 195 with value = 0.4770 +Query 1/1: Action query time = 2.303 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5655 +t=26: Selected seed 195 with value = 0.5655 +Query 1/1: Action query time = 2.361 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6650 +t=42: Selected seed 195 with value = 0.6650 +Query 1/1: Action query time = 2.296 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7702 +t=58: Selected seed 195 with value = 0.7702 +Query 1/1: Action query time = 1.811 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9269 +t=74: Selected seed 195 with value = 0.9269 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s0/2026_08_03-16_22_03--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: turn on the stove +Starting episode 7... +Query 1/1: Action query time = 1.366 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5105 +t=10: Selected seed 195 with value = 0.5105 +Query 1/1: Action query time = 1.354 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5932 +t=26: Selected seed 195 with value = 0.5932 +Query 1/1: Action query time = 1.393 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7013 +t=42: Selected seed 195 with value = 0.7013 +Query 1/1: Action query time = 1.449 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8827 +t=58: Selected seed 195 with value = 0.8827 +Query 1/1: Action query time = 1.463 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=74: Selected seed 195 with value = 0.9903 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s0/2026_08_03-16_22_03--with_future_img--episode=7--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 7 +Total successes: 7 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_03--v2i175_t7_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_03--v2i175_t7_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..d2397f374017040f1bf54bce92d87826c9305fd7 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_03--v2i175_t7_s3.txt @@ -0,0 +1,186 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i175_t7_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,11,19,27,35,43', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 2.021 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4621 +t=10: Selected seed 195 with value = 0.4621 +Query 1/1: Action query time = 2.279 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5801 +t=26: Selected seed 195 with value = 0.5801 +Query 1/1: Action query time = 2.060 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6777 +t=42: Selected seed 195 with value = 0.6777 +Query 1/1: Action query time = 2.516 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7720 +t=58: Selected seed 195 with value = 0.7720 +Query 1/1: Action query time = 2.935 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9288 +t=74: Selected seed 195 with value = 0.9288 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s3/2026_08_03-16_22_03--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 1.818 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4960 +t=10: Selected seed 195 with value = 0.4960 +Query 1/1: Action query time = 1.660 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5762 +t=26: Selected seed 195 with value = 0.5762 +Query 1/1: Action query time = 2.843 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6560 +t=42: Selected seed 195 with value = 0.6560 +Query 1/1: Action query time = 2.922 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7875 +t=58: Selected seed 195 with value = 0.7875 +Query 1/1: Action query time = 2.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9208 +t=74: Selected seed 195 with value = 0.9208 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s3/2026_08_03-16_22_03--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 0.991 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5006 +t=10: Selected seed 195 with value = 0.5006 +Query 1/1: Action query time = 1.548 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5825 +t=26: Selected seed 195 with value = 0.5825 +Query 1/1: Action query time = 2.831 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6581 +t=42: Selected seed 195 with value = 0.6581 +Query 1/1: Action query time = 2.178 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7685 +t=58: Selected seed 195 with value = 0.7685 +Query 1/1: Action query time = 1.692 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9063 +t=74: Selected seed 195 with value = 0.9063 +Query 1/1: Action query time = 2.409 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9959 +t=90: Selected seed 195 with value = 0.9959 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s3/2026_08_03-16_22_03--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 1.359 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4838 +t=10: Selected seed 195 with value = 0.4838 +Query 1/1: Action query time = 1.628 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5879 +t=26: Selected seed 195 with value = 0.5879 +Query 1/1: Action query time = 2.671 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6514 +t=42: Selected seed 195 with value = 0.6514 +Query 1/1: Action query time = 1.785 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7701 +t=58: Selected seed 195 with value = 0.7701 +Query 1/1: Action query time = 2.441 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7087 +t=74: Selected seed 195 with value = 0.7087 +Query 1/1: Action query time = 2.556 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8473 +t=90: Selected seed 195 with value = 0.8473 +Query 1/1: Action query time = 2.249 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9788 +t=106: Selected seed 195 with value = 0.9788 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s3/2026_08_03-16_22_03--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 2.811 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5200 +t=10: Selected seed 195 with value = 0.5200 +Query 1/1: Action query time = 2.471 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5872 +t=26: Selected seed 195 with value = 0.5872 +Query 1/1: Action query time = 2.320 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7081 +t=42: Selected seed 195 with value = 0.7081 +Query 1/1: Action query time = 2.861 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8720 +t=58: Selected seed 195 with value = 0.8720 +Query 1/1: Action query time = 1.841 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=74: Selected seed 195 with value = 0.9960 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s3/2026_08_03-16_22_03--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 1.482 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4780 +t=10: Selected seed 195 with value = 0.4780 +Query 1/1: Action query time = 2.764 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5588 +t=26: Selected seed 195 with value = 0.5588 +Query 1/1: Action query time = 2.608 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6483 +t=42: Selected seed 195 with value = 0.6483 +Query 1/1: Action query time = 2.489 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7392 +t=58: Selected seed 195 with value = 0.7392 +Query 1/1: Action query time = 1.738 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8902 +t=74: Selected seed 195 with value = 0.8902 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s3/2026_08_03-16_22_03--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 6 +Total successes: 6 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_03--v2i175_t7_s7.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_03--v2i175_t7_s7.txt new file mode 100644 index 0000000000000000000000000000000000000000..2ed7e756d40ac83873fae409d0779f5d7e9f93ed --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_03--v2i175_t7_s7.txt @@ -0,0 +1,178 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i175_t7_s7', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='7,15,23,31,39,47', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 2.546 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4857 +t=10: Selected seed 195 with value = 0.4857 +Query 1/1: Action query time = 2.315 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5482 +t=26: Selected seed 195 with value = 0.5482 +Query 1/1: Action query time = 3.000 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6472 +t=42: Selected seed 195 with value = 0.6472 +Query 1/1: Action query time = 2.602 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7555 +t=58: Selected seed 195 with value = 0.7555 +Query 1/1: Action query time = 2.457 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9015 +t=74: Selected seed 195 with value = 0.9015 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s7/2026_08_03-16_22_03--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 1.114 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5240 +t=10: Selected seed 195 with value = 0.5240 +Query 1/1: Action query time = 2.247 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5692 +t=26: Selected seed 195 with value = 0.5692 +Query 1/1: Action query time = 2.571 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6564 +t=42: Selected seed 195 with value = 0.6564 +Query 1/1: Action query time = 2.379 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7815 +t=58: Selected seed 195 with value = 0.7815 +Query 1/1: Action query time = 2.471 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9156 +t=74: Selected seed 195 with value = 0.9156 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s7/2026_08_03-16_22_03--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 1.130 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4775 +t=10: Selected seed 195 with value = 0.4775 +Query 1/1: Action query time = 1.542 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5533 +t=26: Selected seed 195 with value = 0.5533 +Query 1/1: Action query time = 2.766 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6361 +t=42: Selected seed 195 with value = 0.6361 +Query 1/1: Action query time = 2.502 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7379 +t=58: Selected seed 195 with value = 0.7379 +Query 1/1: Action query time = 2.447 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8732 +t=74: Selected seed 195 with value = 0.8732 +Query 1/1: Action query time = 2.231 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9941 +t=90: Selected seed 195 with value = 0.9941 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s7/2026_08_03-16_22_03--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 1.359 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5177 +t=10: Selected seed 195 with value = 0.5177 +Query 1/1: Action query time = 1.646 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5882 +t=26: Selected seed 195 with value = 0.5882 +Query 1/1: Action query time = 2.559 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6809 +t=42: Selected seed 195 with value = 0.6809 +Query 1/1: Action query time = 2.966 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8004 +t=58: Selected seed 195 with value = 0.8004 +Query 1/1: Action query time = 2.551 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9429 +t=74: Selected seed 195 with value = 0.9429 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s7/2026_08_03-16_22_03--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 1.816 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5230 +t=10: Selected seed 195 with value = 0.5230 +Query 1/1: Action query time = 2.993 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5905 +t=26: Selected seed 195 with value = 0.5905 +Query 1/1: Action query time = 2.154 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6867 +t=42: Selected seed 195 with value = 0.6867 +Query 1/1: Action query time = 2.007 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7949 +t=58: Selected seed 195 with value = 0.7949 +Query 1/1: Action query time = 1.960 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9324 +t=74: Selected seed 195 with value = 0.9324 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s7/2026_08_03-16_22_03--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 1.498 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5381 +t=10: Selected seed 195 with value = 0.5381 +Query 1/1: Action query time = 2.216 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6198 +t=26: Selected seed 195 with value = 0.6198 +Query 1/1: Action query time = 2.328 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7245 +t=42: Selected seed 195 with value = 0.7245 +Query 1/1: Action query time = 2.614 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8582 +t=58: Selected seed 195 with value = 0.8582 +Query 1/1: Action query time = 2.600 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=74: Selected seed 195 with value = 0.9912 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s7/2026_08_03-16_22_03--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 6 +Total successes: 6 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_04--v2i175_t7_s5.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_04--v2i175_t7_s5.txt new file mode 100644 index 0000000000000000000000000000000000000000..e3087bff64e959b856e4e8571d77aa86db6195aa --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-16_22_04--v2i175_t7_s5.txt @@ -0,0 +1,178 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000175/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i175_t7_s5', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='5,13,21,29,37,45', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 2.280 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4561 +t=10: Selected seed 195 with value = 0.4561 +Query 1/1: Action query time = 2.348 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5666 +t=26: Selected seed 195 with value = 0.5666 +Query 1/1: Action query time = 2.931 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6449 +t=42: Selected seed 195 with value = 0.6449 +Query 1/1: Action query time = 2.637 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7454 +t=58: Selected seed 195 with value = 0.7454 +Query 1/1: Action query time = 2.595 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8984 +t=74: Selected seed 195 with value = 0.8984 +Query 1/1: Action query time = 1.577 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=90: Selected seed 195 with value = 0.9968 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s5/2026_08_03-16_22_04--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 2.674 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5278 +t=10: Selected seed 195 with value = 0.5278 +Query 1/1: Action query time = 2.621 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6299 +t=26: Selected seed 195 with value = 0.6299 +Query 1/1: Action query time = 2.268 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7000 +t=42: Selected seed 195 with value = 0.7000 +Query 1/1: Action query time = 2.346 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8020 +t=58: Selected seed 195 with value = 0.8020 +Query 1/1: Action query time = 1.517 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9343 +t=74: Selected seed 195 with value = 0.9343 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s5/2026_08_03-16_22_04--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 2.046 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5188 +t=10: Selected seed 195 with value = 0.5188 +Query 1/1: Action query time = 2.123 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6040 +t=26: Selected seed 195 with value = 0.6040 +Query 1/1: Action query time = 2.708 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6791 +t=42: Selected seed 195 with value = 0.6791 +Query 1/1: Action query time = 2.380 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7883 +t=58: Selected seed 195 with value = 0.7883 +Query 1/1: Action query time = 2.016 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9690 +t=74: Selected seed 195 with value = 0.9690 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s5/2026_08_03-16_22_04--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 1.762 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5431 +t=10: Selected seed 195 with value = 0.5431 +Query 1/1: Action query time = 2.404 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6287 +t=26: Selected seed 195 with value = 0.6287 +Query 1/1: Action query time = 3.062 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7202 +t=42: Selected seed 195 with value = 0.7202 +Query 1/1: Action query time = 2.499 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8071 +t=58: Selected seed 195 with value = 0.8071 +Query 1/1: Action query time = 1.412 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9582 +t=74: Selected seed 195 with value = 0.9582 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s5/2026_08_03-16_22_04--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 1.937 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4822 +t=10: Selected seed 195 with value = 0.4822 +Query 1/1: Action query time = 2.400 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5875 +t=26: Selected seed 195 with value = 0.5875 +Query 1/1: Action query time = 2.319 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6455 +t=42: Selected seed 195 with value = 0.6455 +Query 1/1: Action query time = 2.390 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8787 +t=58: Selected seed 195 with value = 0.8787 +Query 1/1: Action query time = 1.611 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=74: Selected seed 195 with value = 0.9985 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s5/2026_08_03-16_22_04--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 2.124 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4764 +t=10: Selected seed 195 with value = 0.4764 +Query 1/1: Action query time = 2.590 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5671 +t=26: Selected seed 195 with value = 0.5671 +Query 1/1: Action query time = 2.590 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6664 +t=42: Selected seed 195 with value = 0.6664 +Query 1/1: Action query time = 2.343 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7808 +t=58: Selected seed 195 with value = 0.7808 +Query 1/1: Action query time = 1.954 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9652 +t=74: Selected seed 195 with value = 0.9652 +Saved rollout MP4 at path ./rollouts/v2i175_t7_s5/2026_08_03-16_22_04--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 6 +Total successes: 6 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_48--v2i350_p1_t1_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_48--v2i350_p1_t1_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..4d16ad81385b980735c9bc8f3e82f215206780e1 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_48--v2i350_p1_t1_s1.txt @@ -0,0 +1,571 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p1_t1_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 4.420 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5690 +t=10: Selected seed 195 with value = 0.5690 +Query 1/1: Action query time = 5.822 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6717 +t=26: Selected seed 195 with value = 0.6717 +Query 1/1: Action query time = 4.101 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7835 +t=42: Selected seed 195 with value = 0.7835 +Query 1/1: Action query time = 4.420 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8991 +t=58: Selected seed 195 with value = 0.8991 +Query 1/1: Action query time = 5.332 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9653 +t=74: Selected seed 195 with value = 0.9653 +Query 1/1: Action query time = 5.082 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9484 +t=90: Selected seed 195 with value = 0.9484 +Query 1/1: Action query time = 3.340 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9565 +t=106: Selected seed 195 with value = 0.9565 +Query 1/1: Action query time = 3.783 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9183 +t=122: Selected seed 195 with value = 0.9183 +Query 1/1: Action query time = 4.233 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9155 +t=138: Selected seed 195 with value = 0.9155 +Query 1/1: Action query time = 4.531 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9120 +t=154: Selected seed 195 with value = 0.9120 +Query 1/1: Action query time = 5.115 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9329 +t=170: Selected seed 195 with value = 0.9329 +Query 1/1: Action query time = 4.910 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9733 +t=186: Selected seed 195 with value = 0.9733 +Query 1/1: Action query time = 5.233 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9892 +t=202: Selected seed 195 with value = 0.9892 +Query 1/1: Action query time = 2.579 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9952 +t=218: Selected seed 195 with value = 0.9952 +Query 1/1: Action query time = 4.297 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=234: Selected seed 195 with value = 0.9958 +Query 1/1: Action query time = 4.976 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9919 +t=250: Selected seed 195 with value = 0.9919 +Query 1/1: Action query time = 4.959 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9862 +t=266: Selected seed 195 with value = 0.9862 +Query 1/1: Action query time = 3.904 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9708 +t=282: Selected seed 195 with value = 0.9708 +Query 1/1: Action query time = 4.387 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9649 +t=298: Selected seed 195 with value = 0.9649 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 5.302 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6207 +t=10: Selected seed 195 with value = 0.6207 +Query 1/1: Action query time = 5.107 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7317 +t=26: Selected seed 195 with value = 0.7317 +Query 1/1: Action query time = 4.893 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8036 +t=42: Selected seed 195 with value = 0.8036 +Query 1/1: Action query time = 5.131 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9027 +t=58: Selected seed 195 with value = 0.9027 +Query 1/1: Action query time = 4.552 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9967 +t=74: Selected seed 195 with value = 0.9967 +Query 1/1: Action query time = 3.968 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 1 (50.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.540 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5727 +t=10: Selected seed 195 with value = 0.5727 +Query 1/1: Action query time = 5.018 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7809 +t=26: Selected seed 195 with value = 0.7809 +Query 1/1: Action query time = 4.786 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7962 +t=42: Selected seed 195 with value = 0.7962 +Query 1/1: Action query time = 5.027 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8946 +t=58: Selected seed 195 with value = 0.8946 +Query 1/1: Action query time = 4.963 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.396 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 3.746 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5747 +t=10: Selected seed 195 with value = 0.5747 +Query 1/1: Action query time = 4.526 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6853 +t=26: Selected seed 195 with value = 0.6853 +Query 1/1: Action query time = 5.977 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7889 +t=42: Selected seed 195 with value = 0.7889 +Query 1/1: Action query time = 5.335 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8762 +t=58: Selected seed 195 with value = 0.8762 +Query 1/1: Action query time = 4.466 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9790 +t=74: Selected seed 195 with value = 0.9790 +Query 1/1: Action query time = 4.339 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9583 +t=90: Selected seed 195 with value = 0.9583 +Query 1/1: Action query time = 4.076 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9609 +t=106: Selected seed 195 with value = 0.9609 +Query 1/1: Action query time = 4.166 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9112 +t=122: Selected seed 195 with value = 0.9112 +Query 1/1: Action query time = 4.525 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8932 +t=138: Selected seed 195 with value = 0.8932 +Query 1/1: Action query time = 3.580 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9085 +t=154: Selected seed 195 with value = 0.9085 +Query 1/1: Action query time = 5.043 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9415 +t=170: Selected seed 195 with value = 0.9415 +Query 1/1: Action query time = 5.013 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9813 +t=186: Selected seed 195 with value = 0.9813 +Query 1/1: Action query time = 4.554 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9668 +t=202: Selected seed 195 with value = 0.9668 +Query 1/1: Action query time = 5.826 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9761 +t=218: Selected seed 195 with value = 0.9761 +Query 1/1: Action query time = 5.173 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9846 +t=234: Selected seed 195 with value = 0.9846 +Query 1/1: Action query time = 5.897 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9937 +t=250: Selected seed 195 with value = 0.9937 +Query 1/1: Action query time = 5.127 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9832 +t=266: Selected seed 195 with value = 0.9832 +Query 1/1: Action query time = 4.760 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9868 +t=282: Selected seed 195 with value = 0.9868 +Query 1/1: Action query time = 4.368 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9802 +t=298: Selected seed 195 with value = 0.9802 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 4 +# successes: 2 (50.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 4.316 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5932 +t=10: Selected seed 195 with value = 0.5932 +Query 1/1: Action query time = 4.595 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7381 +t=26: Selected seed 195 with value = 0.7381 +Query 1/1: Action query time = 3.992 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7631 +t=42: Selected seed 195 with value = 0.7631 +Query 1/1: Action query time = 4.906 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8967 +t=58: Selected seed 195 with value = 0.8967 +Query 1/1: Action query time = 5.201 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.317 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=5--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 3 (60.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 4.973 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5943 +t=10: Selected seed 195 with value = 0.5943 +Query 1/1: Action query time = 4.116 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6514 +t=26: Selected seed 195 with value = 0.6514 +Query 1/1: Action query time = 4.774 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7641 +t=42: Selected seed 195 with value = 0.7641 +Query 1/1: Action query time = 4.370 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8652 +t=58: Selected seed 195 with value = 0.8652 +Query 1/1: Action query time = 4.136 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9962 +t=74: Selected seed 195 with value = 0.9962 +Query 1/1: Action query time = 4.202 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.385 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.284 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.540 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8473 +t=138: Selected seed 195 with value = 0.8473 +Query 1/1: Action query time = 5.484 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9151 +t=154: Selected seed 195 with value = 0.9151 +Query 1/1: Action query time = 3.496 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9708 +t=170: Selected seed 195 with value = 0.9708 +Query 1/1: Action query time = 4.125 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9705 +t=186: Selected seed 195 with value = 0.9705 +Query 1/1: Action query time = 5.168 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9687 +t=202: Selected seed 195 with value = 0.9687 +Query 1/1: Action query time = 4.874 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9700 +t=218: Selected seed 195 with value = 0.9700 +Query 1/1: Action query time = 5.184 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9677 +t=234: Selected seed 195 with value = 0.9677 +Query 1/1: Action query time = 5.182 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9564 +t=250: Selected seed 195 with value = 0.9564 +Query 1/1: Action query time = 4.222 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8721 +t=266: Selected seed 195 with value = 0.8721 +Query 1/1: Action query time = 5.077 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8463 +t=282: Selected seed 195 with value = 0.8463 +Query 1/1: Action query time = 4.848 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8426 +t=298: Selected seed 195 with value = 0.8426 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=6--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 6 +# successes: 3 (50.0%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 3.643 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6148 +t=10: Selected seed 195 with value = 0.6148 +Query 1/1: Action query time = 4.971 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7314 +t=26: Selected seed 195 with value = 0.7314 +Query 1/1: Action query time = 5.522 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7687 +t=42: Selected seed 195 with value = 0.7687 +Query 1/1: Action query time = 5.610 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8919 +t=58: Selected seed 195 with value = 0.8919 +Query 1/1: Action query time = 3.025 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9077 +t=74: Selected seed 195 with value = 0.9077 +Query 1/1: Action query time = 5.475 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9807 +t=90: Selected seed 195 with value = 0.9807 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=7--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 4 (57.1%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 3.972 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6225 +t=10: Selected seed 195 with value = 0.6225 +Query 1/1: Action query time = 3.450 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6672 +t=26: Selected seed 195 with value = 0.6672 +Query 1/1: Action query time = 3.630 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8108 +t=42: Selected seed 195 with value = 0.8108 +Query 1/1: Action query time = 4.475 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9325 +t=58: Selected seed 195 with value = 0.9325 +Query 1/1: Action query time = 4.246 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9956 +t=74: Selected seed 195 with value = 0.9956 +Query 1/1: Action query time = 3.905 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=8--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 5 (62.5%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 2.407 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5374 +t=10: Selected seed 195 with value = 0.5374 +Query 1/1: Action query time = 4.155 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6850 +t=26: Selected seed 195 with value = 0.6850 +Query 1/1: Action query time = 3.085 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7717 +t=42: Selected seed 195 with value = 0.7717 +Query 1/1: Action query time = 2.002 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8935 +t=58: Selected seed 195 with value = 0.8935 +Query 1/1: Action query time = 2.337 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9447 +t=74: Selected seed 195 with value = 0.9447 +Query 1/1: Action query time = 3.136 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=9--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 6 (66.7%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 2.429 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5588 +t=10: Selected seed 195 with value = 0.5588 +Query 1/1: Action query time = 2.682 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6592 +t=26: Selected seed 195 with value = 0.6592 +Query 1/1: Action query time = 2.274 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7901 +t=42: Selected seed 195 with value = 0.7901 +Query 1/1: Action query time = 2.417 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9108 +t=58: Selected seed 195 with value = 0.9108 +Query 1/1: Action query time = 1.626 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=74: Selected seed 195 with value = 0.9987 +Query 1/1: Action query time = 1.928 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=10--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 7 (70.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 2.192 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5720 +t=10: Selected seed 195 with value = 0.5720 +Query 1/1: Action query time = 2.316 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6478 +t=26: Selected seed 195 with value = 0.6478 +Query 1/1: Action query time = 1.879 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7588 +t=42: Selected seed 195 with value = 0.7588 +Query 1/1: Action query time = 1.637 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8297 +t=58: Selected seed 195 with value = 0.8297 +Query 1/1: Action query time = 1.278 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9233 +t=74: Selected seed 195 with value = 0.9233 +Query 1/1: Action query time = 2.157 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9879 +t=90: Selected seed 195 with value = 0.9879 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=11--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 8 (72.7%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 1.456 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5799 +t=10: Selected seed 195 with value = 0.5799 +Query 1/1: Action query time = 1.015 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7421 +t=26: Selected seed 195 with value = 0.7421 +Query 1/1: Action query time = 1.282 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7711 +t=42: Selected seed 195 with value = 0.7711 +Query 1/1: Action query time = 1.484 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8578 +t=58: Selected seed 195 with value = 0.8578 +Query 1/1: Action query time = 1.227 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=74: Selected seed 195 with value = 0.9999 +Query 1/1: Action query time = 1.355 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=12--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 9 (75.0%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 1.933 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6239 +t=10: Selected seed 195 with value = 0.6239 +Query 1/1: Action query time = 1.673 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7303 +t=26: Selected seed 195 with value = 0.7303 +Query 1/1: Action query time = 1.004 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8300 +t=42: Selected seed 195 with value = 0.8300 +Query 1/1: Action query time = 0.964 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9551 +t=58: Selected seed 195 with value = 0.9551 +Query 1/1: Action query time = 0.991 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.303 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s1/2026_08_03-17_25_48--with_future_img--episode=13--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 13 +# successes: 10 (76.9%) +Current task success rate: 0.7692307692307693 +Current total success rate: 0.7692307692307693 +Final results: +Total episodes: 13 +Total successes: 10 +Overall success rate: 0.7692 (76.9%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_48--v2i350_p1_t1_s2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_48--v2i350_p1_t1_s2.txt new file mode 100644 index 0000000000000000000000000000000000000000..dcf6df44bb5d04b40a732cef12bc39fb244de205 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_48--v2i350_p1_t1_s2.txt @@ -0,0 +1,436 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='1', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p1_t1_s2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,6,10,14,18,22,26,30,34,38,42,46', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [1] +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 4.165 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5815 +t=10: Selected seed 195 with value = 0.5815 +Query 1/1: Action query time = 5.267 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7610 +t=26: Selected seed 195 with value = 0.7610 +Query 1/1: Action query time = 4.482 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7927 +t=42: Selected seed 195 with value = 0.7927 +Query 1/1: Action query time = 4.339 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8830 +t=58: Selected seed 195 with value = 0.8830 +Query 1/1: Action query time = 5.137 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.094 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=1--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 3.933 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5756 +t=10: Selected seed 195 with value = 0.5756 +Query 1/1: Action query time = 4.334 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7134 +t=26: Selected seed 195 with value = 0.7134 +Query 1/1: Action query time = 4.486 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7505 +t=42: Selected seed 195 with value = 0.7505 +Query 1/1: Action query time = 5.366 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8882 +t=58: Selected seed 195 with value = 0.8882 +Query 1/1: Action query time = 4.833 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=74: Selected seed 195 with value = 0.9966 +Query 1/1: Action query time = 5.190 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=2--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 5.202 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5113 +t=10: Selected seed 195 with value = 0.5113 +Query 1/1: Action query time = 5.088 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6414 +t=26: Selected seed 195 with value = 0.6414 +Query 1/1: Action query time = 4.417 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7618 +t=42: Selected seed 195 with value = 0.7618 +Query 1/1: Action query time = 3.792 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8909 +t=58: Selected seed 195 with value = 0.8909 +Query 1/1: Action query time = 4.395 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.881 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=3--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 4.445 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5750 +t=10: Selected seed 195 with value = 0.5750 +Query 1/1: Action query time = 4.375 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6983 +t=26: Selected seed 195 with value = 0.6983 +Query 1/1: Action query time = 5.231 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8166 +t=42: Selected seed 195 with value = 0.8166 +Query 1/1: Action query time = 5.285 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9570 +t=58: Selected seed 195 with value = 0.9570 +Query 1/1: Action query time = 4.814 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.497 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=4--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 3.001 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6525 +t=10: Selected seed 195 with value = 0.6525 +Query 1/1: Action query time = 5.086 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8190 +t=26: Selected seed 195 with value = 0.8190 +Query 1/1: Action query time = 4.310 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7728 +t=42: Selected seed 195 with value = 0.7728 +Query 1/1: Action query time = 4.991 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8479 +t=58: Selected seed 195 with value = 0.8479 +Query 1/1: Action query time = 5.446 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.837 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=5--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 3.801 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6706 +t=10: Selected seed 195 with value = 0.6706 +Query 1/1: Action query time = 3.309 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8489 +t=26: Selected seed 195 with value = 0.8489 +Query 1/1: Action query time = 4.410 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7599 +t=42: Selected seed 195 with value = 0.7599 +Query 1/1: Action query time = 3.876 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9186 +t=58: Selected seed 195 with value = 0.9186 +Query 1/1: Action query time = 4.269 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.396 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=6--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 4.820 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5712 +t=10: Selected seed 195 with value = 0.5712 +Query 1/1: Action query time = 4.851 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6783 +t=26: Selected seed 195 with value = 0.6783 +Query 1/1: Action query time = 5.238 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7765 +t=42: Selected seed 195 with value = 0.7765 +Query 1/1: Action query time = 4.621 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8308 +t=58: Selected seed 195 with value = 0.8308 +Query 1/1: Action query time = 5.414 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9423 +t=74: Selected seed 195 with value = 0.9423 +Query 1/1: Action query time = 4.172 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=7--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 4.195 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5497 +t=10: Selected seed 195 with value = 0.5497 +Query 1/1: Action query time = 4.916 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6302 +t=26: Selected seed 195 with value = 0.6302 +Query 1/1: Action query time = 5.062 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7498 +t=42: Selected seed 195 with value = 0.7498 +Query 1/1: Action query time = 5.264 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8430 +t=58: Selected seed 195 with value = 0.8430 +Query 1/1: Action query time = 4.086 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9234 +t=74: Selected seed 195 with value = 0.9234 +Query 1/1: Action query time = 4.900 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.641 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=106: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 3.620 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.882 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.222 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.496 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9816 +t=170: Selected seed 195 with value = 0.9816 +Query 1/1: Action query time = 4.666 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.092 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9972 +t=202: Selected seed 195 with value = 0.9972 +Query 1/1: Action query time = 5.093 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.087 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.465 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9491 +t=250: Selected seed 195 with value = 0.9491 +Query 1/1: Action query time = 4.918 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.005 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.224 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=8--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 8 +# successes: 7 (87.5%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 4.071 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5860 +t=10: Selected seed 195 with value = 0.5860 +Query 1/1: Action query time = 5.578 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6809 +t=26: Selected seed 195 with value = 0.6809 +Query 1/1: Action query time = 4.969 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7809 +t=42: Selected seed 195 with value = 0.7809 +Query 1/1: Action query time = 4.673 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8843 +t=58: Selected seed 195 with value = 0.8843 +Query 1/1: Action query time = 4.237 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.225 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=9--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 8 (88.9%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 4.496 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6026 +t=10: Selected seed 195 with value = 0.6026 +Query 1/1: Action query time = 4.596 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7443 +t=26: Selected seed 195 with value = 0.7443 +Query 1/1: Action query time = 3.649 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7757 +t=42: Selected seed 195 with value = 0.7757 +Query 1/1: Action query time = 4.634 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8882 +t=58: Selected seed 195 with value = 0.8882 +Query 1/1: Action query time = 5.133 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9185 +t=74: Selected seed 195 with value = 0.9185 +Query 1/1: Action query time = 5.151 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9898 +t=90: Selected seed 195 with value = 0.9898 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=10--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 9 (90.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 4.784 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5568 +t=10: Selected seed 195 with value = 0.5568 +Query 1/1: Action query time = 4.938 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6930 +t=26: Selected seed 195 with value = 0.6930 +Query 1/1: Action query time = 2.705 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7779 +t=42: Selected seed 195 with value = 0.7779 +Query 1/1: Action query time = 3.588 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9353 +t=58: Selected seed 195 with value = 0.9353 +Query 1/1: Action query time = 3.558 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.861 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=90: Selected seed 195 with value = 0.9980 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=11--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 10 (90.9%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 3.990 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5316 +t=10: Selected seed 195 with value = 0.5316 +Query 1/1: Action query time = 4.751 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6517 +t=26: Selected seed 195 with value = 0.6517 +Query 1/1: Action query time = 3.769 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7687 +t=42: Selected seed 195 with value = 0.7687 +Query 1/1: Action query time = 4.769 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8588 +t=58: Selected seed 195 with value = 0.8588 +Query 1/1: Action query time = 5.256 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=74: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 4.009 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t1_s2/2026_08_03-17_25_48--with_future_img--episode=12--success=True--task=put_the_bowl_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 11 (91.7%) +Current task success rate: 0.9166666666666666 +Current total success rate: 0.9166666666666666 +Final results: +Total episodes: 12 +Total successes: 11 +Overall success rate: 0.9167 (91.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_49--v2i350_p1_t0_s2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_49--v2i350_p1_t0_s2.txt new file mode 100644 index 0000000000000000000000000000000000000000..dd6a3eb4c651186696ee838b8c886756191b30e8 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_49--v2i350_p1_t0_s2.txt @@ -0,0 +1,484 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p1_t0_s2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,6,10,14,18,22,26,30,34,38,42,46', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 7.145 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4417 +t=10: Selected seed 195 with value = 0.4417 +Query 1/1: Action query time = 5.504 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5147 +t=26: Selected seed 195 with value = 0.5147 +Query 1/1: Action query time = 4.554 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5859 +t=42: Selected seed 195 with value = 0.5859 +Query 1/1: Action query time = 4.573 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6823 +t=58: Selected seed 195 with value = 0.6823 +Query 1/1: Action query time = 4.604 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7847 +t=74: Selected seed 195 with value = 0.7847 +Query 1/1: Action query time = 3.644 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9466 +t=90: Selected seed 195 with value = 0.9466 +Query 1/1: Action query time = 3.287 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9966 +t=106: Selected seed 195 with value = 0.9966 +Query 1/1: Action query time = 4.189 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.480 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3481 +t=10: Selected seed 195 with value = 0.3481 +Query 1/1: Action query time = 5.136 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3129 +t=26: Selected seed 195 with value = 0.3129 +Query 1/1: Action query time = 4.899 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3730 +t=42: Selected seed 195 with value = 0.3730 +Query 1/1: Action query time = 3.485 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4451 +t=58: Selected seed 195 with value = 0.4451 +Query 1/1: Action query time = 5.231 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7305 +t=74: Selected seed 195 with value = 0.7305 +Query 1/1: Action query time = 5.268 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8448 +t=90: Selected seed 195 with value = 0.8448 +Query 1/1: Action query time = 4.585 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.853 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.052 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3934 +t=10: Selected seed 195 with value = 0.3934 +Query 1/1: Action query time = 3.115 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4413 +t=26: Selected seed 195 with value = 0.4413 +Query 1/1: Action query time = 4.704 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5698 +t=42: Selected seed 195 with value = 0.5698 +Query 1/1: Action query time = 4.981 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6626 +t=58: Selected seed 195 with value = 0.6626 +Query 1/1: Action query time = 4.884 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7421 +t=74: Selected seed 195 with value = 0.7421 +Query 1/1: Action query time = 4.705 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8845 +t=90: Selected seed 195 with value = 0.8845 +Query 1/1: Action query time = 4.861 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=106: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 4.052 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 4... +Query 1/1: Action query time = 4.876 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3991 +t=10: Selected seed 195 with value = 0.3991 +Query 1/1: Action query time = 4.955 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4790 +t=26: Selected seed 195 with value = 0.4790 +Query 1/1: Action query time = 4.662 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5552 +t=42: Selected seed 195 with value = 0.5552 +Query 1/1: Action query time = 4.950 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6535 +t=58: Selected seed 195 with value = 0.6535 +Query 1/1: Action query time = 4.907 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7425 +t=74: Selected seed 195 with value = 0.7425 +Query 1/1: Action query time = 3.394 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8864 +t=90: Selected seed 195 with value = 0.8864 +Query 1/1: Action query time = 3.254 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=106: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 4.101 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=4--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 5... +Query 1/1: Action query time = 5.786 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3688 +t=10: Selected seed 195 with value = 0.3688 +Query 1/1: Action query time = 5.683 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4629 +t=26: Selected seed 195 with value = 0.4629 +Query 1/1: Action query time = 4.545 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5565 +t=42: Selected seed 195 with value = 0.5565 +Query 1/1: Action query time = 4.937 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6565 +t=58: Selected seed 195 with value = 0.6565 +Query 1/1: Action query time = 4.692 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7493 +t=74: Selected seed 195 with value = 0.7493 +Query 1/1: Action query time = 5.290 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8837 +t=90: Selected seed 195 with value = 0.8837 +Query 1/1: Action query time = 4.576 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9953 +t=106: Selected seed 195 with value = 0.9953 +Query 1/1: Action query time = 4.859 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=5--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 6... +Query 1/1: Action query time = 3.910 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4239 +t=10: Selected seed 195 with value = 0.4239 +Query 1/1: Action query time = 4.430 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4966 +t=26: Selected seed 195 with value = 0.4966 +Query 1/1: Action query time = 4.283 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5541 +t=42: Selected seed 195 with value = 0.5541 +Query 1/1: Action query time = 4.626 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6730 +t=58: Selected seed 195 with value = 0.6730 +Query 1/1: Action query time = 5.949 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7551 +t=74: Selected seed 195 with value = 0.7551 +Query 1/1: Action query time = 5.441 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9162 +t=90: Selected seed 195 with value = 0.9162 +Query 1/1: Action query time = 4.767 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9995 +t=106: Selected seed 195 with value = 0.9995 +Query 1/1: Action query time = 4.559 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=6--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 7... +Query 1/1: Action query time = 3.932 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4098 +t=10: Selected seed 195 with value = 0.4098 +Query 1/1: Action query time = 5.377 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4329 +t=26: Selected seed 195 with value = 0.4329 +Query 1/1: Action query time = 4.599 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5184 +t=42: Selected seed 195 with value = 0.5184 +Query 1/1: Action query time = 4.472 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6266 +t=58: Selected seed 195 with value = 0.6266 +Query 1/1: Action query time = 4.772 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7199 +t=74: Selected seed 195 with value = 0.7199 +Query 1/1: Action query time = 4.188 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8232 +t=90: Selected seed 195 with value = 0.8232 +Query 1/1: Action query time = 4.857 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9591 +t=106: Selected seed 195 with value = 0.9591 +Query 1/1: Action query time = 3.458 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.323 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=138: Selected seed 195 with value = 0.9985 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=7--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 8... +Query 1/1: Action query time = 5.005 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4050 +t=10: Selected seed 195 with value = 0.4050 +Query 1/1: Action query time = 4.402 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4816 +t=26: Selected seed 195 with value = 0.4816 +Query 1/1: Action query time = 4.308 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5579 +t=42: Selected seed 195 with value = 0.5579 +Query 1/1: Action query time = 5.652 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6566 +t=58: Selected seed 195 with value = 0.6566 +Query 1/1: Action query time = 4.738 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7503 +t=74: Selected seed 195 with value = 0.7503 +Query 1/1: Action query time = 4.484 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8477 +t=90: Selected seed 195 with value = 0.8477 +Query 1/1: Action query time = 4.146 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9921 +t=106: Selected seed 195 with value = 0.9921 +Query 1/1: Action query time = 5.265 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=8--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 9... +Query 1/1: Action query time = 4.405 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4070 +t=10: Selected seed 195 with value = 0.4070 +Query 1/1: Action query time = 4.404 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4533 +t=26: Selected seed 195 with value = 0.4533 +Query 1/1: Action query time = 4.865 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5632 +t=42: Selected seed 195 with value = 0.5632 +Query 1/1: Action query time = 5.221 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6501 +t=58: Selected seed 195 with value = 0.6501 +Query 1/1: Action query time = 5.186 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7605 +t=74: Selected seed 195 with value = 0.7605 +Query 1/1: Action query time = 4.142 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9012 +t=90: Selected seed 195 with value = 0.9012 +Query 1/1: Action query time = 5.033 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9813 +t=106: Selected seed 195 with value = 0.9813 +Query 1/1: Action query time = 4.753 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=9--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 10... +Query 1/1: Action query time = 3.458 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4163 +t=10: Selected seed 195 with value = 0.4163 +Query 1/1: Action query time = 4.572 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4843 +t=26: Selected seed 195 with value = 0.4843 +Query 1/1: Action query time = 4.260 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5402 +t=42: Selected seed 195 with value = 0.5402 +Query 1/1: Action query time = 5.055 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6275 +t=58: Selected seed 195 with value = 0.6275 +Query 1/1: Action query time = 5.265 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7124 +t=74: Selected seed 195 with value = 0.7124 +Query 1/1: Action query time = 3.451 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9157 +t=90: Selected seed 195 with value = 0.9157 +Query 1/1: Action query time = 5.491 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.621 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=10--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 11... +Query 1/1: Action query time = 3.408 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4263 +t=10: Selected seed 195 with value = 0.4263 +Query 1/1: Action query time = 4.020 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4884 +t=26: Selected seed 195 with value = 0.4884 +Query 1/1: Action query time = 5.379 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5516 +t=42: Selected seed 195 with value = 0.5516 +Query 1/1: Action query time = 3.880 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6623 +t=58: Selected seed 195 with value = 0.6623 +Query 1/1: Action query time = 3.607 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7549 +t=74: Selected seed 195 with value = 0.7549 +Query 1/1: Action query time = 3.195 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9129 +t=90: Selected seed 195 with value = 0.9129 +Query 1/1: Action query time = 3.368 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.371 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=11--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 12... +Query 1/1: Action query time = 2.581 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4173 +t=10: Selected seed 195 with value = 0.4173 +Query 1/1: Action query time = 2.863 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4845 +t=26: Selected seed 195 with value = 0.4845 +Query 1/1: Action query time = 3.183 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5706 +t=42: Selected seed 195 with value = 0.5706 +Query 1/1: Action query time = 2.209 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6702 +t=58: Selected seed 195 with value = 0.6702 +Query 1/1: Action query time = 2.563 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7722 +t=74: Selected seed 195 with value = 0.7722 +Query 1/1: Action query time = 2.546 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9230 +t=90: Selected seed 195 with value = 0.9230 +Query 1/1: Action query time = 2.161 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.119 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t0_s2/2026_08_03-17_25_49--with_future_img--episode=12--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 12 +Total successes: 12 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_49--v2i350_p1_t3_s2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_49--v2i350_p1_t3_s2.txt new file mode 100644 index 0000000000000000000000000000000000000000..b442de2ea2312e6b10069612979cdf15ade18df0 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_49--v2i350_p1_t3_s2.txt @@ -0,0 +1,752 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p1_t3_s2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,6,10,14,18,22,26,30,34,38,42,46', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 2.965 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2177 +t=10: Selected seed 195 with value = 0.2177 +Query 1/1: Action query time = 5.330 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2620 +t=26: Selected seed 195 with value = 0.2620 +Query 1/1: Action query time = 4.252 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2963 +t=42: Selected seed 195 with value = 0.2963 +Query 1/1: Action query time = 5.038 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3397 +t=58: Selected seed 195 with value = 0.3397 +Query 1/1: Action query time = 5.205 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3904 +t=74: Selected seed 195 with value = 0.3904 +Query 1/1: Action query time = 5.360 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4574 +t=90: Selected seed 195 with value = 0.4574 +Query 1/1: Action query time = 4.809 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5776 +t=106: Selected seed 195 with value = 0.5776 +Query 1/1: Action query time = 2.865 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6670 +t=122: Selected seed 195 with value = 0.6670 +Query 1/1: Action query time = 4.210 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7855 +t=138: Selected seed 195 with value = 0.7855 +Query 1/1: Action query time = 4.325 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9151 +t=154: Selected seed 195 with value = 0.9151 +Query 1/1: Action query time = 4.810 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9835 +t=170: Selected seed 195 with value = 0.9835 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 4.474 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2131 +t=10: Selected seed 195 with value = 0.2131 +Query 1/1: Action query time = 3.262 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2355 +t=26: Selected seed 195 with value = 0.2355 +Query 1/1: Action query time = 4.801 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3091 +t=42: Selected seed 195 with value = 0.3091 +Query 1/1: Action query time = 5.597 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3495 +t=58: Selected seed 195 with value = 0.3495 +Query 1/1: Action query time = 5.329 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4151 +t=74: Selected seed 195 with value = 0.4151 +Query 1/1: Action query time = 5.190 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4635 +t=90: Selected seed 195 with value = 0.4635 +Query 1/1: Action query time = 3.957 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5729 +t=106: Selected seed 195 with value = 0.5729 +Query 1/1: Action query time = 4.440 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6946 +t=122: Selected seed 195 with value = 0.6946 +Query 1/1: Action query time = 4.521 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7903 +t=138: Selected seed 195 with value = 0.7903 +Query 1/1: Action query time = 3.876 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8655 +t=154: Selected seed 195 with value = 0.8655 +Query 1/1: Action query time = 5.120 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 3.902 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2140 +t=10: Selected seed 195 with value = 0.2140 +Query 1/1: Action query time = 3.778 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2458 +t=26: Selected seed 195 with value = 0.2458 +Query 1/1: Action query time = 5.029 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3072 +t=42: Selected seed 195 with value = 0.3072 +Query 1/1: Action query time = 4.462 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3550 +t=58: Selected seed 195 with value = 0.3550 +Query 1/1: Action query time = 5.319 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4243 +t=74: Selected seed 195 with value = 0.4243 +Query 1/1: Action query time = 4.653 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5029 +t=90: Selected seed 195 with value = 0.5029 +Query 1/1: Action query time = 5.021 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6043 +t=106: Selected seed 195 with value = 0.6043 +Query 1/1: Action query time = 4.957 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7566 +t=122: Selected seed 195 with value = 0.7566 +Query 1/1: Action query time = 4.710 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8356 +t=138: Selected seed 195 with value = 0.8356 +Query 1/1: Action query time = 4.688 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9801 +t=154: Selected seed 195 with value = 0.9801 +Query 1/1: Action query time = 5.139 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9574 +t=170: Selected seed 195 with value = 0.9574 +Query 1/1: Action query time = 4.962 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9528 +t=186: Selected seed 195 with value = 0.9528 +Query 1/1: Action query time = 5.409 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6811 +t=202: Selected seed 195 with value = 0.6811 +Query 1/1: Action query time = 3.891 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8341 +t=218: Selected seed 195 with value = 0.8341 +Query 1/1: Action query time = 4.634 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8825 +t=234: Selected seed 195 with value = 0.8825 +Query 1/1: Action query time = 5.254 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3612 +t=250: Selected seed 195 with value = 0.3612 +Query 1/1: Action query time = 3.901 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9066 +t=266: Selected seed 195 with value = 0.9066 +Query 1/1: Action query time = 4.324 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8127 +t=282: Selected seed 195 with value = 0.8127 +Query 1/1: Action query time = 3.369 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4075 +t=298: Selected seed 195 with value = 0.4075 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=3--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 3 +# successes: 2 (66.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 4.376 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2095 +t=10: Selected seed 195 with value = 0.2095 +Query 1/1: Action query time = 4.998 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2447 +t=26: Selected seed 195 with value = 0.2447 +Query 1/1: Action query time = 4.558 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3111 +t=42: Selected seed 195 with value = 0.3111 +Query 1/1: Action query time = 5.709 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3568 +t=58: Selected seed 195 with value = 0.3568 +Query 1/1: Action query time = 4.084 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4220 +t=74: Selected seed 195 with value = 0.4220 +Query 1/1: Action query time = 3.054 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4915 +t=90: Selected seed 195 with value = 0.4915 +Query 1/1: Action query time = 4.285 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6126 +t=106: Selected seed 195 with value = 0.6126 +Query 1/1: Action query time = 5.263 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7947 +t=122: Selected seed 195 with value = 0.7947 +Query 1/1: Action query time = 4.788 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8770 +t=138: Selected seed 195 with value = 0.8770 +Query 1/1: Action query time = 5.039 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9691 +t=154: Selected seed 195 with value = 0.9691 +Query 1/1: Action query time = 4.960 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=170: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=4--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 4 +# successes: 3 (75.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 5... +Query 1/1: Action query time = 4.497 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2104 +t=10: Selected seed 195 with value = 0.2104 +Query 1/1: Action query time = 4.322 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2609 +t=26: Selected seed 195 with value = 0.2609 +Query 1/1: Action query time = 5.222 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2957 +t=42: Selected seed 195 with value = 0.2957 +Query 1/1: Action query time = 4.634 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3347 +t=58: Selected seed 195 with value = 0.3347 +Query 1/1: Action query time = 3.898 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3953 +t=74: Selected seed 195 with value = 0.3953 +Query 1/1: Action query time = 4.862 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4549 +t=90: Selected seed 195 with value = 0.4549 +Query 1/1: Action query time = 3.367 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5601 +t=106: Selected seed 195 with value = 0.5601 +Query 1/1: Action query time = 3.800 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6569 +t=122: Selected seed 195 with value = 0.6569 +Query 1/1: Action query time = 4.632 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7583 +t=138: Selected seed 195 with value = 0.7583 +Query 1/1: Action query time = 4.964 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8953 +t=154: Selected seed 195 with value = 0.8953 +Query 1/1: Action query time = 4.561 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9905 +t=170: Selected seed 195 with value = 0.9905 +Query 1/1: Action query time = 4.540 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9456 +t=186: Selected seed 195 with value = 0.9456 +Query 1/1: Action query time = 4.816 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9544 +t=202: Selected seed 195 with value = 0.9544 +Query 1/1: Action query time = 5.551 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9670 +t=218: Selected seed 195 with value = 0.9670 +Query 1/1: Action query time = 4.401 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3441 +t=234: Selected seed 195 with value = 0.3441 +Query 1/1: Action query time = 4.712 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3557 +t=250: Selected seed 195 with value = 0.3557 +Query 1/1: Action query time = 4.986 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9298 +t=266: Selected seed 195 with value = 0.9298 +Query 1/1: Action query time = 5.087 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9252 +t=282: Selected seed 195 with value = 0.9252 +Query 1/1: Action query time = 4.346 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9211 +t=298: Selected seed 195 with value = 0.9211 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=5--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 5 +# successes: 3 (60.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 6... +Query 1/1: Action query time = 3.601 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2190 +t=10: Selected seed 195 with value = 0.2190 +Query 1/1: Action query time = 5.003 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2443 +t=26: Selected seed 195 with value = 0.2443 +Query 1/1: Action query time = 5.406 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3137 +t=42: Selected seed 195 with value = 0.3137 +Query 1/1: Action query time = 4.955 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3617 +t=58: Selected seed 195 with value = 0.3617 +Query 1/1: Action query time = 4.892 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4174 +t=74: Selected seed 195 with value = 0.4174 +Query 1/1: Action query time = 3.366 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4894 +t=90: Selected seed 195 with value = 0.4894 +Query 1/1: Action query time = 4.267 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6123 +t=106: Selected seed 195 with value = 0.6123 +Query 1/1: Action query time = 5.337 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8207 +t=122: Selected seed 195 with value = 0.8207 +Query 1/1: Action query time = 4.986 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8799 +t=138: Selected seed 195 with value = 0.8799 +Query 1/1: Action query time = 4.035 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.116 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=6--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 6 +# successes: 4 (66.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 7... +Query 1/1: Action query time = 2.729 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2110 +t=10: Selected seed 195 with value = 0.2110 +Query 1/1: Action query time = 4.952 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2316 +t=26: Selected seed 195 with value = 0.2316 +Query 1/1: Action query time = 5.334 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3093 +t=42: Selected seed 195 with value = 0.3093 +Query 1/1: Action query time = 5.136 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3522 +t=58: Selected seed 195 with value = 0.3522 +Query 1/1: Action query time = 5.452 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4202 +t=74: Selected seed 195 with value = 0.4202 +Query 1/1: Action query time = 4.629 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4787 +t=90: Selected seed 195 with value = 0.4787 +Query 1/1: Action query time = 3.820 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5775 +t=106: Selected seed 195 with value = 0.5775 +Query 1/1: Action query time = 4.168 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6941 +t=122: Selected seed 195 with value = 0.6941 +Query 1/1: Action query time = 4.553 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8204 +t=138: Selected seed 195 with value = 0.8204 +Query 1/1: Action query time = 4.494 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9608 +t=154: Selected seed 195 with value = 0.9608 +Query 1/1: Action query time = 5.128 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9638 +t=170: Selected seed 195 with value = 0.9638 +Query 1/1: Action query time = 4.728 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9527 +t=186: Selected seed 195 with value = 0.9527 +Query 1/1: Action query time = 4.535 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9382 +t=202: Selected seed 195 with value = 0.9382 +Query 1/1: Action query time = 4.452 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9452 +t=218: Selected seed 195 with value = 0.9452 +Query 1/1: Action query time = 3.661 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3739 +t=234: Selected seed 195 with value = 0.3739 +Query 1/1: Action query time = 4.767 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9518 +t=250: Selected seed 195 with value = 0.9518 +Query 1/1: Action query time = 4.889 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9296 +t=266: Selected seed 195 with value = 0.9296 +Query 1/1: Action query time = 4.469 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8454 +t=282: Selected seed 195 with value = 0.8454 +Query 1/1: Action query time = 3.968 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9253 +t=298: Selected seed 195 with value = 0.9253 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=7--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 7 +# successes: 4 (57.1%) + +Task: open the top drawer and put the bowl inside +Starting episode 8... +Query 1/1: Action query time = 5.033 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2114 +t=10: Selected seed 195 with value = 0.2114 +Query 1/1: Action query time = 4.504 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2462 +t=26: Selected seed 195 with value = 0.2462 +Query 1/1: Action query time = 4.855 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3123 +t=42: Selected seed 195 with value = 0.3123 +Query 1/1: Action query time = 3.335 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3577 +t=58: Selected seed 195 with value = 0.3577 +Query 1/1: Action query time = 4.183 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4204 +t=74: Selected seed 195 with value = 0.4204 +Query 1/1: Action query time = 2.677 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5167 +t=90: Selected seed 195 with value = 0.5167 +Query 1/1: Action query time = 3.338 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6101 +t=106: Selected seed 195 with value = 0.6101 +Query 1/1: Action query time = 3.738 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7329 +t=122: Selected seed 195 with value = 0.7329 +Query 1/1: Action query time = 3.099 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7839 +t=138: Selected seed 195 with value = 0.7839 +Query 1/1: Action query time = 3.134 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9321 +t=154: Selected seed 195 with value = 0.9321 +Query 1/1: Action query time = 3.143 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=8--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 8 +# successes: 5 (62.5%) + +Task: open the top drawer and put the bowl inside +Starting episode 9... +Query 1/1: Action query time = 3.592 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2148 +t=10: Selected seed 195 with value = 0.2148 +Query 1/1: Action query time = 3.464 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2461 +t=26: Selected seed 195 with value = 0.2461 +Query 1/1: Action query time = 2.840 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3162 +t=42: Selected seed 195 with value = 0.3162 +Query 1/1: Action query time = 2.742 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3590 +t=58: Selected seed 195 with value = 0.3590 +Query 1/1: Action query time = 3.043 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4263 +t=74: Selected seed 195 with value = 0.4263 +Query 1/1: Action query time = 3.031 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5326 +t=90: Selected seed 195 with value = 0.5326 +Query 1/1: Action query time = 2.211 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6085 +t=106: Selected seed 195 with value = 0.6085 +Query 1/1: Action query time = 2.202 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8088 +t=122: Selected seed 195 with value = 0.8088 +Query 1/1: Action query time = 2.681 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8243 +t=138: Selected seed 195 with value = 0.8243 +Query 1/1: Action query time = 3.085 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9045 +t=154: Selected seed 195 with value = 0.9045 +Query 1/1: Action query time = 2.897 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9683 +t=170: Selected seed 195 with value = 0.9683 +Query 1/1: Action query time = 2.872 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9460 +t=186: Selected seed 195 with value = 0.9460 +Query 1/1: Action query time = 2.932 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3118 +t=202: Selected seed 195 with value = 0.3118 +Query 1/1: Action query time = 2.792 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3254 +t=218: Selected seed 195 with value = 0.3254 +Query 1/1: Action query time = 2.638 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9406 +t=234: Selected seed 195 with value = 0.9406 +Query 1/1: Action query time = 2.408 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3451 +t=250: Selected seed 195 with value = 0.3451 +Query 1/1: Action query time = 1.865 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9117 +t=266: Selected seed 195 with value = 0.9117 +Query 1/1: Action query time = 1.462 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9649 +t=282: Selected seed 195 with value = 0.9649 +Query 1/1: Action query time = 2.395 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8636 +t=298: Selected seed 195 with value = 0.8636 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=9--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 9 +# successes: 5 (55.6%) + +Task: open the top drawer and put the bowl inside +Starting episode 10... +Query 1/1: Action query time = 1.862 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2245 +t=10: Selected seed 195 with value = 0.2245 +Query 1/1: Action query time = 1.810 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2392 +t=26: Selected seed 195 with value = 0.2392 +Query 1/1: Action query time = 2.617 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2971 +t=42: Selected seed 195 with value = 0.2971 +Query 1/1: Action query time = 2.581 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3501 +t=58: Selected seed 195 with value = 0.3501 +Query 1/1: Action query time = 2.527 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3987 +t=74: Selected seed 195 with value = 0.3987 +Query 1/1: Action query time = 2.441 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4573 +t=90: Selected seed 195 with value = 0.4573 +Query 1/1: Action query time = 2.400 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5980 +t=106: Selected seed 195 with value = 0.5980 +Query 1/1: Action query time = 2.374 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6673 +t=122: Selected seed 195 with value = 0.6673 +Query 1/1: Action query time = 2.120 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8156 +t=138: Selected seed 195 with value = 0.8156 +Query 1/1: Action query time = 1.613 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9211 +t=154: Selected seed 195 with value = 0.9211 +Query 1/1: Action query time = 1.510 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=10--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 10 +# successes: 6 (60.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 11... +Query 1/1: Action query time = 2.729 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2192 +t=10: Selected seed 195 with value = 0.2192 +Query 1/1: Action query time = 2.561 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2477 +t=26: Selected seed 195 with value = 0.2477 +Query 1/1: Action query time = 1.888 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3114 +t=42: Selected seed 195 with value = 0.3114 +Query 1/1: Action query time = 1.745 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3556 +t=58: Selected seed 195 with value = 0.3556 +Query 1/1: Action query time = 1.674 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4138 +t=74: Selected seed 195 with value = 0.4138 +Query 1/1: Action query time = 2.235 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4792 +t=90: Selected seed 195 with value = 0.4792 +Query 1/1: Action query time = 2.121 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6019 +t=106: Selected seed 195 with value = 0.6019 +Query 1/1: Action query time = 2.256 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7427 +t=122: Selected seed 195 with value = 0.7427 +Query 1/1: Action query time = 1.908 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8765 +t=138: Selected seed 195 with value = 0.8765 +Query 1/1: Action query time = 1.718 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9765 +t=154: Selected seed 195 with value = 0.9765 +Query 1/1: Action query time = 1.691 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=170: Selected seed 195 with value = 0.9971 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=11--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 11 +# successes: 7 (63.6%) + +Task: open the top drawer and put the bowl inside +Starting episode 12... +Query 1/1: Action query time = 2.033 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2153 +t=10: Selected seed 195 with value = 0.2153 +Query 1/1: Action query time = 1.917 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2452 +t=26: Selected seed 195 with value = 0.2452 +Query 1/1: Action query time = 1.351 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2993 +t=42: Selected seed 195 with value = 0.2993 +Query 1/1: Action query time = 1.347 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3441 +t=58: Selected seed 195 with value = 0.3441 +Query 1/1: Action query time = 1.402 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4011 +t=74: Selected seed 195 with value = 0.4011 +Query 1/1: Action query time = 1.438 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4792 +t=90: Selected seed 195 with value = 0.4792 +Query 1/1: Action query time = 1.475 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5686 +t=106: Selected seed 195 with value = 0.5686 +Query 1/1: Action query time = 1.496 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6991 +t=122: Selected seed 195 with value = 0.6991 +Query 1/1: Action query time = 0.966 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7810 +t=138: Selected seed 195 with value = 0.7810 +Query 1/1: Action query time = 0.976 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9241 +t=154: Selected seed 195 with value = 0.9241 +Query 1/1: Action query time = 0.972 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9933 +t=170: Selected seed 195 with value = 0.9933 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t3_s2/2026_08_03-17_25_49--with_future_img--episode=12--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 12 +# successes: 8 (66.7%) +Current task success rate: 0.6666666666666666 +Current total success rate: 0.6666666666666666 +Final results: +Total episodes: 12 +Total successes: 8 +Overall success rate: 0.6667 (66.7%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_50--v2i350_p1_t2_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_50--v2i350_p1_t2_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..dd177db3ab6dcc677196e2046475a7afab9d7a73 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_25_50--v2i350_p1_t2_s1.txt @@ -0,0 +1,471 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p1_t2_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2] +Using default initial states + +Task: put the wine bottle on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 5.960 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5694 +t=10: Selected seed 195 with value = 0.5694 +Query 1/1: Action query time = 4.434 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6268 +t=26: Selected seed 195 with value = 0.6268 +Query 1/1: Action query time = 5.233 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7547 +t=42: Selected seed 195 with value = 0.7547 +Query 1/1: Action query time = 5.113 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8803 +t=58: Selected seed 195 with value = 0.8803 +Query 1/1: Action query time = 5.444 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9740 +t=74: Selected seed 195 with value = 0.9740 +Query 1/1: Action query time = 4.859 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=1--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 5.532 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5909 +t=10: Selected seed 195 with value = 0.5909 +Query 1/1: Action query time = 5.502 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7043 +t=26: Selected seed 195 with value = 0.7043 +Query 1/1: Action query time = 4.252 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8132 +t=42: Selected seed 195 with value = 0.8132 +Query 1/1: Action query time = 2.640 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9562 +t=58: Selected seed 195 with value = 0.9562 +Query 1/1: Action query time = 3.930 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.893 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=2--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.041 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5819 +t=10: Selected seed 195 with value = 0.5819 +Query 1/1: Action query time = 4.618 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7056 +t=26: Selected seed 195 with value = 0.7056 +Query 1/1: Action query time = 4.718 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8558 +t=42: Selected seed 195 with value = 0.8558 +Query 1/1: Action query time = 3.784 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9727 +t=58: Selected seed 195 with value = 0.9727 +Query 1/1: Action query time = 5.377 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.888 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=3--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 5.085 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5814 +t=10: Selected seed 195 with value = 0.5814 +Query 1/1: Action query time = 4.575 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6523 +t=26: Selected seed 195 with value = 0.6523 +Query 1/1: Action query time = 4.269 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7556 +t=42: Selected seed 195 with value = 0.7556 +Query 1/1: Action query time = 3.584 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8784 +t=58: Selected seed 195 with value = 0.8784 +Query 1/1: Action query time = 3.973 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9769 +t=74: Selected seed 195 with value = 0.9769 +Query 1/1: Action query time = 4.103 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=90: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 5.100 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=4--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 5... +Query 1/1: Action query time = 5.547 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5627 +t=10: Selected seed 195 with value = 0.5627 +Query 1/1: Action query time = 4.635 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6263 +t=26: Selected seed 195 with value = 0.6263 +Query 1/1: Action query time = 5.131 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7494 +t=42: Selected seed 195 with value = 0.7494 +Query 1/1: Action query time = 4.449 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9220 +t=58: Selected seed 195 with value = 0.9220 +Query 1/1: Action query time = 3.868 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9919 +t=74: Selected seed 195 with value = 0.9919 +Query 1/1: Action query time = 4.628 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=5--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 6... +Query 1/1: Action query time = 5.226 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5868 +t=10: Selected seed 195 with value = 0.5868 +Query 1/1: Action query time = 3.726 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7220 +t=26: Selected seed 195 with value = 0.7220 +Query 1/1: Action query time = 4.681 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8267 +t=42: Selected seed 195 with value = 0.8267 +Query 1/1: Action query time = 5.189 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9700 +t=58: Selected seed 195 with value = 0.9700 +Query 1/1: Action query time = 3.985 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=74: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 4.376 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=6--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 7... +Query 1/1: Action query time = 4.284 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5755 +t=10: Selected seed 195 with value = 0.5755 +Query 1/1: Action query time = 3.357 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6700 +t=26: Selected seed 195 with value = 0.6700 +Query 1/1: Action query time = 5.028 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7778 +t=42: Selected seed 195 with value = 0.7778 +Query 1/1: Action query time = 5.370 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9168 +t=58: Selected seed 195 with value = 0.9168 +Query 1/1: Action query time = 4.502 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9851 +t=74: Selected seed 195 with value = 0.9851 +Query 1/1: Action query time = 5.032 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=7--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 8... +Query 1/1: Action query time = 3.363 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5913 +t=10: Selected seed 195 with value = 0.5913 +Query 1/1: Action query time = 4.397 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7078 +t=26: Selected seed 195 with value = 0.7078 +Query 1/1: Action query time = 5.728 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8152 +t=42: Selected seed 195 with value = 0.8152 +Query 1/1: Action query time = 5.095 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9298 +t=58: Selected seed 195 with value = 0.9298 +Query 1/1: Action query time = 5.214 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=74: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 4.730 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=8--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 9... +Query 1/1: Action query time = 4.588 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5494 +t=10: Selected seed 195 with value = 0.5494 +Query 1/1: Action query time = 4.764 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6826 +t=26: Selected seed 195 with value = 0.6826 +Query 1/1: Action query time = 4.840 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7857 +t=42: Selected seed 195 with value = 0.7857 +Query 1/1: Action query time = 4.940 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8821 +t=58: Selected seed 195 with value = 0.8821 +Query 1/1: Action query time = 3.542 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9748 +t=74: Selected seed 195 with value = 0.9748 +Query 1/1: Action query time = 4.819 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=9--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 10... +Query 1/1: Action query time = 4.282 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5241 +t=10: Selected seed 195 with value = 0.5241 +Query 1/1: Action query time = 4.694 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5008 +t=26: Selected seed 195 with value = 0.5008 +Query 1/1: Action query time = 5.550 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6879 +t=42: Selected seed 195 with value = 0.6879 +Query 1/1: Action query time = 4.316 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8198 +t=58: Selected seed 195 with value = 0.8198 +Query 1/1: Action query time = 4.059 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8981 +t=74: Selected seed 195 with value = 0.8981 +Query 1/1: Action query time = 4.834 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9812 +t=90: Selected seed 195 with value = 0.9812 +Query 1/1: Action query time = 5.186 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.338 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9877 +t=122: Selected seed 195 with value = 0.9877 +Query 1/1: Action query time = 5.109 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9887 +t=138: Selected seed 195 with value = 0.9887 +Query 1/1: Action query time = 4.904 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=154: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 4.889 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.015 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.323 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.955 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.304 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.185 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.711 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=266: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 5.048 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9924 +t=282: Selected seed 195 with value = 0.9924 +Query 1/1: Action query time = 4.801 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9867 +t=298: Selected seed 195 with value = 0.9867 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=10--success=False--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: False +# episodes completed so far: 10 +# successes: 9 (90.0%) + +Task: put the wine bottle on top of the cabinet +Starting episode 11... +Query 1/1: Action query time = 3.250 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5780 +t=10: Selected seed 195 with value = 0.5780 +Query 1/1: Action query time = 4.880 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6964 +t=26: Selected seed 195 with value = 0.6964 +Query 1/1: Action query time = 5.355 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7821 +t=42: Selected seed 195 with value = 0.7821 +Query 1/1: Action query time = 4.460 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9239 +t=58: Selected seed 195 with value = 0.9239 +Query 1/1: Action query time = 4.908 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=74: Selected seed 195 with value = 0.9965 +Query 1/1: Action query time = 3.390 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=11--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 11 +# successes: 10 (90.9%) + +Task: put the wine bottle on top of the cabinet +Starting episode 12... +Query 1/1: Action query time = 4.478 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6041 +t=10: Selected seed 195 with value = 0.6041 +Query 1/1: Action query time = 4.035 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7199 +t=26: Selected seed 195 with value = 0.7199 +Query 1/1: Action query time = 4.798 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8229 +t=42: Selected seed 195 with value = 0.8229 +Query 1/1: Action query time = 4.348 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9403 +t=58: Selected seed 195 with value = 0.9403 +Query 1/1: Action query time = 4.512 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.671 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=12--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 12 +# successes: 11 (91.7%) + +Task: put the wine bottle on top of the cabinet +Starting episode 13... +Query 1/1: Action query time = 5.511 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5511 +t=10: Selected seed 195 with value = 0.5511 +Query 1/1: Action query time = 3.484 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6505 +t=26: Selected seed 195 with value = 0.6505 +Query 1/1: Action query time = 4.515 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7884 +t=42: Selected seed 195 with value = 0.7884 +Query 1/1: Action query time = 4.776 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8817 +t=58: Selected seed 195 with value = 0.8817 +Query 1/1: Action query time = 4.487 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9768 +t=74: Selected seed 195 with value = 0.9768 +Query 1/1: Action query time = 4.739 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t2_s1/2026_08_03-17_25_50--with_future_img--episode=13--success=True--task=put_the_wine_bottle_on_top_of_the_c.mp4 +Success: True +# episodes completed so far: 13 +# successes: 12 (92.3%) +Current task success rate: 0.9230769230769231 +Current total success rate: 0.9230769230769231 +Final results: +Total episodes: 13 +Total successes: 12 +Overall success rate: 0.9231 (92.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_39_22--v2i350_p1_t5_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_39_22--v2i350_p1_t5_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..25118748fb572e560ea38d435749f26a5c9ee8a6 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_39_22--v2i350_p1_t5_s3.txt @@ -0,0 +1,508 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p1_t5_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,7,11,15,19,23,27,31,35,39,43,47', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 5.462 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3704 +t=10: Selected seed 195 with value = 0.3704 +Query 1/1: Action query time = 4.975 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4253 +t=26: Selected seed 195 with value = 0.4253 +Query 1/1: Action query time = 5.583 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5278 +t=42: Selected seed 195 with value = 0.5278 +Query 1/1: Action query time = 5.087 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6820 +t=58: Selected seed 195 with value = 0.6820 +Query 1/1: Action query time = 4.224 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8007 +t=74: Selected seed 195 with value = 0.8007 +Query 1/1: Action query time = 3.778 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9155 +t=90: Selected seed 195 with value = 0.9155 +Query 1/1: Action query time = 4.334 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.477 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=122: Selected seed 195 with value = 0.9992 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 3.466 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3938 +t=10: Selected seed 195 with value = 0.3938 +Query 1/1: Action query time = 4.216 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4471 +t=26: Selected seed 195 with value = 0.4471 +Query 1/1: Action query time = 4.798 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5466 +t=42: Selected seed 195 with value = 0.5466 +Query 1/1: Action query time = 5.138 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5437 +t=58: Selected seed 195 with value = 0.5437 +Query 1/1: Action query time = 3.662 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6049 +t=74: Selected seed 195 with value = 0.6049 +Query 1/1: Action query time = 3.898 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6959 +t=90: Selected seed 195 with value = 0.6959 +Query 1/1: Action query time = 4.824 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7880 +t=106: Selected seed 195 with value = 0.7880 +Query 1/1: Action query time = 4.656 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9372 +t=122: Selected seed 195 with value = 0.9372 +Query 1/1: Action query time = 4.675 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9532 +t=138: Selected seed 195 with value = 0.9532 +Query 1/1: Action query time = 4.716 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 2.628 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4051 +t=10: Selected seed 195 with value = 0.4051 +Query 1/1: Action query time = 2.601 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4714 +t=26: Selected seed 195 with value = 0.4714 +Query 1/1: Action query time = 5.417 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5889 +t=42: Selected seed 195 with value = 0.5889 +Query 1/1: Action query time = 5.319 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6387 +t=58: Selected seed 195 with value = 0.6387 +Query 1/1: Action query time = 4.634 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7548 +t=74: Selected seed 195 with value = 0.7548 +Query 1/1: Action query time = 4.345 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8783 +t=90: Selected seed 195 with value = 0.8783 +Query 1/1: Action query time = 5.177 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9899 +t=106: Selected seed 195 with value = 0.9899 +Query 1/1: Action query time = 4.514 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 4.994 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4100 +t=10: Selected seed 195 with value = 0.4100 +Query 1/1: Action query time = 3.286 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4746 +t=26: Selected seed 195 with value = 0.4746 +Query 1/1: Action query time = 4.809 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5570 +t=42: Selected seed 195 with value = 0.5570 +Query 1/1: Action query time = 5.167 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6029 +t=58: Selected seed 195 with value = 0.6029 +Query 1/1: Action query time = 4.824 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7338 +t=74: Selected seed 195 with value = 0.7338 +Query 1/1: Action query time = 4.303 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7961 +t=90: Selected seed 195 with value = 0.7961 +Query 1/1: Action query time = 5.621 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8883 +t=106: Selected seed 195 with value = 0.8883 +Query 1/1: Action query time = 3.721 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=122: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 4.721 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9989 +t=138: Selected seed 195 with value = 0.9989 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=4--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 5... +Query 1/1: Action query time = 5.197 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3918 +t=10: Selected seed 195 with value = 0.3918 +Query 1/1: Action query time = 3.427 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4549 +t=26: Selected seed 195 with value = 0.4549 +Query 1/1: Action query time = 2.842 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5415 +t=42: Selected seed 195 with value = 0.5415 +Query 1/1: Action query time = 4.914 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6122 +t=58: Selected seed 195 with value = 0.6122 +Query 1/1: Action query time = 4.955 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7021 +t=74: Selected seed 195 with value = 0.7021 +Query 1/1: Action query time = 4.814 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8635 +t=90: Selected seed 195 with value = 0.8635 +Query 1/1: Action query time = 4.775 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9865 +t=106: Selected seed 195 with value = 0.9865 +Query 1/1: Action query time = 3.782 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=122: Selected seed 195 with value = 0.9985 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=5--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 6... +Query 1/1: Action query time = 4.878 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3772 +t=10: Selected seed 195 with value = 0.3772 +Query 1/1: Action query time = 4.414 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4826 +t=26: Selected seed 195 with value = 0.4826 +Query 1/1: Action query time = 3.929 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5749 +t=42: Selected seed 195 with value = 0.5749 +Query 1/1: Action query time = 3.430 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6450 +t=58: Selected seed 195 with value = 0.6450 +Query 1/1: Action query time = 4.931 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7553 +t=74: Selected seed 195 with value = 0.7553 +Query 1/1: Action query time = 5.444 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8694 +t=90: Selected seed 195 with value = 0.8694 +Query 1/1: Action query time = 3.663 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9815 +t=106: Selected seed 195 with value = 0.9815 +Query 1/1: Action query time = 5.232 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=6--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 7... +Query 1/1: Action query time = 4.688 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3576 +t=10: Selected seed 195 with value = 0.3576 +Query 1/1: Action query time = 5.454 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4164 +t=26: Selected seed 195 with value = 0.4164 +Query 1/1: Action query time = 4.565 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4913 +t=42: Selected seed 195 with value = 0.4913 +Query 1/1: Action query time = 4.734 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6189 +t=58: Selected seed 195 with value = 0.6189 +Query 1/1: Action query time = 4.662 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6533 +t=74: Selected seed 195 with value = 0.6533 +Query 1/1: Action query time = 5.010 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7302 +t=90: Selected seed 195 with value = 0.7302 +Query 1/1: Action query time = 4.606 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8347 +t=106: Selected seed 195 with value = 0.8347 +Query 1/1: Action query time = 5.114 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=122: Selected seed 195 with value = 0.9971 +Query 1/1: Action query time = 4.673 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=138: Selected seed 195 with value = 0.9981 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=7--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 8... +Query 1/1: Action query time = 4.932 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3815 +t=10: Selected seed 195 with value = 0.3815 +Query 1/1: Action query time = 4.651 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4431 +t=26: Selected seed 195 with value = 0.4431 +Query 1/1: Action query time = 3.782 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5078 +t=42: Selected seed 195 with value = 0.5078 +Query 1/1: Action query time = 4.156 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6410 +t=58: Selected seed 195 with value = 0.6410 +Query 1/1: Action query time = 2.448 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7538 +t=74: Selected seed 195 with value = 0.7538 +Query 1/1: Action query time = 4.614 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8778 +t=90: Selected seed 195 with value = 0.8778 +Query 1/1: Action query time = 5.227 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.387 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=8--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 9... +Query 1/1: Action query time = 4.467 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3783 +t=10: Selected seed 195 with value = 0.3783 +Query 1/1: Action query time = 4.694 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4182 +t=26: Selected seed 195 with value = 0.4182 +Query 1/1: Action query time = 5.255 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5177 +t=42: Selected seed 195 with value = 0.5177 +Query 1/1: Action query time = 4.304 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5670 +t=58: Selected seed 195 with value = 0.5670 +Query 1/1: Action query time = 4.430 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6736 +t=74: Selected seed 195 with value = 0.6736 +Query 1/1: Action query time = 4.550 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7124 +t=90: Selected seed 195 with value = 0.7124 +Query 1/1: Action query time = 4.220 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8154 +t=106: Selected seed 195 with value = 0.8154 +Query 1/1: Action query time = 3.392 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9934 +t=122: Selected seed 195 with value = 0.9934 +Query 1/1: Action query time = 4.383 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=138: Selected seed 195 with value = 0.9977 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=9--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 10... +Query 1/1: Action query time = 4.246 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3999 +t=10: Selected seed 195 with value = 0.3999 +Query 1/1: Action query time = 3.816 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4396 +t=26: Selected seed 195 with value = 0.4396 +Query 1/1: Action query time = 3.085 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5675 +t=42: Selected seed 195 with value = 0.5675 +Query 1/1: Action query time = 3.568 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5979 +t=58: Selected seed 195 with value = 0.5979 +Query 1/1: Action query time = 2.464 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7171 +t=74: Selected seed 195 with value = 0.7171 +Query 1/1: Action query time = 1.519 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8361 +t=90: Selected seed 195 with value = 0.8361 +Query 1/1: Action query time = 1.865 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9880 +t=106: Selected seed 195 with value = 0.9880 +Query 1/1: Action query time = 2.206 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=122: Selected seed 195 with value = 0.9992 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=10--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 11... +Query 1/1: Action query time = 2.134 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3907 +t=10: Selected seed 195 with value = 0.3907 +Query 1/1: Action query time = 3.106 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4453 +t=26: Selected seed 195 with value = 0.4453 +Query 1/1: Action query time = 3.219 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5738 +t=42: Selected seed 195 with value = 0.5738 +Query 1/1: Action query time = 3.092 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5718 +t=58: Selected seed 195 with value = 0.5718 +Query 1/1: Action query time = 2.647 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7307 +t=74: Selected seed 195 with value = 0.7307 +Query 1/1: Action query time = 2.702 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7838 +t=90: Selected seed 195 with value = 0.7838 +Query 1/1: Action query time = 1.921 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8878 +t=106: Selected seed 195 with value = 0.8878 +Query 1/1: Action query time = 1.463 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.930 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=138: Selected seed 195 with value = 0.9977 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=11--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 12... +Query 1/1: Action query time = 2.957 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3524 +t=10: Selected seed 195 with value = 0.3524 +Query 1/1: Action query time = 2.438 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4385 +t=26: Selected seed 195 with value = 0.4385 +Query 1/1: Action query time = 2.124 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4871 +t=42: Selected seed 195 with value = 0.4871 +Query 1/1: Action query time = 2.289 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5374 +t=58: Selected seed 195 with value = 0.5374 +Query 1/1: Action query time = 2.353 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6471 +t=74: Selected seed 195 with value = 0.6471 +Query 1/1: Action query time = 1.617 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7294 +t=90: Selected seed 195 with value = 0.7294 +Query 1/1: Action query time = 1.801 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8244 +t=106: Selected seed 195 with value = 0.8244 +Query 1/1: Action query time = 2.663 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9462 +t=122: Selected seed 195 with value = 0.9462 +Query 1/1: Action query time = 2.339 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=138: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s3/2026_08_03-17_39_22--with_future_img--episode=12--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 12 +Total successes: 12 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_39_23--v2i350_p1_t5_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_39_23--v2i350_p1_t5_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..c44a6d41fd00407f87ff7354570621697aaafe15 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_39_23--v2i350_p1_t5_s0.txt @@ -0,0 +1,603 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p1_t5_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,4,8,12,16,20,24,28,32,36,40,44,48', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 5.024 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3788 +t=10: Selected seed 195 with value = 0.3788 +Query 1/1: Action query time = 4.973 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4350 +t=26: Selected seed 195 with value = 0.4350 +Query 1/1: Action query time = 5.601 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5264 +t=42: Selected seed 195 with value = 0.5264 +Query 1/1: Action query time = 5.175 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5737 +t=58: Selected seed 195 with value = 0.5737 +Query 1/1: Action query time = 4.145 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6118 +t=74: Selected seed 195 with value = 0.6118 +Query 1/1: Action query time = 4.081 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7106 +t=90: Selected seed 195 with value = 0.7106 +Query 1/1: Action query time = 4.557 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9302 +t=106: Selected seed 195 with value = 0.9302 +Query 1/1: Action query time = 3.989 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.434 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 4.509 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3301 +t=10: Selected seed 195 with value = 0.3301 +Query 1/1: Action query time = 4.641 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4269 +t=26: Selected seed 195 with value = 0.4269 +Query 1/1: Action query time = 4.856 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5472 +t=42: Selected seed 195 with value = 0.5472 +Query 1/1: Action query time = 4.702 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6439 +t=58: Selected seed 195 with value = 0.6439 +Query 1/1: Action query time = 4.802 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6898 +t=74: Selected seed 195 with value = 0.6898 +Query 1/1: Action query time = 5.161 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7457 +t=90: Selected seed 195 with value = 0.7457 +Query 1/1: Action query time = 5.238 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9771 +t=106: Selected seed 195 with value = 0.9771 +Query 1/1: Action query time = 5.027 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.476 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=138: Selected seed 195 with value = 0.9976 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 1.999 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4035 +t=10: Selected seed 195 with value = 0.4035 +Query 1/1: Action query time = 4.259 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4291 +t=26: Selected seed 195 with value = 0.4291 +Query 1/1: Action query time = 4.968 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5174 +t=42: Selected seed 195 with value = 0.5174 +Query 1/1: Action query time = 5.298 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5535 +t=58: Selected seed 195 with value = 0.5535 +Query 1/1: Action query time = 5.272 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5619 +t=74: Selected seed 195 with value = 0.5619 +Query 1/1: Action query time = 4.589 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6725 +t=90: Selected seed 195 with value = 0.6725 +Query 1/1: Action query time = 4.750 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8027 +t=106: Selected seed 195 with value = 0.8027 +Query 1/1: Action query time = 3.671 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9243 +t=122: Selected seed 195 with value = 0.9243 +Query 1/1: Action query time = 3.122 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=138: Selected seed 195 with value = 0.9980 +Query 1/1: Action query time = 4.607 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=154: Selected seed 195 with value = 0.9991 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 4.411 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3762 +t=10: Selected seed 195 with value = 0.3762 +Query 1/1: Action query time = 5.129 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4438 +t=26: Selected seed 195 with value = 0.4438 +Query 1/1: Action query time = 4.930 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5405 +t=42: Selected seed 195 with value = 0.5405 +Query 1/1: Action query time = 4.221 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6306 +t=58: Selected seed 195 with value = 0.6306 +Query 1/1: Action query time = 5.346 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7349 +t=74: Selected seed 195 with value = 0.7349 +Query 1/1: Action query time = 3.121 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8245 +t=90: Selected seed 195 with value = 0.8245 +Query 1/1: Action query time = 5.050 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9488 +t=106: Selected seed 195 with value = 0.9488 +Query 1/1: Action query time = 4.682 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9950 +t=122: Selected seed 195 with value = 0.9950 +Query 1/1: Action query time = 5.517 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.337 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.012 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.599 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=186: Selected seed 195 with value = 0.9991 +Query 1/1: Action query time = 4.834 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=202: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 4.704 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.294 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.959 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.120 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=266: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 5.118 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9911 +t=282: Selected seed 195 with value = 0.9911 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=4--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 5... +Query 1/1: Action query time = 4.811 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3655 +t=10: Selected seed 195 with value = 0.3655 +Query 1/1: Action query time = 5.759 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4241 +t=26: Selected seed 195 with value = 0.4241 +Query 1/1: Action query time = 2.972 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5451 +t=42: Selected seed 195 with value = 0.5451 +Query 1/1: Action query time = 5.338 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6320 +t=58: Selected seed 195 with value = 0.6320 +Query 1/1: Action query time = 4.720 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6805 +t=74: Selected seed 195 with value = 0.6805 +Query 1/1: Action query time = 4.489 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7988 +t=90: Selected seed 195 with value = 0.7988 +Query 1/1: Action query time = 4.398 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=106: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 3.618 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.458 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=5--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 6... +Query 1/1: Action query time = 4.554 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3701 +t=10: Selected seed 195 with value = 0.3701 +Query 1/1: Action query time = 5.436 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4286 +t=26: Selected seed 195 with value = 0.4286 +Query 1/1: Action query time = 4.930 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5169 +t=42: Selected seed 195 with value = 0.5169 +Query 1/1: Action query time = 5.237 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5931 +t=58: Selected seed 195 with value = 0.5931 +Query 1/1: Action query time = 4.537 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6626 +t=74: Selected seed 195 with value = 0.6626 +Query 1/1: Action query time = 5.043 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7473 +t=90: Selected seed 195 with value = 0.7473 +Query 1/1: Action query time = 5.457 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9584 +t=106: Selected seed 195 with value = 0.9584 +Query 1/1: Action query time = 4.243 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=122: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 4.142 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=138: Selected seed 195 with value = 0.9998 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=6--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 7... +Query 1/1: Action query time = 4.279 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3667 +t=10: Selected seed 195 with value = 0.3667 +Query 1/1: Action query time = 5.339 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4376 +t=26: Selected seed 195 with value = 0.4376 +Query 1/1: Action query time = 5.534 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5467 +t=42: Selected seed 195 with value = 0.5467 +Query 1/1: Action query time = 4.276 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6269 +t=58: Selected seed 195 with value = 0.6269 +Query 1/1: Action query time = 4.236 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7336 +t=74: Selected seed 195 with value = 0.7336 +Query 1/1: Action query time = 4.530 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8961 +t=90: Selected seed 195 with value = 0.8961 +Query 1/1: Action query time = 5.434 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.468 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=122: Selected seed 195 with value = 0.9991 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=7--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 8... +Query 1/1: Action query time = 4.347 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3689 +t=10: Selected seed 195 with value = 0.3689 +Query 1/1: Action query time = 4.594 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4415 +t=26: Selected seed 195 with value = 0.4415 +Query 1/1: Action query time = 4.656 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5046 +t=42: Selected seed 195 with value = 0.5046 +Query 1/1: Action query time = 4.220 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5710 +t=58: Selected seed 195 with value = 0.5710 +Query 1/1: Action query time = 3.028 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6165 +t=74: Selected seed 195 with value = 0.6165 +Query 1/1: Action query time = 4.161 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6920 +t=90: Selected seed 195 with value = 0.6920 +Query 1/1: Action query time = 4.653 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8341 +t=106: Selected seed 195 with value = 0.8341 +Query 1/1: Action query time = 3.810 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9355 +t=122: Selected seed 195 with value = 0.9355 +Query 1/1: Action query time = 3.172 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9965 +t=138: Selected seed 195 with value = 0.9965 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=8--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 9... +Query 1/1: Action query time = 1.884 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4097 +t=10: Selected seed 195 with value = 0.4097 +Query 1/1: Action query time = 2.729 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4846 +t=26: Selected seed 195 with value = 0.4846 +Query 1/1: Action query time = 3.150 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5624 +t=42: Selected seed 195 with value = 0.5624 +Query 1/1: Action query time = 3.733 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6014 +t=58: Selected seed 195 with value = 0.6014 +Query 1/1: Action query time = 2.993 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7370 +t=74: Selected seed 195 with value = 0.7370 +Query 1/1: Action query time = 2.964 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8318 +t=90: Selected seed 195 with value = 0.8318 +Query 1/1: Action query time = 2.946 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9341 +t=106: Selected seed 195 with value = 0.9341 +Query 1/1: Action query time = 2.250 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.320 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=138: Selected seed 195 with value = 0.9994 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=9--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 10... +Query 1/1: Action query time = 1.914 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3917 +t=10: Selected seed 195 with value = 0.3917 +Query 1/1: Action query time = 1.342 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4396 +t=26: Selected seed 195 with value = 0.4396 +Query 1/1: Action query time = 1.472 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5056 +t=42: Selected seed 195 with value = 0.5056 +Query 1/1: Action query time = 3.016 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6598 +t=58: Selected seed 195 with value = 0.6598 +Query 1/1: Action query time = 2.628 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7487 +t=74: Selected seed 195 with value = 0.7487 +Query 1/1: Action query time = 2.360 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8921 +t=90: Selected seed 195 with value = 0.8921 +Query 1/1: Action query time = 2.312 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.384 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=10--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 11... +Query 1/1: Action query time = 2.708 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3672 +t=10: Selected seed 195 with value = 0.3672 +Query 1/1: Action query time = 2.281 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4155 +t=26: Selected seed 195 with value = 0.4155 +Query 1/1: Action query time = 0.963 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5372 +t=42: Selected seed 195 with value = 0.5372 +Query 1/1: Action query time = 0.982 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5379 +t=58: Selected seed 195 with value = 0.5379 +Query 1/1: Action query time = 1.357 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6221 +t=74: Selected seed 195 with value = 0.6221 +Query 1/1: Action query time = 1.413 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6786 +t=90: Selected seed 195 with value = 0.6786 +Query 1/1: Action query time = 1.943 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8595 +t=106: Selected seed 195 with value = 0.8595 +Query 1/1: Action query time = 2.395 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.586 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=138: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=11--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 12... +Query 1/1: Action query time = 2.535 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3652 +t=10: Selected seed 195 with value = 0.3652 +Query 1/1: Action query time = 2.185 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4090 +t=26: Selected seed 195 with value = 0.4090 +Query 1/1: Action query time = 1.960 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5091 +t=42: Selected seed 195 with value = 0.5091 +Query 1/1: Action query time = 0.959 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5805 +t=58: Selected seed 195 with value = 0.5805 +Query 1/1: Action query time = 0.971 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6734 +t=74: Selected seed 195 with value = 0.6734 +Query 1/1: Action query time = 0.993 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8440 +t=90: Selected seed 195 with value = 0.8440 +Query 1/1: Action query time = 1.557 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9787 +t=106: Selected seed 195 with value = 0.9787 +Query 1/1: Action query time = 1.501 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.404 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9357 +t=138: Selected seed 195 with value = 0.9357 +Query 1/1: Action query time = 1.387 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9811 +t=154: Selected seed 195 with value = 0.9811 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=12--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 13... +Query 1/1: Action query time = 1.448 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3793 +t=10: Selected seed 195 with value = 0.3793 +Query 1/1: Action query time = 1.518 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4365 +t=26: Selected seed 195 with value = 0.4365 +Query 1/1: Action query time = 0.953 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5341 +t=42: Selected seed 195 with value = 0.5341 +Query 1/1: Action query time = 0.958 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6520 +t=58: Selected seed 195 with value = 0.6520 +Query 1/1: Action query time = 0.950 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7517 +t=74: Selected seed 195 with value = 0.7517 +Query 1/1: Action query time = 0.977 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8664 +t=90: Selected seed 195 with value = 0.8664 +Query 1/1: Action query time = 0.971 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9819 +t=106: Selected seed 195 with value = 0.9819 +Query 1/1: Action query time = 0.968 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p1_t5_s0/2026_08_03-17_39_23--with_future_img--episode=13--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 13 +Total successes: 13 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_49_38--v2i350_p2_t0_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_49_38--v2i350_p2_t0_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..53bc3895d19a6bcd3874fdedb33aefde0eab4a33 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_49_38--v2i350_p2_t0_s1.txt @@ -0,0 +1,567 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='0', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p2_t0_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [0] +Using default initial states + +Task: open the middle drawer of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.229 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4010 +t=10: Selected seed 195 with value = 0.4010 +Query 1/1: Action query time = 3.440 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4868 +t=26: Selected seed 195 with value = 0.4868 +Query 1/1: Action query time = 4.621 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5593 +t=42: Selected seed 195 with value = 0.5593 +Query 1/1: Action query time = 4.644 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6447 +t=58: Selected seed 195 with value = 0.6447 +Query 1/1: Action query time = 4.945 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7518 +t=74: Selected seed 195 with value = 0.7518 +Query 1/1: Action query time = 5.329 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8659 +t=90: Selected seed 195 with value = 0.8659 +Query 1/1: Action query time = 5.360 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.859 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9946 +t=122: Selected seed 195 with value = 0.9946 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=1--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.089 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4065 +t=10: Selected seed 195 with value = 0.4065 +Query 1/1: Action query time = 3.842 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4486 +t=26: Selected seed 195 with value = 0.4486 +Query 1/1: Action query time = 5.184 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5604 +t=42: Selected seed 195 with value = 0.5604 +Query 1/1: Action query time = 5.622 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6593 +t=58: Selected seed 195 with value = 0.6593 +Query 1/1: Action query time = 4.999 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7557 +t=74: Selected seed 195 with value = 0.7557 +Query 1/1: Action query time = 4.645 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8605 +t=90: Selected seed 195 with value = 0.8605 +Query 1/1: Action query time = 5.202 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9932 +t=106: Selected seed 195 with value = 0.9932 +Query 1/1: Action query time = 4.643 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=2--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 3... +Query 1/1: Action query time = 5.020 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4402 +t=10: Selected seed 195 with value = 0.4402 +Query 1/1: Action query time = 3.673 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5148 +t=26: Selected seed 195 with value = 0.5148 +Query 1/1: Action query time = 3.795 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5712 +t=42: Selected seed 195 with value = 0.5712 +Query 1/1: Action query time = 3.937 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6540 +t=58: Selected seed 195 with value = 0.6540 +Query 1/1: Action query time = 5.513 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7484 +t=74: Selected seed 195 with value = 0.7484 +Query 1/1: Action query time = 5.033 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9131 +t=90: Selected seed 195 with value = 0.9131 +Query 1/1: Action query time = 5.072 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.603 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=3--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 4... +Query 1/1: Action query time = 3.910 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4014 +t=10: Selected seed 195 with value = 0.4014 +Query 1/1: Action query time = 3.083 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4707 +t=26: Selected seed 195 with value = 0.4707 +Query 1/1: Action query time = 4.733 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5454 +t=42: Selected seed 195 with value = 0.5454 +Query 1/1: Action query time = 5.239 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6318 +t=58: Selected seed 195 with value = 0.6318 +Query 1/1: Action query time = 5.212 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7617 +t=74: Selected seed 195 with value = 0.7617 +Query 1/1: Action query time = 5.409 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9050 +t=90: Selected seed 195 with value = 0.9050 +Query 1/1: Action query time = 4.382 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9894 +t=106: Selected seed 195 with value = 0.9894 +Query 1/1: Action query time = 4.541 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=122: Selected seed 195 with value = 0.9999 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=4--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 5... +Query 1/1: Action query time = 5.228 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4077 +t=10: Selected seed 195 with value = 0.4077 +Query 1/1: Action query time = 4.613 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4746 +t=26: Selected seed 195 with value = 0.4746 +Query 1/1: Action query time = 5.285 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5760 +t=42: Selected seed 195 with value = 0.5760 +Query 1/1: Action query time = 4.567 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6175 +t=58: Selected seed 195 with value = 0.6175 +Query 1/1: Action query time = 5.028 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7121 +t=74: Selected seed 195 with value = 0.7121 +Query 1/1: Action query time = 4.961 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8735 +t=90: Selected seed 195 with value = 0.8735 +Query 1/1: Action query time = 4.772 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9940 +t=106: Selected seed 195 with value = 0.9940 +Query 1/1: Action query time = 4.129 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=5--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: open the middle drawer of the cabinet +Starting episode 6... +Query 1/1: Action query time = 4.350 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3730 +t=10: Selected seed 195 with value = 0.3730 +Query 1/1: Action query time = 3.790 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3405 +t=26: Selected seed 195 with value = 0.3405 +Query 1/1: Action query time = 4.678 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4278 +t=42: Selected seed 195 with value = 0.4278 +Query 1/1: Action query time = 5.042 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5627 +t=58: Selected seed 195 with value = 0.5627 +Query 1/1: Action query time = 5.009 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7446 +t=74: Selected seed 195 with value = 0.7446 +Query 1/1: Action query time = 4.816 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8387 +t=90: Selected seed 195 with value = 0.8387 +Query 1/1: Action query time = 5.687 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.418 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9471 +t=122: Selected seed 195 with value = 0.9471 +Query 1/1: Action query time = 4.469 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7578 +t=138: Selected seed 195 with value = 0.7578 +Query 1/1: Action query time = 3.338 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7829 +t=154: Selected seed 195 with value = 0.7829 +Query 1/1: Action query time = 4.351 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7713 +t=170: Selected seed 195 with value = 0.7713 +Query 1/1: Action query time = 4.723 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7474 +t=186: Selected seed 195 with value = 0.7474 +Query 1/1: Action query time = 4.663 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7540 +t=202: Selected seed 195 with value = 0.7540 +Query 1/1: Action query time = 4.690 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7575 +t=218: Selected seed 195 with value = 0.7575 +Query 1/1: Action query time = 5.307 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7534 +t=234: Selected seed 195 with value = 0.7534 +Query 1/1: Action query time = 4.939 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7682 +t=250: Selected seed 195 with value = 0.7682 +Query 1/1: Action query time = 4.127 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7871 +t=266: Selected seed 195 with value = 0.7871 +Query 1/1: Action query time = 4.726 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8163 +t=282: Selected seed 195 with value = 0.8163 +Query 1/1: Action query time = 3.265 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8398 +t=298: Selected seed 195 with value = 0.8398 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=6--success=False--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: False +# episodes completed so far: 6 +# successes: 5 (83.3%) + +Task: open the middle drawer of the cabinet +Starting episode 7... +Query 1/1: Action query time = 4.149 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3815 +t=10: Selected seed 195 with value = 0.3815 +Query 1/1: Action query time = 4.553 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4551 +t=26: Selected seed 195 with value = 0.4551 +Query 1/1: Action query time = 5.524 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5445 +t=42: Selected seed 195 with value = 0.5445 +Query 1/1: Action query time = 4.658 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6256 +t=58: Selected seed 195 with value = 0.6256 +Query 1/1: Action query time = 4.004 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7619 +t=74: Selected seed 195 with value = 0.7619 +Query 1/1: Action query time = 4.333 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8953 +t=90: Selected seed 195 with value = 0.8953 +Query 1/1: Action query time = 4.339 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9958 +t=106: Selected seed 195 with value = 0.9958 +Query 1/1: Action query time = 5.093 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=122: Selected seed 195 with value = 0.9999 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=7--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 7 +# successes: 6 (85.7%) + +Task: open the middle drawer of the cabinet +Starting episode 8... +Query 1/1: Action query time = 3.811 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3835 +t=10: Selected seed 195 with value = 0.3835 +Query 1/1: Action query time = 5.248 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4346 +t=26: Selected seed 195 with value = 0.4346 +Query 1/1: Action query time = 4.250 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5218 +t=42: Selected seed 195 with value = 0.5218 +Query 1/1: Action query time = 5.422 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6186 +t=58: Selected seed 195 with value = 0.6186 +Query 1/1: Action query time = 5.084 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7591 +t=74: Selected seed 195 with value = 0.7591 +Query 1/1: Action query time = 5.019 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8967 +t=90: Selected seed 195 with value = 0.8967 +Query 1/1: Action query time = 2.593 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.849 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=8--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 8 +# successes: 7 (87.5%) + +Task: open the middle drawer of the cabinet +Starting episode 9... +Query 1/1: Action query time = 4.143 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3744 +t=10: Selected seed 195 with value = 0.3744 +Query 1/1: Action query time = 3.918 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4319 +t=26: Selected seed 195 with value = 0.4319 +Query 1/1: Action query time = 4.760 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5246 +t=42: Selected seed 195 with value = 0.5246 +Query 1/1: Action query time = 4.091 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6489 +t=58: Selected seed 195 with value = 0.6489 +Query 1/1: Action query time = 3.874 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7984 +t=74: Selected seed 195 with value = 0.7984 +Query 1/1: Action query time = 2.912 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8632 +t=90: Selected seed 195 with value = 0.8632 +Query 1/1: Action query time = 3.561 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9168 +t=106: Selected seed 195 with value = 0.9168 +Query 1/1: Action query time = 4.865 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.836 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=9--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 9 +# successes: 8 (88.9%) + +Task: open the middle drawer of the cabinet +Starting episode 10... +Query 1/1: Action query time = 4.905 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3823 +t=10: Selected seed 195 with value = 0.3823 +Query 1/1: Action query time = 3.438 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4551 +t=26: Selected seed 195 with value = 0.4551 +Query 1/1: Action query time = 4.995 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5358 +t=42: Selected seed 195 with value = 0.5358 +Query 1/1: Action query time = 4.305 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6403 +t=58: Selected seed 195 with value = 0.6403 +Query 1/1: Action query time = 3.151 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7497 +t=74: Selected seed 195 with value = 0.7497 +Query 1/1: Action query time = 3.167 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9065 +t=90: Selected seed 195 with value = 0.9065 +Query 1/1: Action query time = 3.642 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9898 +t=106: Selected seed 195 with value = 0.9898 +Query 1/1: Action query time = 3.232 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=10--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 10 +# successes: 9 (90.0%) + +Task: open the middle drawer of the cabinet +Starting episode 11... +Query 1/1: Action query time = 3.479 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3878 +t=10: Selected seed 195 with value = 0.3878 +Query 1/1: Action query time = 3.984 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4535 +t=26: Selected seed 195 with value = 0.4535 +Query 1/1: Action query time = 2.435 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5413 +t=42: Selected seed 195 with value = 0.5413 +Query 1/1: Action query time = 2.537 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6583 +t=58: Selected seed 195 with value = 0.6583 +Query 1/1: Action query time = 2.764 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7517 +t=74: Selected seed 195 with value = 0.7517 +Query 1/1: Action query time = 2.103 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8903 +t=90: Selected seed 195 with value = 0.8903 +Query 1/1: Action query time = 2.267 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=106: Selected seed 195 with value = 0.9954 +Query 1/1: Action query time = 2.768 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=11--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 11 +# successes: 10 (90.9%) + +Task: open the middle drawer of the cabinet +Starting episode 12... +Query 1/1: Action query time = 1.393 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3761 +t=10: Selected seed 195 with value = 0.3761 +Query 1/1: Action query time = 1.604 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4547 +t=26: Selected seed 195 with value = 0.4547 +Query 1/1: Action query time = 1.269 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5501 +t=42: Selected seed 195 with value = 0.5501 +Query 1/1: Action query time = 1.010 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6322 +t=58: Selected seed 195 with value = 0.6322 +Query 1/1: Action query time = 1.007 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7468 +t=74: Selected seed 195 with value = 0.7468 +Query 1/1: Action query time = 1.512 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9173 +t=90: Selected seed 195 with value = 0.9173 +Query 1/1: Action query time = 1.519 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.415 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=12--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 12 +# successes: 11 (91.7%) + +Task: open the middle drawer of the cabinet +Starting episode 13... +Query 1/1: Action query time = 1.221 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3803 +t=10: Selected seed 195 with value = 0.3803 +Query 1/1: Action query time = 1.034 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4699 +t=26: Selected seed 195 with value = 0.4699 +Query 1/1: Action query time = 1.949 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5336 +t=42: Selected seed 195 with value = 0.5336 +Query 1/1: Action query time = 1.902 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6713 +t=58: Selected seed 195 with value = 0.6713 +Query 1/1: Action query time = 1.694 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7988 +t=74: Selected seed 195 with value = 0.7988 +Query 1/1: Action query time = 1.995 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9299 +t=90: Selected seed 195 with value = 0.9299 +Query 1/1: Action query time = 1.643 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=106: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 1.492 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t0_s1/2026_08_03-17_49_38--with_future_img--episode=13--success=True--task=open_the_middle_drawer_of_the_cabin.mp4 +Success: True +# episodes completed so far: 13 +# successes: 12 (92.3%) +Current task success rate: 0.9230769230769231 +Current total success rate: 0.9230769230769231 +Final results: +Total episodes: 13 +Total successes: 12 +Overall success rate: 0.9231 (92.3%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_49_40--v2i350_p2_t3_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_49_40--v2i350_p2_t3_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..40fcbc979007673cfbd5b5a82afcbe6e56fcbbbe --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_49_40--v2i350_p2_t3_s1.txt @@ -0,0 +1,759 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p2_t3_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 3.777 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2223 +t=10: Selected seed 195 with value = 0.2223 +Query 1/1: Action query time = 5.558 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2478 +t=26: Selected seed 195 with value = 0.2478 +Query 1/1: Action query time = 4.347 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3040 +t=42: Selected seed 195 with value = 0.3040 +Query 1/1: Action query time = 4.721 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3540 +t=58: Selected seed 195 with value = 0.3540 +Query 1/1: Action query time = 5.075 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4197 +t=74: Selected seed 195 with value = 0.4197 +Query 1/1: Action query time = 5.211 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4889 +t=90: Selected seed 195 with value = 0.4889 +Query 1/1: Action query time = 3.765 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6175 +t=106: Selected seed 195 with value = 0.6175 +Query 1/1: Action query time = 4.004 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7262 +t=122: Selected seed 195 with value = 0.7262 +Query 1/1: Action query time = 4.371 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8623 +t=138: Selected seed 195 with value = 0.8623 +Query 1/1: Action query time = 4.870 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 3.757 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2103 +t=10: Selected seed 195 with value = 0.2103 +Query 1/1: Action query time = 3.900 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2626 +t=26: Selected seed 195 with value = 0.2626 +Query 1/1: Action query time = 3.895 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2940 +t=42: Selected seed 195 with value = 0.2940 +Query 1/1: Action query time = 4.065 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3379 +t=58: Selected seed 195 with value = 0.3379 +Query 1/1: Action query time = 4.655 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4065 +t=74: Selected seed 195 with value = 0.4065 +Query 1/1: Action query time = 5.256 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4489 +t=90: Selected seed 195 with value = 0.4489 +Query 1/1: Action query time = 4.449 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5667 +t=106: Selected seed 195 with value = 0.5667 +Query 1/1: Action query time = 5.545 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6528 +t=122: Selected seed 195 with value = 0.6528 +Query 1/1: Action query time = 4.729 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7632 +t=138: Selected seed 195 with value = 0.7632 +Query 1/1: Action query time = 4.765 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8769 +t=154: Selected seed 195 with value = 0.8769 +Query 1/1: Action query time = 4.769 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9947 +t=170: Selected seed 195 with value = 0.9947 +Query 1/1: Action query time = 5.474 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 4.185 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2280 +t=10: Selected seed 195 with value = 0.2280 +Query 1/1: Action query time = 3.737 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2727 +t=26: Selected seed 195 with value = 0.2727 +Query 1/1: Action query time = 3.659 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3032 +t=42: Selected seed 195 with value = 0.3032 +Query 1/1: Action query time = 5.022 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3480 +t=58: Selected seed 195 with value = 0.3480 +Query 1/1: Action query time = 5.114 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4182 +t=74: Selected seed 195 with value = 0.4182 +Query 1/1: Action query time = 4.474 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4782 +t=90: Selected seed 195 with value = 0.4782 +Query 1/1: Action query time = 5.157 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5815 +t=106: Selected seed 195 with value = 0.5815 +Query 1/1: Action query time = 4.529 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6815 +t=122: Selected seed 195 with value = 0.6815 +Query 1/1: Action query time = 5.145 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6157 +t=138: Selected seed 195 with value = 0.6157 +Query 1/1: Action query time = 4.235 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7551 +t=154: Selected seed 195 with value = 0.7551 +Query 1/1: Action query time = 4.970 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8638 +t=170: Selected seed 195 with value = 0.8638 +Query 1/1: Action query time = 4.740 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9878 +t=186: Selected seed 195 with value = 0.9878 +Query 1/1: Action query time = 3.546 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=202: Selected seed 195 with value = 0.9992 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 5.419 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2172 +t=10: Selected seed 195 with value = 0.2172 +Query 1/1: Action query time = 4.024 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2614 +t=26: Selected seed 195 with value = 0.2614 +Query 1/1: Action query time = 4.765 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3017 +t=42: Selected seed 195 with value = 0.3017 +Query 1/1: Action query time = 4.524 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3390 +t=58: Selected seed 195 with value = 0.3390 +Query 1/1: Action query time = 4.314 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4183 +t=74: Selected seed 195 with value = 0.4183 +Query 1/1: Action query time = 4.844 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4603 +t=90: Selected seed 195 with value = 0.4603 +Query 1/1: Action query time = 4.779 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5667 +t=106: Selected seed 195 with value = 0.5667 +Query 1/1: Action query time = 5.022 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6375 +t=122: Selected seed 195 with value = 0.6375 +Query 1/1: Action query time = 4.961 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8112 +t=138: Selected seed 195 with value = 0.8112 +Query 1/1: Action query time = 4.537 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8983 +t=154: Selected seed 195 with value = 0.8983 +Query 1/1: Action query time = 4.171 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=4--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 5... +Query 1/1: Action query time = 4.877 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2322 +t=10: Selected seed 195 with value = 0.2322 +Query 1/1: Action query time = 5.363 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2450 +t=26: Selected seed 195 with value = 0.2450 +Query 1/1: Action query time = 5.655 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3154 +t=42: Selected seed 195 with value = 0.3154 +Query 1/1: Action query time = 5.178 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3538 +t=58: Selected seed 195 with value = 0.3538 +Query 1/1: Action query time = 5.274 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3962 +t=74: Selected seed 195 with value = 0.3962 +Query 1/1: Action query time = 4.171 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4628 +t=90: Selected seed 195 with value = 0.4628 +Query 1/1: Action query time = 4.779 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5662 +t=106: Selected seed 195 with value = 0.5662 +Query 1/1: Action query time = 5.397 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6633 +t=122: Selected seed 195 with value = 0.6633 +Query 1/1: Action query time = 3.547 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7704 +t=138: Selected seed 195 with value = 0.7704 +Query 1/1: Action query time = 2.537 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9340 +t=154: Selected seed 195 with value = 0.9340 +Query 1/1: Action query time = 4.293 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9672 +t=170: Selected seed 195 with value = 0.9672 +Query 1/1: Action query time = 4.043 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9445 +t=186: Selected seed 195 with value = 0.9445 +Query 1/1: Action query time = 4.134 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3311 +t=202: Selected seed 195 with value = 0.3311 +Query 1/1: Action query time = 3.424 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3335 +t=218: Selected seed 195 with value = 0.3335 +Query 1/1: Action query time = 5.651 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9314 +t=234: Selected seed 195 with value = 0.9314 +Query 1/1: Action query time = 5.413 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9198 +t=250: Selected seed 195 with value = 0.9198 +Query 1/1: Action query time = 4.261 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8962 +t=266: Selected seed 195 with value = 0.8962 +Query 1/1: Action query time = 4.684 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7544 +t=282: Selected seed 195 with value = 0.7544 +Query 1/1: Action query time = 4.615 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4171 +t=298: Selected seed 195 with value = 0.4171 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=5--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 5 +# successes: 4 (80.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 6... +Query 1/1: Action query time = 4.145 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2169 +t=10: Selected seed 195 with value = 0.2169 +Query 1/1: Action query time = 5.289 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2329 +t=26: Selected seed 195 with value = 0.2329 +Query 1/1: Action query time = 3.601 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3004 +t=42: Selected seed 195 with value = 0.3004 +Query 1/1: Action query time = 2.868 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3471 +t=58: Selected seed 195 with value = 0.3471 +Query 1/1: Action query time = 4.923 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4054 +t=74: Selected seed 195 with value = 0.4054 +Query 1/1: Action query time = 4.882 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4584 +t=90: Selected seed 195 with value = 0.4584 +Query 1/1: Action query time = 5.055 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5630 +t=106: Selected seed 195 with value = 0.5630 +Query 1/1: Action query time = 4.239 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6809 +t=122: Selected seed 195 with value = 0.6809 +Query 1/1: Action query time = 4.831 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7583 +t=138: Selected seed 195 with value = 0.7583 +Query 1/1: Action query time = 5.036 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8715 +t=154: Selected seed 195 with value = 0.8715 +Query 1/1: Action query time = 5.720 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9817 +t=170: Selected seed 195 with value = 0.9817 +Query 1/1: Action query time = 5.299 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=6--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 6 +# successes: 5 (83.3%) + +Task: open the top drawer and put the bowl inside +Starting episode 7... +Query 1/1: Action query time = 4.121 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2145 +t=10: Selected seed 195 with value = 0.2145 +Query 1/1: Action query time = 5.162 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2558 +t=26: Selected seed 195 with value = 0.2558 +Query 1/1: Action query time = 3.501 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2982 +t=42: Selected seed 195 with value = 0.2982 +Query 1/1: Action query time = 3.924 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3345 +t=58: Selected seed 195 with value = 0.3345 +Query 1/1: Action query time = 5.310 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3930 +t=74: Selected seed 195 with value = 0.3930 +Query 1/1: Action query time = 6.058 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4654 +t=90: Selected seed 195 with value = 0.4654 +Query 1/1: Action query time = 5.358 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5525 +t=106: Selected seed 195 with value = 0.5525 +Query 1/1: Action query time = 3.817 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6488 +t=122: Selected seed 195 with value = 0.6488 +Query 1/1: Action query time = 4.502 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7812 +t=138: Selected seed 195 with value = 0.7812 +Query 1/1: Action query time = 5.011 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8704 +t=154: Selected seed 195 with value = 0.8704 +Query 1/1: Action query time = 3.937 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9796 +t=170: Selected seed 195 with value = 0.9796 +Query 1/1: Action query time = 4.225 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=7--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 7 +# successes: 6 (85.7%) + +Task: open the top drawer and put the bowl inside +Starting episode 8... +Query 1/1: Action query time = 5.045 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2170 +t=10: Selected seed 195 with value = 0.2170 +Query 1/1: Action query time = 2.923 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2590 +t=26: Selected seed 195 with value = 0.2590 +Query 1/1: Action query time = 3.996 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2990 +t=42: Selected seed 195 with value = 0.2990 +Query 1/1: Action query time = 5.053 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3421 +t=58: Selected seed 195 with value = 0.3421 +Query 1/1: Action query time = 4.675 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4006 +t=74: Selected seed 195 with value = 0.4006 +Query 1/1: Action query time = 4.797 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4553 +t=90: Selected seed 195 with value = 0.4553 +Query 1/1: Action query time = 4.357 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5647 +t=106: Selected seed 195 with value = 0.5647 +Query 1/1: Action query time = 5.079 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6996 +t=122: Selected seed 195 with value = 0.6996 +Query 1/1: Action query time = 4.716 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8085 +t=138: Selected seed 195 with value = 0.8085 +Query 1/1: Action query time = 4.658 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9172 +t=154: Selected seed 195 with value = 0.9172 +Query 1/1: Action query time = 3.903 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9975 +t=170: Selected seed 195 with value = 0.9975 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=8--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 8 +# successes: 7 (87.5%) + +Task: open the top drawer and put the bowl inside +Starting episode 9... +Query 1/1: Action query time = 4.102 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2165 +t=10: Selected seed 195 with value = 0.2165 +Query 1/1: Action query time = 3.950 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2553 +t=26: Selected seed 195 with value = 0.2553 +Query 1/1: Action query time = 3.968 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2946 +t=42: Selected seed 195 with value = 0.2946 +Query 1/1: Action query time = 3.849 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3317 +t=58: Selected seed 195 with value = 0.3317 +Query 1/1: Action query time = 2.445 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3831 +t=74: Selected seed 195 with value = 0.3831 +Query 1/1: Action query time = 2.726 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4610 +t=90: Selected seed 195 with value = 0.4610 +Query 1/1: Action query time = 3.717 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5451 +t=106: Selected seed 195 with value = 0.5451 +Query 1/1: Action query time = 3.890 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6084 +t=122: Selected seed 195 with value = 0.6084 +Query 1/1: Action query time = 3.894 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7065 +t=138: Selected seed 195 with value = 0.7065 +Query 1/1: Action query time = 2.261 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8594 +t=154: Selected seed 195 with value = 0.8594 +Query 1/1: Action query time = 1.620 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9771 +t=170: Selected seed 195 with value = 0.9771 +Query 1/1: Action query time = 1.522 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=186: Selected seed 195 with value = 0.9990 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=9--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 9 +# successes: 8 (88.9%) + +Task: open the top drawer and put the bowl inside +Starting episode 10... +Query 1/1: Action query time = 3.926 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2030 +t=10: Selected seed 195 with value = 0.2030 +Query 1/1: Action query time = 3.190 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2659 +t=26: Selected seed 195 with value = 0.2659 +Query 1/1: Action query time = 2.716 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2993 +t=42: Selected seed 195 with value = 0.2993 +Query 1/1: Action query time = 2.412 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3360 +t=58: Selected seed 195 with value = 0.3360 +Query 1/1: Action query time = 2.769 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4007 +t=74: Selected seed 195 with value = 0.4007 +Query 1/1: Action query time = 3.364 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4667 +t=90: Selected seed 195 with value = 0.4667 +Query 1/1: Action query time = 3.448 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5647 +t=106: Selected seed 195 with value = 0.5647 +Query 1/1: Action query time = 3.444 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6999 +t=122: Selected seed 195 with value = 0.6999 +Query 1/1: Action query time = 2.926 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8101 +t=138: Selected seed 195 with value = 0.8101 +Query 1/1: Action query time = 1.649 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8930 +t=154: Selected seed 195 with value = 0.8930 +Query 1/1: Action query time = 1.377 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=170: Selected seed 195 with value = 0.9963 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=10--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 10 +# successes: 9 (90.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 11... +Query 1/1: Action query time = 2.436 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2095 +t=10: Selected seed 195 with value = 0.2095 +Query 1/1: Action query time = 2.359 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2445 +t=26: Selected seed 195 with value = 0.2445 +Query 1/1: Action query time = 2.308 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3117 +t=42: Selected seed 195 with value = 0.3117 +Query 1/1: Action query time = 2.357 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3570 +t=58: Selected seed 195 with value = 0.3570 +Query 1/1: Action query time = 2.376 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4153 +t=74: Selected seed 195 with value = 0.4153 +Query 1/1: Action query time = 2.399 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5114 +t=90: Selected seed 195 with value = 0.5114 +Query 1/1: Action query time = 2.381 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6133 +t=106: Selected seed 195 with value = 0.6133 +Query 1/1: Action query time = 2.398 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7944 +t=122: Selected seed 195 with value = 0.7944 +Query 1/1: Action query time = 2.308 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8660 +t=138: Selected seed 195 with value = 0.8660 +Query 1/1: Action query time = 1.480 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=154: Selected seed 195 with value = 0.9980 +Query 1/1: Action query time = 1.623 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9540 +t=170: Selected seed 195 with value = 0.9540 +Query 1/1: Action query time = 2.172 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9538 +t=186: Selected seed 195 with value = 0.9538 +Query 1/1: Action query time = 1.253 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9460 +t=202: Selected seed 195 with value = 0.9460 +Query 1/1: Action query time = 1.194 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3328 +t=218: Selected seed 195 with value = 0.3328 +Query 1/1: Action query time = 1.196 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9317 +t=234: Selected seed 195 with value = 0.9317 +Query 1/1: Action query time = 1.792 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9204 +t=250: Selected seed 195 with value = 0.9204 +Query 1/1: Action query time = 1.758 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9229 +t=266: Selected seed 195 with value = 0.9229 +Query 1/1: Action query time = 1.756 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3163 +t=282: Selected seed 195 with value = 0.3163 +Query 1/1: Action query time = 1.700 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9130 +t=298: Selected seed 195 with value = 0.9130 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=11--success=False--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: False +# episodes completed so far: 11 +# successes: 9 (81.8%) + +Task: open the top drawer and put the bowl inside +Starting episode 12... +Query 1/1: Action query time = 1.423 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2142 +t=10: Selected seed 195 with value = 0.2142 +Query 1/1: Action query time = 1.373 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2633 +t=26: Selected seed 195 with value = 0.2633 +Query 1/1: Action query time = 1.364 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3009 +t=42: Selected seed 195 with value = 0.3009 +Query 1/1: Action query time = 1.364 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3380 +t=58: Selected seed 195 with value = 0.3380 +Query 1/1: Action query time = 1.363 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4076 +t=74: Selected seed 195 with value = 0.4076 +Query 1/1: Action query time = 1.354 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4597 +t=90: Selected seed 195 with value = 0.4597 +Query 1/1: Action query time = 1.362 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5608 +t=106: Selected seed 195 with value = 0.5608 +Query 1/1: Action query time = 1.311 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6156 +t=122: Selected seed 195 with value = 0.6156 +Query 1/1: Action query time = 1.464 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7888 +t=138: Selected seed 195 with value = 0.7888 +Query 1/1: Action query time = 1.460 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8808 +t=154: Selected seed 195 with value = 0.8808 +Query 1/1: Action query time = 1.464 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=12--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 12 +# successes: 10 (83.3%) + +Task: open the top drawer and put the bowl inside +Starting episode 13... +Query 1/1: Action query time = 0.982 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2047 +t=10: Selected seed 195 with value = 0.2047 +Query 1/1: Action query time = 1.058 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2376 +t=26: Selected seed 195 with value = 0.2376 +Query 1/1: Action query time = 1.266 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3111 +t=42: Selected seed 195 with value = 0.3111 +Query 1/1: Action query time = 1.232 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3550 +t=58: Selected seed 195 with value = 0.3550 +Query 1/1: Action query time = 1.259 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4180 +t=74: Selected seed 195 with value = 0.4180 +Query 1/1: Action query time = 1.255 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5020 +t=90: Selected seed 195 with value = 0.5020 +Query 1/1: Action query time = 1.296 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5938 +t=106: Selected seed 195 with value = 0.5938 +Query 1/1: Action query time = 1.253 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7604 +t=122: Selected seed 195 with value = 0.7604 +Query 1/1: Action query time = 1.149 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7978 +t=138: Selected seed 195 with value = 0.7978 +Query 1/1: Action query time = 1.186 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9486 +t=154: Selected seed 195 with value = 0.9486 +Query 1/1: Action query time = 1.220 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=170: Selected seed 195 with value = 0.9886 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s1/2026_08_03-17_49_40--with_future_img--episode=13--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 13 +# successes: 11 (84.6%) +Current task success rate: 0.8461538461538461 +Current total success rate: 0.8461538461538461 +Final results: +Total episodes: 13 +Total successes: 11 +Overall success rate: 0.8462 (84.6%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_49_40--v2i350_p2_t3_s3.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_49_40--v2i350_p2_t3_s3.txt new file mode 100644 index 0000000000000000000000000000000000000000..e01a4b0787067190100a8e0398031c26361c664b --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-17_49_40--v2i350_p2_t3_s3.txt @@ -0,0 +1,624 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='3', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p2_t3_s3', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='3,7,11,15,19,23,27,31,35,39,43,47', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [3] +Using default initial states + +Task: open the top drawer and put the bowl inside +Starting episode 1... +Query 1/1: Action query time = 4.190 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2130 +t=10: Selected seed 195 with value = 0.2130 +Query 1/1: Action query time = 5.386 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2584 +t=26: Selected seed 195 with value = 0.2584 +Query 1/1: Action query time = 5.307 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2948 +t=42: Selected seed 195 with value = 0.2948 +Query 1/1: Action query time = 4.883 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3361 +t=58: Selected seed 195 with value = 0.3361 +Query 1/1: Action query time = 4.620 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3916 +t=74: Selected seed 195 with value = 0.3916 +Query 1/1: Action query time = 4.847 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4608 +t=90: Selected seed 195 with value = 0.4608 +Query 1/1: Action query time = 3.997 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5689 +t=106: Selected seed 195 with value = 0.5689 +Query 1/1: Action query time = 4.677 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6553 +t=122: Selected seed 195 with value = 0.6553 +Query 1/1: Action query time = 4.529 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8170 +t=138: Selected seed 195 with value = 0.8170 +Query 1/1: Action query time = 5.007 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8838 +t=154: Selected seed 195 with value = 0.8838 +Query 1/1: Action query time = 4.441 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9855 +t=170: Selected seed 195 with value = 0.9855 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=1--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 2... +Query 1/1: Action query time = 3.888 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2322 +t=10: Selected seed 195 with value = 0.2322 +Query 1/1: Action query time = 3.959 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2524 +t=26: Selected seed 195 with value = 0.2524 +Query 1/1: Action query time = 4.707 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3039 +t=42: Selected seed 195 with value = 0.3039 +Query 1/1: Action query time = 4.461 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3511 +t=58: Selected seed 195 with value = 0.3511 +Query 1/1: Action query time = 4.847 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4074 +t=74: Selected seed 195 with value = 0.4074 +Query 1/1: Action query time = 5.745 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4586 +t=90: Selected seed 195 with value = 0.4586 +Query 1/1: Action query time = 4.817 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5856 +t=106: Selected seed 195 with value = 0.5856 +Query 1/1: Action query time = 4.612 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6853 +t=122: Selected seed 195 with value = 0.6853 +Query 1/1: Action query time = 3.393 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7934 +t=138: Selected seed 195 with value = 0.7934 +Query 1/1: Action query time = 4.779 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9104 +t=154: Selected seed 195 with value = 0.9104 +Query 1/1: Action query time = 4.815 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=2--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 3... +Query 1/1: Action query time = 3.066 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2036 +t=10: Selected seed 195 with value = 0.2036 +Query 1/1: Action query time = 5.030 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2491 +t=26: Selected seed 195 with value = 0.2491 +Query 1/1: Action query time = 4.326 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3038 +t=42: Selected seed 195 with value = 0.3038 +Query 1/1: Action query time = 4.639 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3508 +t=58: Selected seed 195 with value = 0.3508 +Query 1/1: Action query time = 4.653 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4019 +t=74: Selected seed 195 with value = 0.4019 +Query 1/1: Action query time = 4.985 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4633 +t=90: Selected seed 195 with value = 0.4633 +Query 1/1: Action query time = 5.042 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5771 +t=106: Selected seed 195 with value = 0.5771 +Query 1/1: Action query time = 5.545 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6850 +t=122: Selected seed 195 with value = 0.6850 +Query 1/1: Action query time = 4.586 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8088 +t=138: Selected seed 195 with value = 0.8088 +Query 1/1: Action query time = 3.676 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9535 +t=154: Selected seed 195 with value = 0.9535 +Query 1/1: Action query time = 4.616 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=170: Selected seed 195 with value = 0.9976 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=3--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 4... +Query 1/1: Action query time = 3.591 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2306 +t=10: Selected seed 195 with value = 0.2306 +Query 1/1: Action query time = 3.549 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2541 +t=26: Selected seed 195 with value = 0.2541 +Query 1/1: Action query time = 5.188 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3022 +t=42: Selected seed 195 with value = 0.3022 +Query 1/1: Action query time = 4.072 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3386 +t=58: Selected seed 195 with value = 0.3386 +Query 1/1: Action query time = 4.976 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3891 +t=74: Selected seed 195 with value = 0.3891 +Query 1/1: Action query time = 4.535 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4538 +t=90: Selected seed 195 with value = 0.4538 +Query 1/1: Action query time = 4.354 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5620 +t=106: Selected seed 195 with value = 0.5620 +Query 1/1: Action query time = 5.124 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6183 +t=122: Selected seed 195 with value = 0.6183 +Query 1/1: Action query time = 5.222 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8151 +t=138: Selected seed 195 with value = 0.8151 +Query 1/1: Action query time = 5.099 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8896 +t=154: Selected seed 195 with value = 0.8896 +Query 1/1: Action query time = 5.835 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=4--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 5... +Query 1/1: Action query time = 3.254 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2214 +t=10: Selected seed 195 with value = 0.2214 +Query 1/1: Action query time = 4.063 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2555 +t=26: Selected seed 195 with value = 0.2555 +Query 1/1: Action query time = 5.500 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2962 +t=42: Selected seed 195 with value = 0.2962 +Query 1/1: Action query time = 5.756 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3292 +t=58: Selected seed 195 with value = 0.3292 +Query 1/1: Action query time = 5.311 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3815 +t=74: Selected seed 195 with value = 0.3815 +Query 1/1: Action query time = 5.010 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4517 +t=90: Selected seed 195 with value = 0.4517 +Query 1/1: Action query time = 4.756 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5413 +t=106: Selected seed 195 with value = 0.5413 +Query 1/1: Action query time = 4.436 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6193 +t=122: Selected seed 195 with value = 0.6193 +Query 1/1: Action query time = 4.421 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7000 +t=138: Selected seed 195 with value = 0.7000 +Query 1/1: Action query time = 4.425 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9009 +t=154: Selected seed 195 with value = 0.9009 +Query 1/1: Action query time = 3.947 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=5--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 6... +Query 1/1: Action query time = 3.851 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2082 +t=10: Selected seed 195 with value = 0.2082 +Query 1/1: Action query time = 4.232 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2416 +t=26: Selected seed 195 with value = 0.2416 +Query 1/1: Action query time = 4.377 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3064 +t=42: Selected seed 195 with value = 0.3064 +Query 1/1: Action query time = 3.915 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3551 +t=58: Selected seed 195 with value = 0.3551 +Query 1/1: Action query time = 5.198 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4177 +t=74: Selected seed 195 with value = 0.4177 +Query 1/1: Action query time = 4.991 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4661 +t=90: Selected seed 195 with value = 0.4661 +Query 1/1: Action query time = 4.382 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5915 +t=106: Selected seed 195 with value = 0.5915 +Query 1/1: Action query time = 4.703 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6888 +t=122: Selected seed 195 with value = 0.6888 +Query 1/1: Action query time = 4.688 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8213 +t=138: Selected seed 195 with value = 0.8213 +Query 1/1: Action query time = 3.906 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9736 +t=154: Selected seed 195 with value = 0.9736 +Query 1/1: Action query time = 4.068 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9932 +t=170: Selected seed 195 with value = 0.9932 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=6--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 7... +Query 1/1: Action query time = 4.214 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2230 +t=10: Selected seed 195 with value = 0.2230 +Query 1/1: Action query time = 3.266 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2456 +t=26: Selected seed 195 with value = 0.2456 +Query 1/1: Action query time = 3.475 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3025 +t=42: Selected seed 195 with value = 0.3025 +Query 1/1: Action query time = 5.291 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3417 +t=58: Selected seed 195 with value = 0.3417 +Query 1/1: Action query time = 5.173 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3938 +t=74: Selected seed 195 with value = 0.3938 +Query 1/1: Action query time = 4.842 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4640 +t=90: Selected seed 195 with value = 0.4640 +Query 1/1: Action query time = 4.659 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5676 +t=106: Selected seed 195 with value = 0.5676 +Query 1/1: Action query time = 5.348 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6568 +t=122: Selected seed 195 with value = 0.6568 +Query 1/1: Action query time = 4.041 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7900 +t=138: Selected seed 195 with value = 0.7900 +Query 1/1: Action query time = 5.011 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8830 +t=154: Selected seed 195 with value = 0.8830 +Query 1/1: Action query time = 5.018 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=170: Selected seed 195 with value = 0.9886 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=7--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 8... +Query 1/1: Action query time = 3.942 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2258 +t=10: Selected seed 195 with value = 0.2258 +Query 1/1: Action query time = 4.809 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2508 +t=26: Selected seed 195 with value = 0.2508 +Query 1/1: Action query time = 3.113 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2996 +t=42: Selected seed 195 with value = 0.2996 +Query 1/1: Action query time = 4.500 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3487 +t=58: Selected seed 195 with value = 0.3487 +Query 1/1: Action query time = 5.004 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4005 +t=74: Selected seed 195 with value = 0.4005 +Query 1/1: Action query time = 5.851 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4581 +t=90: Selected seed 195 with value = 0.4581 +Query 1/1: Action query time = 5.491 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5754 +t=106: Selected seed 195 with value = 0.5754 +Query 1/1: Action query time = 3.997 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6722 +t=122: Selected seed 195 with value = 0.6722 +Query 1/1: Action query time = 4.339 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7960 +t=138: Selected seed 195 with value = 0.7960 +Query 1/1: Action query time = 4.548 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9202 +t=154: Selected seed 195 with value = 0.9202 +Query 1/1: Action query time = 3.442 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=8--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 9... +Query 1/1: Action query time = 4.365 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2220 +t=10: Selected seed 195 with value = 0.2220 +Query 1/1: Action query time = 5.064 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2561 +t=26: Selected seed 195 with value = 0.2561 +Query 1/1: Action query time = 2.676 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2900 +t=42: Selected seed 195 with value = 0.2900 +Query 1/1: Action query time = 3.985 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3336 +t=58: Selected seed 195 with value = 0.3336 +Query 1/1: Action query time = 4.691 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3924 +t=74: Selected seed 195 with value = 0.3924 +Query 1/1: Action query time = 3.201 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4618 +t=90: Selected seed 195 with value = 0.4618 +Query 1/1: Action query time = 5.158 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5513 +t=106: Selected seed 195 with value = 0.5513 +Query 1/1: Action query time = 5.127 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6631 +t=122: Selected seed 195 with value = 0.6631 +Query 1/1: Action query time = 5.422 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7670 +t=138: Selected seed 195 with value = 0.7670 +Query 1/1: Action query time = 5.104 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8752 +t=154: Selected seed 195 with value = 0.8752 +Query 1/1: Action query time = 4.767 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=9--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 10... +Query 1/1: Action query time = 3.041 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2269 +t=10: Selected seed 195 with value = 0.2269 +Query 1/1: Action query time = 3.245 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2578 +t=26: Selected seed 195 with value = 0.2578 +Query 1/1: Action query time = 3.928 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2999 +t=42: Selected seed 195 with value = 0.2999 +Query 1/1: Action query time = 3.994 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3473 +t=58: Selected seed 195 with value = 0.3473 +Query 1/1: Action query time = 3.782 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4130 +t=74: Selected seed 195 with value = 0.4130 +Query 1/1: Action query time = 3.663 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4791 +t=90: Selected seed 195 with value = 0.4791 +Query 1/1: Action query time = 3.707 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5984 +t=106: Selected seed 195 with value = 0.5984 +Query 1/1: Action query time = 4.062 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6826 +t=122: Selected seed 195 with value = 0.6826 +Query 1/1: Action query time = 3.919 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8456 +t=138: Selected seed 195 with value = 0.8456 +Query 1/1: Action query time = 3.670 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9857 +t=154: Selected seed 195 with value = 0.9857 +Query 1/1: Action query time = 2.708 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=170: Selected seed 195 with value = 0.9981 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=10--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 11... +Query 1/1: Action query time = 1.866 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2069 +t=10: Selected seed 195 with value = 0.2069 +Query 1/1: Action query time = 2.397 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2558 +t=26: Selected seed 195 with value = 0.2558 +Query 1/1: Action query time = 3.605 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2959 +t=42: Selected seed 195 with value = 0.2959 +Query 1/1: Action query time = 3.498 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3416 +t=58: Selected seed 195 with value = 0.3416 +Query 1/1: Action query time = 3.544 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4024 +t=74: Selected seed 195 with value = 0.4024 +Query 1/1: Action query time = 3.498 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4688 +t=90: Selected seed 195 with value = 0.4688 +Query 1/1: Action query time = 2.391 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5550 +t=106: Selected seed 195 with value = 0.5550 +Query 1/1: Action query time = 2.452 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6954 +t=122: Selected seed 195 with value = 0.6954 +Query 1/1: Action query time = 2.968 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8071 +t=138: Selected seed 195 with value = 0.8071 +Query 1/1: Action query time = 3.361 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9314 +t=154: Selected seed 195 with value = 0.9314 +Query 1/1: Action query time = 3.473 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9907 +t=170: Selected seed 195 with value = 0.9907 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=11--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: open the top drawer and put the bowl inside +Starting episode 12... +Query 1/1: Action query time = 1.489 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2156 +t=10: Selected seed 195 with value = 0.2156 +Query 1/1: Action query time = 1.699 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2420 +t=26: Selected seed 195 with value = 0.2420 +Query 1/1: Action query time = 2.547 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.2978 +t=42: Selected seed 195 with value = 0.2978 +Query 1/1: Action query time = 2.570 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3319 +t=58: Selected seed 195 with value = 0.3319 +Query 1/1: Action query time = 2.503 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3929 +t=74: Selected seed 195 with value = 0.3929 +Query 1/1: Action query time = 2.486 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4497 +t=90: Selected seed 195 with value = 0.4497 +Query 1/1: Action query time = 2.446 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5534 +t=106: Selected seed 195 with value = 0.5534 +Query 1/1: Action query time = 2.423 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6857 +t=122: Selected seed 195 with value = 0.6857 +Query 1/1: Action query time = 2.400 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7927 +t=138: Selected seed 195 with value = 0.7927 +Query 1/1: Action query time = 2.215 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9150 +t=154: Selected seed 195 with value = 0.9150 +Query 1/1: Action query time = 1.728 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9952 +t=170: Selected seed 195 with value = 0.9952 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t3_s3/2026_08_03-17_49_40--with_future_img--episode=12--success=True--task=open_the_top_drawer_and_put_the_bow.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 12 +Total successes: 12 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-18_02_50--v2i350_p2_t4_s0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-18_02_50--v2i350_p2_t4_s0.txt new file mode 100644 index 0000000000000000000000000000000000000000..24a85c5c76cd7bb01a47463d17edcfb75d13d746 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-18_02_50--v2i350_p2_t4_s0.txt @@ -0,0 +1,415 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p2_t4_s0', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='0,4,8,12,16,20,24,28,32,36,40,44,48', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [4] +Using default initial states + +Task: put the bowl on top of the cabinet +Starting episode 1... +Query 1/1: Action query time = 2.946 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5493 +t=10: Selected seed 195 with value = 0.5493 +Query 1/1: Action query time = 4.218 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6880 +t=26: Selected seed 195 with value = 0.6880 +Query 1/1: Action query time = 4.623 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7872 +t=42: Selected seed 195 with value = 0.7872 +Query 1/1: Action query time = 4.907 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9638 +t=58: Selected seed 195 with value = 0.9638 +Query 1/1: Action query time = 4.586 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.041 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=1--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 2... +Query 1/1: Action query time = 4.048 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5682 +t=10: Selected seed 195 with value = 0.5682 +Query 1/1: Action query time = 4.998 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6823 +t=26: Selected seed 195 with value = 0.6823 +Query 1/1: Action query time = 5.173 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7749 +t=42: Selected seed 195 with value = 0.7749 +Query 1/1: Action query time = 3.462 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9446 +t=58: Selected seed 195 with value = 0.9446 +Query 1/1: Action query time = 3.993 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.424 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=2--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 3... +Query 1/1: Action query time = 4.156 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5846 +t=10: Selected seed 195 with value = 0.5846 +Query 1/1: Action query time = 4.077 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7021 +t=26: Selected seed 195 with value = 0.7021 +Query 1/1: Action query time = 4.678 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7785 +t=42: Selected seed 195 with value = 0.7785 +Query 1/1: Action query time = 4.866 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9411 +t=58: Selected seed 195 with value = 0.9411 +Query 1/1: Action query time = 4.720 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.370 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=3--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 4... +Query 1/1: Action query time = 2.271 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5735 +t=10: Selected seed 195 with value = 0.5735 +Query 1/1: Action query time = 1.375 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6628 +t=26: Selected seed 195 with value = 0.6628 +Query 1/1: Action query time = 4.115 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8030 +t=42: Selected seed 195 with value = 0.8030 +Query 1/1: Action query time = 6.177 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9222 +t=58: Selected seed 195 with value = 0.9222 +Query 1/1: Action query time = 5.043 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.504 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=4--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 5... +Query 1/1: Action query time = 5.264 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5964 +t=10: Selected seed 195 with value = 0.5964 +Query 1/1: Action query time = 4.095 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6990 +t=26: Selected seed 195 with value = 0.6990 +Query 1/1: Action query time = 3.027 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8286 +t=42: Selected seed 195 with value = 0.8286 +Query 1/1: Action query time = 3.862 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9688 +t=58: Selected seed 195 with value = 0.9688 +Query 1/1: Action query time = 4.839 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.738 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=5--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 6... +Query 1/1: Action query time = 5.004 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5887 +t=10: Selected seed 195 with value = 0.5887 +Query 1/1: Action query time = 4.293 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6819 +t=26: Selected seed 195 with value = 0.6819 +Query 1/1: Action query time = 5.092 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8144 +t=42: Selected seed 195 with value = 0.8144 +Query 1/1: Action query time = 5.127 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9379 +t=58: Selected seed 195 with value = 0.9379 +Query 1/1: Action query time = 5.410 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=74: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 4.432 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=6--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 7... +Query 1/1: Action query time = 2.004 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5643 +t=10: Selected seed 195 with value = 0.5643 +Query 1/1: Action query time = 4.085 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7018 +t=26: Selected seed 195 with value = 0.7018 +Query 1/1: Action query time = 4.423 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8027 +t=42: Selected seed 195 with value = 0.8027 +Query 1/1: Action query time = 4.535 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9555 +t=58: Selected seed 195 with value = 0.9555 +Query 1/1: Action query time = 3.658 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.127 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=7--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 8... +Query 1/1: Action query time = 4.511 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5578 +t=10: Selected seed 195 with value = 0.5578 +Query 1/1: Action query time = 5.220 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6445 +t=26: Selected seed 195 with value = 0.6445 +Query 1/1: Action query time = 4.658 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7538 +t=42: Selected seed 195 with value = 0.7538 +Query 1/1: Action query time = 3.007 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9161 +t=58: Selected seed 195 with value = 0.9161 +Query 1/1: Action query time = 4.458 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9972 +t=74: Selected seed 195 with value = 0.9972 +Query 1/1: Action query time = 3.019 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=8--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 9... +Query 1/1: Action query time = 5.800 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6051 +t=10: Selected seed 195 with value = 0.6051 +Query 1/1: Action query time = 5.816 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7242 +t=26: Selected seed 195 with value = 0.7242 +Query 1/1: Action query time = 4.809 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7186 +t=42: Selected seed 195 with value = 0.7186 +Query 1/1: Action query time = 4.530 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8740 +t=58: Selected seed 195 with value = 0.8740 +Query 1/1: Action query time = 4.980 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.408 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=9--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 10... +Query 1/1: Action query time = 5.461 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5700 +t=10: Selected seed 195 with value = 0.5700 +Query 1/1: Action query time = 5.394 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6858 +t=26: Selected seed 195 with value = 0.6858 +Query 1/1: Action query time = 3.248 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7736 +t=42: Selected seed 195 with value = 0.7736 +Query 1/1: Action query time = 3.987 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9334 +t=58: Selected seed 195 with value = 0.9334 +Query 1/1: Action query time = 3.446 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.114 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=10--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 11... +Query 1/1: Action query time = 4.758 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5767 +t=10: Selected seed 195 with value = 0.5767 +Query 1/1: Action query time = 4.477 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7081 +t=26: Selected seed 195 with value = 0.7081 +Query 1/1: Action query time = 4.275 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7673 +t=42: Selected seed 195 with value = 0.7673 +Query 1/1: Action query time = 4.053 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9308 +t=58: Selected seed 195 with value = 0.9308 +Query 1/1: Action query time = 4.521 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9927 +t=74: Selected seed 195 with value = 0.9927 +Query 1/1: Action query time = 4.934 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=11--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 12... +Query 1/1: Action query time = 3.919 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6006 +t=10: Selected seed 195 with value = 0.6006 +Query 1/1: Action query time = 5.654 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7161 +t=26: Selected seed 195 with value = 0.7161 +Query 1/1: Action query time = 4.631 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8093 +t=42: Selected seed 195 with value = 0.8093 +Query 1/1: Action query time = 4.562 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9470 +t=58: Selected seed 195 with value = 0.9470 +Query 1/1: Action query time = 4.558 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=74: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.417 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=12--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: put the bowl on top of the cabinet +Starting episode 13... +Query 1/1: Action query time = 3.864 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5740 +t=10: Selected seed 195 with value = 0.5740 +Query 1/1: Action query time = 5.883 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6881 +t=26: Selected seed 195 with value = 0.6881 +Query 1/1: Action query time = 3.449 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7632 +t=42: Selected seed 195 with value = 0.7632 +Query 1/1: Action query time = 4.993 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9233 +t=58: Selected seed 195 with value = 0.9233 +Query 1/1: Action query time = 4.312 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=74: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 4.687 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t4_s0/2026_08_03-18_02_50--with_future_img--episode=13--success=True--task=put_the_bowl_on_top_of_the_cabinet.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 13 +Total successes: 13 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-18_02_51--v2i350_p2_t5_s1.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-18_02_51--v2i350_p2_t5_s1.txt new file mode 100644 index 0000000000000000000000000000000000000000..2e37f3b925aad255709aef2c0fbd8eadd15384c1 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-18_02_51--v2i350_p2_t5_s1.txt @@ -0,0 +1,579 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='5', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p2_t5_s1', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='1,5,9,13,17,21,25,29,33,37,41,45,49', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [5] +Using default initial states + +Task: push the plate to the front of the stove +Starting episode 1... +Query 1/1: Action query time = 6.561 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3735 +t=10: Selected seed 195 with value = 0.3735 +Query 1/1: Action query time = 5.321 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4242 +t=26: Selected seed 195 with value = 0.4242 +Query 1/1: Action query time = 5.047 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5573 +t=42: Selected seed 195 with value = 0.5573 +Query 1/1: Action query time = 4.872 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6421 +t=58: Selected seed 195 with value = 0.6421 +Query 1/1: Action query time = 4.508 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6992 +t=74: Selected seed 195 with value = 0.6992 +Query 1/1: Action query time = 4.075 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8022 +t=90: Selected seed 195 with value = 0.8022 +Query 1/1: Action query time = 5.052 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9185 +t=106: Selected seed 195 with value = 0.9185 +Query 1/1: Action query time = 5.207 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.567 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=1--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 2... +Query 1/1: Action query time = 5.361 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3881 +t=10: Selected seed 195 with value = 0.3881 +Query 1/1: Action query time = 4.337 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4367 +t=26: Selected seed 195 with value = 0.4367 +Query 1/1: Action query time = 3.413 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5438 +t=42: Selected seed 195 with value = 0.5438 +Query 1/1: Action query time = 4.252 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6521 +t=58: Selected seed 195 with value = 0.6521 +Query 1/1: Action query time = 5.297 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7173 +t=74: Selected seed 195 with value = 0.7173 +Query 1/1: Action query time = 5.609 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8637 +t=90: Selected seed 195 with value = 0.8637 +Query 1/1: Action query time = 5.341 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9664 +t=106: Selected seed 195 with value = 0.9664 +Query 1/1: Action query time = 5.091 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.386 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=2--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 3... +Query 1/1: Action query time = 4.126 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4073 +t=10: Selected seed 195 with value = 0.4073 +Query 1/1: Action query time = 5.509 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4592 +t=26: Selected seed 195 with value = 0.4592 +Query 1/1: Action query time = 5.037 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5526 +t=42: Selected seed 195 with value = 0.5526 +Query 1/1: Action query time = 5.259 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6334 +t=58: Selected seed 195 with value = 0.6334 +Query 1/1: Action query time = 3.859 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6905 +t=74: Selected seed 195 with value = 0.6905 +Query 1/1: Action query time = 4.985 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8510 +t=90: Selected seed 195 with value = 0.8510 +Query 1/1: Action query time = 4.743 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9999 +t=106: Selected seed 195 with value = 0.9999 +Query 1/1: Action query time = 4.020 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=3--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 4... +Query 1/1: Action query time = 4.137 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3827 +t=10: Selected seed 195 with value = 0.3827 +Query 1/1: Action query time = 4.408 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4369 +t=26: Selected seed 195 with value = 0.4369 +Query 1/1: Action query time = 5.344 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5329 +t=42: Selected seed 195 with value = 0.5329 +Query 1/1: Action query time = 4.553 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5976 +t=58: Selected seed 195 with value = 0.5976 +Query 1/1: Action query time = 4.794 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6290 +t=74: Selected seed 195 with value = 0.6290 +Query 1/1: Action query time = 3.968 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7895 +t=90: Selected seed 195 with value = 0.7895 +Query 1/1: Action query time = 4.096 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9279 +t=106: Selected seed 195 with value = 0.9279 +Query 1/1: Action query time = 4.882 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 4.810 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9974 +t=138: Selected seed 195 with value = 0.9974 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=4--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 5... +Query 1/1: Action query time = 4.821 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3773 +t=10: Selected seed 195 with value = 0.3773 +Query 1/1: Action query time = 4.361 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4774 +t=26: Selected seed 195 with value = 0.4774 +Query 1/1: Action query time = 4.859 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5571 +t=42: Selected seed 195 with value = 0.5571 +Query 1/1: Action query time = 6.066 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5922 +t=58: Selected seed 195 with value = 0.5922 +Query 1/1: Action query time = 5.012 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6473 +t=74: Selected seed 195 with value = 0.6473 +Query 1/1: Action query time = 4.984 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7443 +t=90: Selected seed 195 with value = 0.7443 +Query 1/1: Action query time = 4.439 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8347 +t=106: Selected seed 195 with value = 0.8347 +Query 1/1: Action query time = 4.504 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9789 +t=122: Selected seed 195 with value = 0.9789 +Query 1/1: Action query time = 2.821 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9919 +t=138: Selected seed 195 with value = 0.9919 +Query 1/1: Action query time = 3.834 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=5--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 6... +Query 1/1: Action query time = 4.737 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3662 +t=10: Selected seed 195 with value = 0.3662 +Query 1/1: Action query time = 4.550 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4350 +t=26: Selected seed 195 with value = 0.4350 +Query 1/1: Action query time = 4.848 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5080 +t=42: Selected seed 195 with value = 0.5080 +Query 1/1: Action query time = 5.557 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6079 +t=58: Selected seed 195 with value = 0.6079 +Query 1/1: Action query time = 4.608 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6529 +t=74: Selected seed 195 with value = 0.6529 +Query 1/1: Action query time = 4.232 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8205 +t=90: Selected seed 195 with value = 0.8205 +Query 1/1: Action query time = 4.187 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9934 +t=106: Selected seed 195 with value = 0.9934 +Query 1/1: Action query time = 5.705 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 5.804 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=138: Selected seed 195 with value = 0.9979 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=6--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 7... +Query 1/1: Action query time = 4.427 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3654 +t=10: Selected seed 195 with value = 0.3654 +Query 1/1: Action query time = 5.691 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4393 +t=26: Selected seed 195 with value = 0.4393 +Query 1/1: Action query time = 4.499 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5260 +t=42: Selected seed 195 with value = 0.5260 +Query 1/1: Action query time = 5.224 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5798 +t=58: Selected seed 195 with value = 0.5798 +Query 1/1: Action query time = 4.161 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6157 +t=74: Selected seed 195 with value = 0.6157 +Query 1/1: Action query time = 4.421 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7104 +t=90: Selected seed 195 with value = 0.7104 +Query 1/1: Action query time = 4.718 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7995 +t=106: Selected seed 195 with value = 0.7995 +Query 1/1: Action query time = 5.361 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9747 +t=122: Selected seed 195 with value = 0.9747 +Query 1/1: Action query time = 4.908 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9987 +t=138: Selected seed 195 with value = 0.9987 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=7--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 8... +Query 1/1: Action query time = 5.658 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3555 +t=10: Selected seed 195 with value = 0.3555 +Query 1/1: Action query time = 5.419 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4539 +t=26: Selected seed 195 with value = 0.4539 +Query 1/1: Action query time = 4.614 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4663 +t=42: Selected seed 195 with value = 0.4663 +Query 1/1: Action query time = 4.893 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5506 +t=58: Selected seed 195 with value = 0.5506 +Query 1/1: Action query time = 5.526 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6158 +t=74: Selected seed 195 with value = 0.6158 +Query 1/1: Action query time = 2.816 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7078 +t=90: Selected seed 195 with value = 0.7078 +Query 1/1: Action query time = 4.496 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8036 +t=106: Selected seed 195 with value = 0.8036 +Query 1/1: Action query time = 4.814 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9346 +t=122: Selected seed 195 with value = 0.9346 +Query 1/1: Action query time = 3.625 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9996 +t=138: Selected seed 195 with value = 0.9996 +Query 1/1: Action query time = 4.782 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=8--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 9... +Query 1/1: Action query time = 5.400 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3869 +t=10: Selected seed 195 with value = 0.3869 +Query 1/1: Action query time = 3.651 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4417 +t=26: Selected seed 195 with value = 0.4417 +Query 1/1: Action query time = 3.609 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5181 +t=42: Selected seed 195 with value = 0.5181 +Query 1/1: Action query time = 3.128 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5904 +t=58: Selected seed 195 with value = 0.5904 +Query 1/1: Action query time = 3.441 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6698 +t=74: Selected seed 195 with value = 0.6698 +Query 1/1: Action query time = 1.845 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7567 +t=90: Selected seed 195 with value = 0.7567 +Query 1/1: Action query time = 2.972 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9009 +t=106: Selected seed 195 with value = 0.9009 +Query 1/1: Action query time = 2.596 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 3.010 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=9--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 10... +Query 1/1: Action query time = 3.469 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3544 +t=10: Selected seed 195 with value = 0.3544 +Query 1/1: Action query time = 2.582 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4156 +t=26: Selected seed 195 with value = 0.4156 +Query 1/1: Action query time = 2.557 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5088 +t=42: Selected seed 195 with value = 0.5088 +Query 1/1: Action query time = 2.803 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5414 +t=58: Selected seed 195 with value = 0.5414 +Query 1/1: Action query time = 2.368 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5829 +t=74: Selected seed 195 with value = 0.5829 +Query 1/1: Action query time = 2.339 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6689 +t=90: Selected seed 195 with value = 0.6689 +Query 1/1: Action query time = 2.160 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7613 +t=106: Selected seed 195 with value = 0.7613 +Query 1/1: Action query time = 1.545 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9269 +t=122: Selected seed 195 with value = 0.9269 +Query 1/1: Action query time = 1.889 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=138: Selected seed 195 with value = 0.9981 +Query 1/1: Action query time = 2.830 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=10--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 11... +Query 1/1: Action query time = 1.901 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3840 +t=10: Selected seed 195 with value = 0.3840 +Query 1/1: Action query time = 1.974 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4342 +t=26: Selected seed 195 with value = 0.4342 +Query 1/1: Action query time = 2.455 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5594 +t=42: Selected seed 195 with value = 0.5594 +Query 1/1: Action query time = 2.427 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6603 +t=58: Selected seed 195 with value = 0.6603 +Query 1/1: Action query time = 2.468 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7001 +t=74: Selected seed 195 with value = 0.7001 +Query 1/1: Action query time = 1.613 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8242 +t=90: Selected seed 195 with value = 0.8242 +Query 1/1: Action query time = 1.481 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9615 +t=106: Selected seed 195 with value = 0.9615 +Query 1/1: Action query time = 2.374 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 2.269 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=11--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 12... +Query 1/1: Action query time = 2.699 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3618 +t=10: Selected seed 195 with value = 0.3618 +Query 1/1: Action query time = 1.870 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4205 +t=26: Selected seed 195 with value = 0.4205 +Query 1/1: Action query time = 1.211 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5660 +t=42: Selected seed 195 with value = 0.5660 +Query 1/1: Action query time = 1.713 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5587 +t=58: Selected seed 195 with value = 0.5587 +Query 1/1: Action query time = 1.819 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6556 +t=74: Selected seed 195 with value = 0.6556 +Query 1/1: Action query time = 1.821 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7085 +t=90: Selected seed 195 with value = 0.7085 +Query 1/1: Action query time = 1.641 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8135 +t=106: Selected seed 195 with value = 0.8135 +Query 1/1: Action query time = 0.958 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9985 +t=122: Selected seed 195 with value = 0.9985 +Query 1/1: Action query time = 0.971 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9954 +t=138: Selected seed 195 with value = 0.9954 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=12--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) + +Task: push the plate to the front of the stove +Starting episode 13... +Query 1/1: Action query time = 1.804 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.3614 +t=10: Selected seed 195 with value = 0.3614 +Query 1/1: Action query time = 1.786 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4418 +t=26: Selected seed 195 with value = 0.4418 +Query 1/1: Action query time = 1.803 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4719 +t=42: Selected seed 195 with value = 0.4719 +Query 1/1: Action query time = 1.648 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5906 +t=58: Selected seed 195 with value = 0.5906 +Query 1/1: Action query time = 1.781 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6360 +t=74: Selected seed 195 with value = 0.6360 +Query 1/1: Action query time = 1.794 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7220 +t=90: Selected seed 195 with value = 0.7220 +Query 1/1: Action query time = 1.618 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8327 +t=106: Selected seed 195 with value = 0.8327 +Query 1/1: Action query time = 0.977 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9830 +t=122: Selected seed 195 with value = 0.9830 +Query 1/1: Action query time = 0.973 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9981 +t=138: Selected seed 195 with value = 0.9981 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t5_s1/2026_08_03-18_02_51--with_future_img--episode=13--success=True--task=push_the_plate_to_the_front_of_the_.mp4 +Success: True +# episodes completed so far: 13 +# successes: 13 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 13 +Total successes: 13 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-18_02_52--v2i350_p2_t7_s2.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-18_02_52--v2i350_p2_t7_s2.txt new file mode 100644 index 0000000000000000000000000000000000000000..9f285f69fd6c19e1a3d9e709168908e4ca2d9fd2 --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_08_03-18_02_52--v2i350_p2_t7_s2.txt @@ -0,0 +1,344 @@ +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/new_task7/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task7_forget_cl_v2_from350_2gpu/checkpoints/iter_000000350/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='7', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='/home/azureuser/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='v2i350_p2_t7_s2', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=False, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', eval_episode_ids='2,6,10,14,18,22,26,30,34,38,42,46', dump_final_states_path='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200, disable_value_stop=False) +Using serial inference (parallel inference disabled) +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [7] +Using default initial states + +Task: turn on the stove +Starting episode 1... +Query 1/1: Action query time = 2.353 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4764 +t=10: Selected seed 195 with value = 0.4764 +Query 1/1: Action query time = 3.347 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5652 +t=26: Selected seed 195 with value = 0.5652 +Query 1/1: Action query time = 4.649 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6586 +t=42: Selected seed 195 with value = 0.6586 +Query 1/1: Action query time = 5.480 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7660 +t=58: Selected seed 195 with value = 0.7660 +Query 1/1: Action query time = 4.986 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9102 +t=74: Selected seed 195 with value = 0.9102 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=1--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 1 +# successes: 1 (100.0%) + +Task: turn on the stove +Starting episode 2... +Query 1/1: Action query time = 4.613 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4864 +t=10: Selected seed 195 with value = 0.4864 +Query 1/1: Action query time = 4.847 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5549 +t=26: Selected seed 195 with value = 0.5549 +Query 1/1: Action query time = 4.734 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6572 +t=42: Selected seed 195 with value = 0.6572 +Query 1/1: Action query time = 5.224 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7911 +t=58: Selected seed 195 with value = 0.7911 +Query 1/1: Action query time = 4.902 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9168 +t=74: Selected seed 195 with value = 0.9168 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=2--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 2 +# successes: 2 (100.0%) + +Task: turn on the stove +Starting episode 3... +Query 1/1: Action query time = 4.230 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4847 +t=10: Selected seed 195 with value = 0.4847 +Query 1/1: Action query time = 5.060 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5418 +t=26: Selected seed 195 with value = 0.5418 +Query 1/1: Action query time = 3.295 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6560 +t=42: Selected seed 195 with value = 0.6560 +Query 1/1: Action query time = 5.373 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7781 +t=58: Selected seed 195 with value = 0.7781 +Query 1/1: Action query time = 5.067 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9225 +t=74: Selected seed 195 with value = 0.9225 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=3--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 3 +# successes: 3 (100.0%) + +Task: turn on the stove +Starting episode 4... +Query 1/1: Action query time = 5.431 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4931 +t=10: Selected seed 195 with value = 0.4931 +Query 1/1: Action query time = 5.136 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5613 +t=26: Selected seed 195 with value = 0.5613 +Query 1/1: Action query time = 3.164 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6440 +t=42: Selected seed 195 with value = 0.6440 +Query 1/1: Action query time = 3.815 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7749 +t=58: Selected seed 195 with value = 0.7749 +Query 1/1: Action query time = 3.886 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8782 +t=74: Selected seed 195 with value = 0.8782 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=4--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 4 +# successes: 4 (100.0%) + +Task: turn on the stove +Starting episode 5... +Query 1/1: Action query time = 4.385 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4802 +t=10: Selected seed 195 with value = 0.4802 +Query 1/1: Action query time = 4.470 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5603 +t=26: Selected seed 195 with value = 0.5603 +Query 1/1: Action query time = 4.722 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7081 +t=42: Selected seed 195 with value = 0.7081 +Query 1/1: Action query time = 5.303 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8582 +t=58: Selected seed 195 with value = 0.8582 +Query 1/1: Action query time = 5.691 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=74: Selected seed 195 with value = 0.9968 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=5--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 5 +# successes: 5 (100.0%) + +Task: turn on the stove +Starting episode 6... +Query 1/1: Action query time = 4.145 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4959 +t=10: Selected seed 195 with value = 0.4959 +Query 1/1: Action query time = 4.643 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5799 +t=26: Selected seed 195 with value = 0.5799 +Query 1/1: Action query time = 5.337 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6814 +t=42: Selected seed 195 with value = 0.6814 +Query 1/1: Action query time = 4.595 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8330 +t=58: Selected seed 195 with value = 0.8330 +Query 1/1: Action query time = 5.025 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9675 +t=74: Selected seed 195 with value = 0.9675 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=6--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 6 +# successes: 6 (100.0%) + +Task: turn on the stove +Starting episode 7... +Query 1/1: Action query time = 4.390 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4792 +t=10: Selected seed 195 with value = 0.4792 +Query 1/1: Action query time = 4.173 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5511 +t=26: Selected seed 195 with value = 0.5511 +Query 1/1: Action query time = 5.106 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6849 +t=42: Selected seed 195 with value = 0.6849 +Query 1/1: Action query time = 5.911 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8070 +t=58: Selected seed 195 with value = 0.8070 +Query 1/1: Action query time = 5.848 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9315 +t=74: Selected seed 195 with value = 0.9315 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=7--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 7 +# successes: 7 (100.0%) + +Task: turn on the stove +Starting episode 8... +Query 1/1: Action query time = 4.217 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4726 +t=10: Selected seed 195 with value = 0.4726 +Query 1/1: Action query time = 4.023 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5666 +t=26: Selected seed 195 with value = 0.5666 +Query 1/1: Action query time = 3.721 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6430 +t=42: Selected seed 195 with value = 0.6430 +Query 1/1: Action query time = 5.574 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7721 +t=58: Selected seed 195 with value = 0.7721 +Query 1/1: Action query time = 5.272 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8988 +t=74: Selected seed 195 with value = 0.8988 +Query 1/1: Action query time = 4.351 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9968 +t=90: Selected seed 195 with value = 0.9968 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=8--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 8 +# successes: 8 (100.0%) + +Task: turn on the stove +Starting episode 9... +Query 1/1: Action query time = 4.311 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5154 +t=10: Selected seed 195 with value = 0.5154 +Query 1/1: Action query time = 3.177 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6431 +t=26: Selected seed 195 with value = 0.6431 +Query 1/1: Action query time = 4.348 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7239 +t=42: Selected seed 195 with value = 0.7239 +Query 1/1: Action query time = 3.957 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8498 +t=58: Selected seed 195 with value = 0.8498 +Query 1/1: Action query time = 4.644 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9716 +t=74: Selected seed 195 with value = 0.9716 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=9--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 9 +# successes: 9 (100.0%) + +Task: turn on the stove +Starting episode 10... +Query 1/1: Action query time = 4.184 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5348 +t=10: Selected seed 195 with value = 0.5348 +Query 1/1: Action query time = 3.061 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6062 +t=26: Selected seed 195 with value = 0.6062 +Query 1/1: Action query time = 4.250 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7239 +t=42: Selected seed 195 with value = 0.7239 +Query 1/1: Action query time = 4.075 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8407 +t=58: Selected seed 195 with value = 0.8407 +Query 1/1: Action query time = 4.359 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9894 +t=74: Selected seed 195 with value = 0.9894 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=10--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 10 +# successes: 10 (100.0%) + +Task: turn on the stove +Starting episode 11... +Query 1/1: Action query time = 5.013 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5181 +t=10: Selected seed 195 with value = 0.5181 +Query 1/1: Action query time = 4.807 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6032 +t=26: Selected seed 195 with value = 0.6032 +Query 1/1: Action query time = 4.295 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6968 +t=42: Selected seed 195 with value = 0.6968 +Query 1/1: Action query time = 3.616 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8379 +t=58: Selected seed 195 with value = 0.8379 +Query 1/1: Action query time = 4.785 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9658 +t=74: Selected seed 195 with value = 0.9658 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=11--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 11 +# successes: 11 (100.0%) + +Task: turn on the stove +Starting episode 12... +Query 1/1: Action query time = 5.047 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4632 +t=10: Selected seed 195 with value = 0.4632 +Query 1/1: Action query time = 4.855 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5481 +t=26: Selected seed 195 with value = 0.5481 +Query 1/1: Action query time = 5.129 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6538 +t=42: Selected seed 195 with value = 0.6538 +Query 1/1: Action query time = 3.460 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7514 +t=58: Selected seed 195 with value = 0.7514 +Query 1/1: Action query time = 4.859 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8735 +t=74: Selected seed 195 with value = 0.8735 +Query 1/1: Action query time = 4.890 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9991 +t=90: Selected seed 195 with value = 0.9991 +Saved rollout MP4 at path ./rollouts/v2i350_p2_t7_s2/2026_08_03-18_02_52--with_future_img--episode=12--success=True--task=turn_on_the_stove.mp4 +Success: True +# episodes completed so far: 12 +# successes: 12 (100.0%) +Current task success rate: 1.0 +Current total success rate: 1.0 +Final results: +Total episodes: 12 +Total successes: 12 +Overall success rate: 1.0000 (100.0%) diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal_gpu0-cosmos-2026_07_31-00_55_25--CLbest_t15--gpu0.txt b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal_gpu0-cosmos-2026_07_31-00_55_25--CLbest_t15--gpu0.txt new file mode 100644 index 0000000000000000000000000000000000000000..f1ca857dc9d4c23609adcc8c9c52dd0d639a972b --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal_gpu0-cosmos-2026_07_31-00_55_25--CLbest_t15--gpu0.txt @@ -0,0 +1,1098 @@ +[worker gpu=0] tasks=[1, 3, 5] CUDA_VISIBLE_DEVICES=0 +[worker gpu=0] starting task_id=1 +Using default initial states + +Task: put the bowl on the stove +Starting episode 1... +Query 1/1: Action query time = 1.918 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4688 +t=10: Selected seed 195 with value = 0.4688 +Query 1/1: Action query time = 1.330 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5660 +t=26: Selected seed 195 with value = 0.5660 +Query 1/1: Action query time = 1.324 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6636 +t=42: Selected seed 195 with value = 0.6636 +Query 1/1: Action query time = 1.359 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7204 +t=58: Selected seed 195 with value = 0.7204 +Query 1/1: Action query time = 1.610 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7782 +t=74: Selected seed 195 with value = 0.7782 +Query 1/1: Action query time = 1.826 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8154 +t=90: Selected seed 195 with value = 0.8154 +Query 1/1: Action query time = 0.962 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9034 +t=106: Selected seed 195 with value = 0.9034 +Query 1/1: Action query time = 0.963 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8696 +t=122: Selected seed 195 with value = 0.8696 +Query 1/1: Action query time = 0.968 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9494 +t=138: Selected seed 195 with value = 0.9494 +Query 1/1: Action query time = 1.327 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=154: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 1.312 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9942 +t=170: Selected seed 195 with value = 0.9942 +Query 1/1: Action query time = 1.318 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9928 +t=186: Selected seed 195 with value = 0.9928 +Query 1/1: Action query time = 1.401 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9813 +t=202: Selected seed 195 with value = 0.9813 +Query 1/1: Action query time = 1.382 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9718 +t=218: Selected seed 195 with value = 0.9718 +Query 1/1: Action query time = 1.505 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9643 +t=234: Selected seed 195 with value = 0.9643 +Query 1/1: Action query time = 1.915 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9455 +t=250: Selected seed 195 with value = 0.9455 +Query 1/1: Action query time = 0.979 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9046 +t=266: Selected seed 195 with value = 0.9046 +Query 1/1: Action query time = 0.968 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8400 +t=282: Selected seed 195 with value = 0.8400 +Query 1/1: Action query time = 0.966 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7957 +t=298: Selected seed 195 with value = 0.7957 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=1--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 1 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 2... +Query 1/1: Action query time = 1.012 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4693 +t=10: Selected seed 195 with value = 0.4693 +Query 1/1: Action query time = 1.039 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5675 +t=26: Selected seed 195 with value = 0.5675 +Query 1/1: Action query time = 1.053 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6234 +t=42: Selected seed 195 with value = 0.6234 +Query 1/1: Action query time = 1.459 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7376 +t=58: Selected seed 195 with value = 0.7376 +Query 1/1: Action query time = 1.287 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8125 +t=74: Selected seed 195 with value = 0.8125 +Query 1/1: Action query time = 1.288 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9185 +t=90: Selected seed 195 with value = 0.9185 +Query 1/1: Action query time = 1.240 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9952 +t=106: Selected seed 195 with value = 0.9952 +Query 1/1: Action query time = 1.182 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9869 +t=122: Selected seed 195 with value = 0.9869 +Query 1/1: Action query time = 1.238 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9878 +t=138: Selected seed 195 with value = 0.9878 +Query 1/1: Action query time = 1.121 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9751 +t=154: Selected seed 195 with value = 0.9751 +Query 1/1: Action query time = 1.066 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9651 +t=170: Selected seed 195 with value = 0.9651 +Query 1/1: Action query time = 1.081 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9469 +t=186: Selected seed 195 with value = 0.9469 +Query 1/1: Action query time = 1.756 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9321 +t=202: Selected seed 195 with value = 0.9321 +Query 1/1: Action query time = 1.582 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9172 +t=218: Selected seed 195 with value = 0.9172 +Query 1/1: Action query time = 1.701 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8908 +t=234: Selected seed 195 with value = 0.8908 +Query 1/1: Action query time = 1.645 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8706 +t=250: Selected seed 195 with value = 0.8706 +Query 1/1: Action query time = 1.806 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8869 +t=266: Selected seed 195 with value = 0.8869 +Query 1/1: Action query time = 1.654 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9157 +t=282: Selected seed 195 with value = 0.9157 +Query 1/1: Action query time = 1.421 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9142 +t=298: Selected seed 195 with value = 0.9142 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=2--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 2 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 3... +Query 1/1: Action query time = 1.438 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4812 +t=10: Selected seed 195 with value = 0.4812 +Query 1/1: Action query time = 1.578 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5827 +t=26: Selected seed 195 with value = 0.5827 +Query 1/1: Action query time = 1.773 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6445 +t=42: Selected seed 195 with value = 0.6445 +Query 1/1: Action query time = 1.830 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7414 +t=58: Selected seed 195 with value = 0.7414 +Query 1/1: Action query time = 1.202 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8086 +t=74: Selected seed 195 with value = 0.8086 +Query 1/1: Action query time = 1.172 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9357 +t=90: Selected seed 195 with value = 0.9357 +Query 1/1: Action query time = 1.218 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9912 +t=106: Selected seed 195 with value = 0.9912 +Query 1/1: Action query time = 1.399 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9957 +t=122: Selected seed 195 with value = 0.9957 +Query 1/1: Action query time = 1.214 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9969 +t=138: Selected seed 195 with value = 0.9969 +Query 1/1: Action query time = 1.209 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9890 +t=154: Selected seed 195 with value = 0.9890 +Query 1/1: Action query time = 1.188 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9837 +t=170: Selected seed 195 with value = 0.9837 +Query 1/1: Action query time = 1.231 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9787 +t=186: Selected seed 195 with value = 0.9787 +Query 1/1: Action query time = 1.250 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9690 +t=202: Selected seed 195 with value = 0.9690 +Query 1/1: Action query time = 1.345 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9446 +t=218: Selected seed 195 with value = 0.9446 +Query 1/1: Action query time = 0.969 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9014 +t=234: Selected seed 195 with value = 0.9014 +Query 1/1: Action query time = 0.962 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8385 +t=250: Selected seed 195 with value = 0.8385 +Query 1/1: Action query time = 1.015 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8081 +t=266: Selected seed 195 with value = 0.8081 +Query 1/1: Action query time = 1.189 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8588 +t=282: Selected seed 195 with value = 0.8588 +Query 1/1: Action query time = 1.250 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9149 +t=298: Selected seed 195 with value = 0.9149 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=3--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 3 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 4... +Query 1/1: Action query time = 0.982 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4863 +t=10: Selected seed 195 with value = 0.4863 +Query 1/1: Action query time = 0.953 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5931 +t=26: Selected seed 195 with value = 0.5931 +Query 1/1: Action query time = 1.246 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6666 +t=42: Selected seed 195 with value = 0.6666 +Query 1/1: Action query time = 1.243 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6734 +t=58: Selected seed 195 with value = 0.6734 +Query 1/1: Action query time = 1.490 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7043 +t=74: Selected seed 195 with value = 0.7043 +Query 1/1: Action query time = 1.265 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7522 +t=90: Selected seed 195 with value = 0.7522 +Query 1/1: Action query time = 1.429 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8003 +t=106: Selected seed 195 with value = 0.8003 +Query 1/1: Action query time = 1.154 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8638 +t=122: Selected seed 195 with value = 0.8638 +Query 1/1: Action query time = 1.062 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9070 +t=138: Selected seed 195 with value = 0.9070 +Query 1/1: Action query time = 0.987 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9256 +t=154: Selected seed 195 with value = 0.9256 +Query 1/1: Action query time = 1.044 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9915 +t=170: Selected seed 195 with value = 0.9915 +Query 1/1: Action query time = 1.317 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9898 +t=186: Selected seed 195 with value = 0.9898 +Query 1/1: Action query time = 1.336 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=202: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 1.249 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9917 +t=218: Selected seed 195 with value = 0.9917 +Query 1/1: Action query time = 1.008 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9854 +t=234: Selected seed 195 with value = 0.9854 +Query 1/1: Action query time = 1.248 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9824 +t=250: Selected seed 195 with value = 0.9824 +Query 1/1: Action query time = 1.278 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9680 +t=266: Selected seed 195 with value = 0.9680 +Query 1/1: Action query time = 1.314 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9452 +t=282: Selected seed 195 with value = 0.9452 +Query 1/1: Action query time = 0.998 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9192 +t=298: Selected seed 195 with value = 0.9192 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=4--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 4 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 5... +Query 1/1: Action query time = 1.283 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4905 +t=10: Selected seed 195 with value = 0.4905 +Query 1/1: Action query time = 1.039 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5462 +t=26: Selected seed 195 with value = 0.5462 +Query 1/1: Action query time = 1.214 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6197 +t=42: Selected seed 195 with value = 0.6197 +Query 1/1: Action query time = 1.228 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7273 +t=58: Selected seed 195 with value = 0.7273 +Query 1/1: Action query time = 1.226 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8420 +t=74: Selected seed 195 with value = 0.8420 +Query 1/1: Action query time = 1.253 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=90: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 1.296 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.298 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.318 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.269 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=154: Selected seed 195 with value = 0.9997 +Query 1/1: Action query time = 1.208 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.247 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=186: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.246 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.227 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.211 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.243 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.969 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=266: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 0.976 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=282: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 0.962 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9940 +t=298: Selected seed 195 with value = 0.9940 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=5--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=5--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 5 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 6... +Query 1/1: Action query time = 1.272 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4946 +t=10: Selected seed 195 with value = 0.4946 +Query 1/1: Action query time = 1.332 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5150 +t=26: Selected seed 195 with value = 0.5150 +Query 1/1: Action query time = 1.411 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6237 +t=42: Selected seed 195 with value = 0.6237 +Query 1/1: Action query time = 1.743 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7430 +t=58: Selected seed 195 with value = 0.7430 +Query 1/1: Action query time = 1.652 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8317 +t=74: Selected seed 195 with value = 0.8317 +Query 1/1: Action query time = 1.627 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9799 +t=90: Selected seed 195 with value = 0.9799 +Query 1/1: Action query time = 1.625 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.628 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.610 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.648 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.709 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=170: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 1.720 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9945 +t=186: Selected seed 195 with value = 0.9945 +Query 1/1: Action query time = 1.671 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9895 +t=202: Selected seed 195 with value = 0.9895 +Query 1/1: Action query time = 1.397 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9950 +t=218: Selected seed 195 with value = 0.9950 +Query 1/1: Action query time = 1.384 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9872 +t=234: Selected seed 195 with value = 0.9872 +Query 1/1: Action query time = 1.458 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9524 +t=250: Selected seed 195 with value = 0.9524 +Query 1/1: Action query time = 0.971 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9986 +t=266: Selected seed 195 with value = 0.9986 +Query 1/1: Action query time = 0.971 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9938 +t=282: Selected seed 195 with value = 0.9938 +Query 1/1: Action query time = 0.979 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9871 +t=298: Selected seed 195 with value = 0.9871 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=6--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=6--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 6 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 7... +Query 1/1: Action query time = 1.341 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4754 +t=10: Selected seed 195 with value = 0.4754 +Query 1/1: Action query time = 1.400 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5437 +t=26: Selected seed 195 with value = 0.5437 +Query 1/1: Action query time = 1.832 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6059 +t=42: Selected seed 195 with value = 0.6059 +Query 1/1: Action query time = 1.817 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6838 +t=58: Selected seed 195 with value = 0.6838 +Query 1/1: Action query time = 1.834 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7938 +t=74: Selected seed 195 with value = 0.7938 +Query 1/1: Action query time = 1.763 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8971 +t=90: Selected seed 195 with value = 0.8971 +Query 1/1: Action query time = 1.712 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9632 +t=106: Selected seed 195 with value = 0.9632 +Query 1/1: Action query time = 1.711 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.440 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.428 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9983 +t=154: Selected seed 195 with value = 0.9983 +Query 1/1: Action query time = 1.717 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=170: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 1.572 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9992 +t=186: Selected seed 195 with value = 0.9992 +Query 1/1: Action query time = 1.550 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=202: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.737 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=218: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.452 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=234: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.476 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=250: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.986 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=266: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.979 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=282: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 0.975 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=7--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=7--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 7 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 8... +Query 1/1: Action query time = 1.612 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4618 +t=10: Selected seed 195 with value = 0.4618 +Query 1/1: Action query time = 1.852 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5981 +t=26: Selected seed 195 with value = 0.5981 +Query 1/1: Action query time = 1.634 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6682 +t=42: Selected seed 195 with value = 0.6682 +Query 1/1: Action query time = 1.364 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7313 +t=58: Selected seed 195 with value = 0.7313 +Query 1/1: Action query time = 0.983 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7741 +t=74: Selected seed 195 with value = 0.7741 +Query 1/1: Action query time = 0.959 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8097 +t=90: Selected seed 195 with value = 0.8097 +Query 1/1: Action query time = 0.965 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8761 +t=106: Selected seed 195 with value = 0.8761 +Query 1/1: Action query time = 1.659 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9045 +t=122: Selected seed 195 with value = 0.9045 +Query 1/1: Action query time = 1.954 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9482 +t=138: Selected seed 195 with value = 0.9482 +Query 1/1: Action query time = 1.596 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9841 +t=154: Selected seed 195 with value = 0.9841 +Query 1/1: Action query time = 1.643 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9872 +t=170: Selected seed 195 with value = 0.9872 +Query 1/1: Action query time = 1.307 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9816 +t=186: Selected seed 195 with value = 0.9816 +Query 1/1: Action query time = 1.315 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9698 +t=202: Selected seed 195 with value = 0.9698 +Query 1/1: Action query time = 0.980 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9600 +t=218: Selected seed 195 with value = 0.9600 +Query 1/1: Action query time = 0.972 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9449 +t=234: Selected seed 195 with value = 0.9449 +Query 1/1: Action query time = 0.967 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9245 +t=250: Selected seed 195 with value = 0.9245 +Query 1/1: Action query time = 1.245 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9014 +t=266: Selected seed 195 with value = 0.9014 +Query 1/1: Action query time = 1.063 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8717 +t=282: Selected seed 195 with value = 0.8717 +Query 1/1: Action query time = 1.296 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8640 +t=298: Selected seed 195 with value = 0.8640 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=8--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=8--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 8 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 9... +Query 1/1: Action query time = 0.986 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4650 +t=10: Selected seed 195 with value = 0.4650 +Query 1/1: Action query time = 0.990 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5382 +t=26: Selected seed 195 with value = 0.5382 +Query 1/1: Action query time = 1.294 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6196 +t=42: Selected seed 195 with value = 0.6196 +Query 1/1: Action query time = 1.320 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7161 +t=58: Selected seed 195 with value = 0.7161 +Query 1/1: Action query time = 1.502 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8332 +t=74: Selected seed 195 with value = 0.8332 +Query 1/1: Action query time = 1.357 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9741 +t=90: Selected seed 195 with value = 0.9741 +Query 1/1: Action query time = 1.355 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.301 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9995 +t=122: Selected seed 195 with value = 0.9995 +Query 1/1: Action query time = 1.239 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9929 +t=138: Selected seed 195 with value = 0.9929 +Query 1/1: Action query time = 0.970 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9827 +t=154: Selected seed 195 with value = 0.9827 +Query 1/1: Action query time = 0.967 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9939 +t=170: Selected seed 195 with value = 0.9939 +Query 1/1: Action query time = 0.968 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9908 +t=186: Selected seed 195 with value = 0.9908 +Query 1/1: Action query time = 1.672 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9908 +t=202: Selected seed 195 with value = 0.9908 +Query 1/1: Action query time = 1.725 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9938 +t=218: Selected seed 195 with value = 0.9938 +Query 1/1: Action query time = 1.571 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9931 +t=234: Selected seed 195 with value = 0.9931 +Query 1/1: Action query time = 1.459 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9925 +t=250: Selected seed 195 with value = 0.9925 +Query 1/1: Action query time = 1.617 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=266: Selected seed 195 with value = 0.9914 +Query 1/1: Action query time = 1.379 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9914 +t=282: Selected seed 195 with value = 0.9914 +Query 1/1: Action query time = 0.998 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9877 +t=298: Selected seed 195 with value = 0.9877 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=9--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 9 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 10... +Query 1/1: Action query time = 1.429 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5103 +t=10: Selected seed 195 with value = 0.5103 +Query 1/1: Action query time = 1.290 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5485 +t=26: Selected seed 195 with value = 0.5485 +Query 1/1: Action query time = 1.337 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6205 +t=42: Selected seed 195 with value = 0.6205 +Query 1/1: Action query time = 0.970 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7628 +t=58: Selected seed 195 with value = 0.7628 +Query 1/1: Action query time = 0.974 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8437 +t=74: Selected seed 195 with value = 0.8437 +Query 1/1: Action query time = 1.119 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9884 +t=90: Selected seed 195 with value = 0.9884 +Query 1/1: Action query time = 1.300 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.307 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=122: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.203 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=138: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.013 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=154: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.209 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=170: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.232 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9998 +t=186: Selected seed 195 with value = 0.9998 +Query 1/1: Action query time = 1.113 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=202: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 1.157 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9959 +t=218: Selected seed 195 with value = 0.9959 +Query 1/1: Action query time = 1.228 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9834 +t=234: Selected seed 195 with value = 0.9834 +Query 1/1: Action query time = 1.265 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9925 +t=250: Selected seed 195 with value = 0.9925 +Query 1/1: Action query time = 1.248 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9756 +t=266: Selected seed 195 with value = 0.9756 +Query 1/1: Action query time = 1.266 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9492 +t=282: Selected seed 195 with value = 0.9492 +Query 1/1: Action query time = 1.207 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9243 +t=298: Selected seed 195 with value = 0.9243 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=10--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=10--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 10 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 11... +Query 1/1: Action query time = 1.756 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4734 +t=10: Selected seed 195 with value = 0.4734 +Query 1/1: Action query time = 1.794 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5736 +t=26: Selected seed 195 with value = 0.5736 +Query 1/1: Action query time = 1.913 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6610 +t=42: Selected seed 195 with value = 0.6610 +Query 1/1: Action query time = 1.501 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7186 +t=58: Selected seed 195 with value = 0.7186 +Query 1/1: Action query time = 1.544 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7465 +t=74: Selected seed 195 with value = 0.7465 +Query 1/1: Action query time = 1.023 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7911 +t=90: Selected seed 195 with value = 0.7911 +Query 1/1: Action query time = 1.030 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8509 +t=106: Selected seed 195 with value = 0.8509 +Query 1/1: Action query time = 0.997 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8802 +t=122: Selected seed 195 with value = 0.8802 +Query 1/1: Action query time = 1.790 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9796 +t=138: Selected seed 195 with value = 0.9796 +Query 1/1: Action query time = 1.686 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9888 +t=154: Selected seed 195 with value = 0.9888 +Query 1/1: Action query time = 1.612 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9994 +t=170: Selected seed 195 with value = 0.9994 +Query 1/1: Action query time = 1.462 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9973 +t=186: Selected seed 195 with value = 0.9973 +Query 1/1: Action query time = 1.557 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9980 +t=202: Selected seed 195 with value = 0.9980 +Query 1/1: Action query time = 1.351 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9977 +t=218: Selected seed 195 with value = 0.9977 +Query 1/1: Action query time = 1.247 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9988 +t=234: Selected seed 195 with value = 0.9988 +Query 1/1: Action query time = 0.975 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9990 +t=250: Selected seed 195 with value = 0.9990 +Query 1/1: Action query time = 0.973 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9976 +t=266: Selected seed 195 with value = 0.9976 +Query 1/1: Action query time = 0.965 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9908 +t=282: Selected seed 195 with value = 0.9908 +Query 1/1: Action query time = 1.743 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9818 +t=298: Selected seed 195 with value = 0.9818 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=11--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=11--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 11 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 12... +Query 1/1: Action query time = 1.394 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4934 +t=10: Selected seed 195 with value = 0.4934 +Query 1/1: Action query time = 1.415 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5412 +t=26: Selected seed 195 with value = 0.5412 +Query 1/1: Action query time = 1.507 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6310 +t=42: Selected seed 195 with value = 0.6310 +Query 1/1: Action query time = 1.455 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7460 +t=58: Selected seed 195 with value = 0.7460 +Query 1/1: Action query time = 1.495 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8609 +t=74: Selected seed 195 with value = 0.8609 +Query 1/1: Action query time = 1.448 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=90: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.409 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=106: Selected seed 195 with value = 1.0000 +Query 1/1: Action query time = 1.469 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9963 +t=122: Selected seed 195 with value = 0.9963 +Query 1/1: Action query time = 1.530 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9903 +t=138: Selected seed 195 with value = 0.9903 +Query 1/1: Action query time = 1.522 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9857 +t=154: Selected seed 195 with value = 0.9857 +Query 1/1: Action query time = 1.648 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9886 +t=170: Selected seed 195 with value = 0.9886 +Query 1/1: Action query time = 1.395 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9847 +t=186: Selected seed 195 with value = 0.9847 +Query 1/1: Action query time = 1.390 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9837 +t=202: Selected seed 195 with value = 0.9837 +Query 1/1: Action query time = 1.126 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9805 +t=218: Selected seed 195 with value = 0.9805 +Query 1/1: Action query time = 0.997 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9632 +t=234: Selected seed 195 with value = 0.9632 +Query 1/1: Action query time = 0.982 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9784 +t=250: Selected seed 195 with value = 0.9784 +Query 1/1: Action query time = 0.970 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9775 +t=266: Selected seed 195 with value = 0.9775 +Query 1/1: Action query time = 0.983 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9777 +t=282: Selected seed 195 with value = 0.9777 +Query 1/1: Action query time = 1.772 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9851 +t=298: Selected seed 195 with value = 0.9851 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=12--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=12--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 12 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 13... +Query 1/1: Action query time = 1.246 sec +t=10: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.4933 +t=10: Selected seed 195 with value = 0.4933 +Query 1/1: Action query time = 1.198 sec +t=26: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.5339 +t=26: Selected seed 195 with value = 0.5339 +Query 1/1: Action query time = 1.194 sec +t=42: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.6353 +t=42: Selected seed 195 with value = 0.6353 +Query 1/1: Action query time = 1.517 sec +t=58: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.7365 +t=58: Selected seed 195 with value = 0.7365 +Query 1/1: Action query time = 1.893 sec +t=74: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.8786 +t=74: Selected seed 195 with value = 0.8786 +Query 1/1: Action query time = 1.654 sec +t=90: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9766 +t=90: Selected seed 195 with value = 0.9766 +Query 1/1: Action query time = 1.355 sec +t=106: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9970 +t=106: Selected seed 195 with value = 0.9970 +Query 1/1: Action query time = 1.384 sec +t=122: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9964 +t=122: Selected seed 195 with value = 0.9964 +Query 1/1: Action query time = 1.039 sec +t=138: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9960 +t=138: Selected seed 195 with value = 0.9960 +Query 1/1: Action query time = 1.020 sec +t=154: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=154: Selected seed 195 with value = 0.9971 +Query 1/1: Action query time = 0.981 sec +t=170: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9979 +t=170: Selected seed 195 with value = 0.9979 +Query 1/1: Action query time = 1.573 sec +t=186: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9971 +t=186: Selected seed 195 with value = 0.9971 +Query 1/1: Action query time = 1.614 sec +t=202: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9779 +t=202: Selected seed 195 with value = 0.9779 +Query 1/1: Action query time = 1.769 sec +t=218: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9900 +t=218: Selected seed 195 with value = 0.9900 +Query 1/1: Action query time = 1.493 sec +t=234: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9676 +t=234: Selected seed 195 with value = 0.9676 +Query 1/1: Action query time = 1.403 sec +t=250: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9961 +t=250: Selected seed 195 with value = 0.9961 +Query 1/1: Action query time = 1.556 sec +t=266: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9984 +t=266: Selected seed 195 with value = 0.9984 +Query 1/1: Action query time = 1.364 sec +t=282: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 0.9997 +t=282: Selected seed 195 with value = 0.9997 +Query 1/1: Action query time = 1.367 sec +t=298: Current base seed: 195 +Query 1/1 (seed 195): Predicted value = 1.0000 +t=298: Selected seed 195 with value = 1.0000 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--episode=13--success=False--task=put_the_bowl_on_the_stove.mp4 +Saved rollout MP4 at path ./rollouts/CLbest_t15/2026_07_31-00_55_25--with_future_img--episode=13--success=False--task=put_the_bowl_on_the_stove.mp4 +Success: False +# episodes completed so far: 13 +# successes: 0 (0.0%) + +Task: put the bowl on the stove +Starting episode 14... diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/eval_CLbest_g1_t24.log b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/eval_CLbest_g1_t24.log new file mode 100644 index 0000000000000000000000000000000000000000..51b3a636b1b7d465e76eab6d8ca650af457930cb --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/eval_CLbest_g1_t24.log @@ -0,0 +1,34405 @@ +Uninstalled 1 package in 2ms +Installed 1 package in 17ms +[07-31 01:10:30|CRITICAL|cosmos_policy/_src/predict2/checkpointer/dcp.py:188:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[robosuite WARNING] No private macro file found! (__init__.py:7) +[robosuite WARNING] It is recommended to use a private macro file (__init__.py:8) +[robosuite WARNING] To setup, run: python /home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/robosuite/scripts/setup_macros.py (__init__.py:9) +Using LIBERO constants: + NUM_ACTIONS_CHUNK = 16 + ACTION_DIM = 7 + PROPRIO_DIM = 9 +If needed, manually set the correct constants in `projects/cosmos/cosmos_policy/constants.py`! +Loaded T5 text embeddings from /home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl onto device cuda:0 +Instantiating pretrained Cosmos model... +[07-31 01:10:32|INFO|cosmos_policy/_src/imaginaire/visualize/video.py:31:] No module named 'ffmpegcv' +[07-31 01:10:32|CRITICAL|cosmos_policy/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_policy._src.predict2.configs.video2world.experiment +[07-31 01:10:32|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v1... +[07-31 01:10:32|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v2... +[07-31 01:10:32|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering noframedrop_nocameramove_video_augmentor_v1... +[07-31 01:10:32|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering nocameramove_video_augmentor_v1... +[07-31 01:10:32|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor... +[07-31 01:10:32|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor_without_embeddings... +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_only2 +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_only2 +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_only2 +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_only2 +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_only2 +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_only2 +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_only2 +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_high_sigma +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_high_sigma +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_high_sigma +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_high_sigma +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma2 +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma2 +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma2 +[07-31 01:10:36|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt +[07-31 01:10:36|CRITICAL|cosmos_policy/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_policy.config.experiment +[07-31 01:10:36|INFO|cosmos_policy/_src/imaginaire/utils/checkpoint_db.py:927:get_checkpoint_by_hf] Downloading checkpoint from HuggingFace: nvidia/Cosmos-Predict2-2B-Video2World/model-480p-16fps.pt +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_base_stage +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_base_stage_inference_only +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_cl_stage +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_cl_stage_inference_only +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero__inference_only +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_task6_ft +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_task6_regen_cl +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_robocasa_50_demos_per_task +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_robocasa_50_demos_per_task__inference +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80 +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__inference_only +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func +[07-31 01:10:37|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func__inference_only +[07-31 01:10:39|INFO|cosmos_policy/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: /home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model +[07-31 01:10:39|INFO|cosmos_policy/_src/imaginaire/utils/misc.py:152:set_random_seed] Using random seed 0. +[07-31 01:10:39|INFO|cosmos_policy/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from /home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model +[07-31 01:10:39|WARNING|cosmos_policy/_src/predict2/models/text2world_model.py:145:__init__] DiffusionModel: precision torch.bfloat16 +[07-31 01:10:39|INFO|cosmos_policy/_src/imaginaire/utils/checkpoint_db.py:927:get_checkpoint_by_hf] Downloading checkpoint from HuggingFace: nvidia/Cosmos-Predict2-2B-Video2World/tokenizer/tokenizer.pth +[07-31 01:10:40|INFO|cosmos_policy/tokenizers/wan2pt1.py:165:_policy_video_vae] loading /home/azureuser/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2-2B-Video2World/snapshots/f50c09f5d8ab133a90cac3f4886a6471e9ba3f18/tokenizer/tokenizer.pth +[07-31 01:10:40|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on DiffusionModel: set_up_tokenizer: 0.88 s +[07-31 01:10:40|CRITICAL|cosmos_policy/_src/predict2/models/text2world_model.py:173:__init__] Using mean loss reduce with loss scale 10.0 +[07-31 01:10:40|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[07-31 01:10:40|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[07-31 01:10:40|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: [crossattn_emb] +[07-31 01:10:40|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1805:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720, for every 1 blocks. Total blocks: 28 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[07-31 01:10:41|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on meta to cuda and broadcast model states: 0.14 s +[07-31 01:10:41|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on Creating PyTorch model: 0.77 s +[07-31 01:10:41|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on Creating PyTorch model and ema if enabled: 0.77 s +[07-31 01:10:41|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on DiffusionModel: set_up_model: 0.77 s +[07-31 01:10:41|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on instantiate model: 1.68 s +[07-31 01:10:41|INFO|cosmos_policy/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model +/home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/torch/distributed/checkpoint/state_dict_loader.py:153: UserWarning: torch.distributed is disabled, unavailable or uninitialized, assuming the intent is to load in a single process. + warnings.warn( +[07-31 01:10:43|CRITICAL|cosmos_policy/_src/predict2/models/text2world_model.py:839:load_state_dict] load model in non-strict mode +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 01:10:43|CRITICAL|cosmos_policy/_src/predict2/models/text2world_model.py:840:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=[], unexpected_keys=[], incorrect_shapes=[]) +INFO:cosmos_policy.experiments.robot.robot_utils:Logging to local log file: cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal-cosmos-2026_07_31-01_10_31--CLbest_g1_t24.txt +INFO:cosmos_policy.experiments.robot.robot_utils:Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2,4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='CLbest_g1_t24', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=True, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200) +INFO:cosmos_policy.experiments.robot.robot_utils:Using serial inference (parallel inference disabled) +INFO:cosmos_policy.experiments.robot.robot_utils:Task suite: libero_goal +INFO:cosmos_policy.experiments.robot.robot_utils:Number of tasks: 10 +INFO:cosmos_policy.experiments.robot.robot_utils:Running only task IDs: [2, 4] +Eval config: PolicyEvalConfig(suite='libero', model_family='cosmos', config='cosmos_predict2_2b_480p_libero_cl_stage_inference_only', ckpt_path='/home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model', planning_model_config_name='', planning_model_ckpt_path='', config_file='cosmos_policy/config/config.py', packnet_mask_path='', packnet_first_cl_libero_task=6, use_third_person_image=True, num_third_person_images=1, use_wrist_image=True, num_wrist_images=1, use_proprio=True, flip_images=True, use_variance_scale=False, use_jpeg_compression=True, ar_future_prediction=False, ar_value_prediction=False, ar_qvalue_prediction=False, num_denoising_steps_action=5, num_denoising_steps_future_state=1, num_denoising_steps_value=1, unnormalize_actions=True, normalize_proprio=True, dataset_stats_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/libero_goal_regen/dataset_statistics.json', t5_text_embeddings_path='/home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl', trained_with_image_aug=True, chunk_size=16, num_open_loop_steps=16, deterministic=True, deterministic_reset=False, deterministic_reset_seed=None, use_ensemble_future_state_predictions=False, num_future_state_predictions_in_ensemble=3, future_state_ensemble_aggregation_scheme='average', use_ensemble_value_predictions=False, num_value_predictions_in_ensemble=5, value_ensemble_aggregation_scheme='average', search_depth=1, mask_current_state_action_for_value_prediction=False, mask_future_state_for_qvalue_prediction=False, num_queries_best_of_n=1, use_parallel_inference=False, available_gpus='0', parallel_timeout=15, parallel_tasks_across_gpus=False, task_suite_name='libero_goal', num_trials_per_task=50, task_ids_to_run='2,4', initial_states_path='DEFAULT', env_img_res=256, local_log_dir='cosmos_policy/experiments/robot/libero/logs/libero_goal', run_id_note='CLbest_g1_t24', use_wandb=False, wandb_entity='YOUR_ENTITY', wandb_project='YOUR_PROJECT', seed=195, randomize_seed=False, data_collection=False, jpeg_compress=True, reward_stop_threshold=None, reward_stop_consecutive_steps=3, use_gpt_reward=False, gpt_reward_model='gpt-5.4-mini', gpt_reward_stop_threshold=4, gpt_reward_timeout_s=30.0, save_rollout_video=True, eval_hdf5_path='', eval_hdf5_demo_key='demo_0', data_gen_demo_ids='', replay_actions=False, replay_action_path_dir='', replay_hdf5_path='', data_generation=False, max_generation_steps=200) +Using serial inference (parallel inference disabled) +[info] using task orders [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] +Task suite: libero_goal +Number of tasks: 10 +Running only task IDs: [2, 4] + 0%| | 0/10 [00:00 +Traceback (most recent call last): + File "/home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/robosuite/utils/binding_utils.py", line 199, in __del__ + self.gl_ctx.free() + File "/home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/robosuite/renderers/context/egl_context.py", line 149, in free + EGL.eglMakeCurrent(EGL_DISPLAY, EGL.EGL_NO_SURFACE, EGL.EGL_NO_SURFACE, EGL.EGL_NO_CONTEXT) + File "/home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/OpenGL/error.py", line 230, in glCheckError + raise self._errorClass( +OpenGL.raw.EGL._errors.EGLError: EGLError( + err = EGL_NOT_INITIALIZED, + baseOperation = eglMakeCurrent, + cArguments = ( + , + , + , + , + ), + result = 0 +) +Exception ignored in: +Traceback (most recent call last): + File "/home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/robosuite/renderers/context/egl_context.py", line 155, in __del__ + self.free() + File "/home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/robosuite/renderers/context/egl_context.py", line 149, in free + EGL.eglMakeCurrent(EGL_DISPLAY, EGL.EGL_NO_SURFACE, EGL.EGL_NO_SURFACE, EGL.EGL_NO_CONTEXT) + File "/home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/OpenGL/error.py", line 230, in glCheckError + raise self._errorClass( +OpenGL.raw.EGL._errors.EGLError: EGLError( + err = EGL_NOT_INITIALIZED, + baseOperation = eglMakeCurrent, + cArguments = ( + , + , + , + , + ), + result = 0 +) +sys:1: DeprecationWarning: builtin type swigvarlink has no __module__ attribute diff --git a/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/eval_CLbest_t1-5.log b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/eval_CLbest_t1-5.log new file mode 100644 index 0000000000000000000000000000000000000000..a1284bf4d1408471e440919bc1b193bee996f0db --- /dev/null +++ b/REGEN-main/cosmos_policy/experiments/robot/libero/logs/libero_goal/eval_CLbest_t1-5.log @@ -0,0 +1,12855 @@ +Uninstalled 1 package in 3ms +Installed 1 package in 17ms +[07-31 00:55:17|CRITICAL|cosmos_policy/_src/predict2/checkpointer/dcp.py:188:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[robosuite WARNING] No private macro file found! (__init__.py:7) +[robosuite WARNING] It is recommended to use a private macro file (__init__.py:8) +[robosuite WARNING] To setup, run: python /home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/robosuite/scripts/setup_macros.py (__init__.py:9) +Using LIBERO constants: + NUM_ACTIONS_CHUNK = 16 + ACTION_DIM = 7 + PROPRIO_DIM = 9 +If needed, manually set the correct constants in `projects/cosmos/cosmos_policy/constants.py`! +[info] using task orders [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] +[parallel_tasks_across_gpus] task assignment: {0: [1, 3, 5], 1: [2, 4]} +[07-31 00:55:25|CRITICAL|cosmos_policy/_src/predict2/checkpointer/dcp.py:188:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[robosuite WARNING] No private macro file found! (__init__.py:7) +[robosuite WARNING] It is recommended to use a private macro file (__init__.py:8) +[robosuite WARNING] To setup, run: python /home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/robosuite/scripts/setup_macros.py (__init__.py:9) +[07-31 00:55:25|CRITICAL|cosmos_policy/_src/predict2/checkpointer/dcp.py:188:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[robosuite WARNING] No private macro file found! (__init__.py:7) +[robosuite WARNING] It is recommended to use a private macro file (__init__.py:8) +[robosuite WARNING] To setup, run: python /home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/robosuite/scripts/setup_macros.py (__init__.py:9) +Using LIBERO constants: + NUM_ACTIONS_CHUNK = 16 + ACTION_DIM = 7 + PROPRIO_DIM = 9 +If needed, manually set the correct constants in `projects/cosmos/cosmos_policy/constants.py`! +Loaded T5 text embeddings from /home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl onto device cuda:0 +Instantiating pretrained Cosmos model... +[07-31 00:55:26|INFO|cosmos_policy/_src/imaginaire/visualize/video.py:31:] No module named 'ffmpegcv' +Using LIBERO constants: + NUM_ACTIONS_CHUNK = 16 + ACTION_DIM = 7 + PROPRIO_DIM = 9 +If needed, manually set the correct constants in `projects/cosmos/cosmos_policy/constants.py`! +Loaded T5 text embeddings from /home/azureuser/REGEN-main/LIBERO-Cosmos-Policy/success_only/t5_embeddings.pkl onto device cuda:0 +Instantiating pretrained Cosmos model... +[07-31 00:55:26|INFO|cosmos_policy/_src/imaginaire/visualize/video.py:31:] No module named 'ffmpegcv' +[07-31 00:55:26|CRITICAL|cosmos_policy/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_policy._src.predict2.configs.video2world.experiment +[07-31 00:55:26|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v1... +[07-31 00:55:26|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v2... +[07-31 00:55:26|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering noframedrop_nocameramove_video_augmentor_v1... +[07-31 00:55:26|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering nocameramove_video_augmentor_v1... +[07-31 00:55:26|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor... +[07-31 00:55:26|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor_without_embeddings... +[07-31 00:55:27|CRITICAL|cosmos_policy/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_policy._src.predict2.configs.video2world.experiment +[07-31 00:55:27|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v1... +[07-31 00:55:27|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering video_basic_augmentor_v2... +[07-31 00:55:27|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering noframedrop_nocameramove_video_augmentor_v1... +[07-31 00:55:27|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering nocameramove_video_augmentor_v1... +[07-31 00:55:27|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor... +[07-31 00:55:27|INFO|cosmos_policy/_src/predict2/datasets/augmentor_provider.py:94:augmentor_register] registering image_basic_augmentor_without_embeddings... +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_only2 +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_only2 +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_only2 +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_only2 +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_only2 +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_only2 +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_only2 +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_only2 +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_only2 +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_high_sigma +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_high_sigma +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_high_sigma +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_high_sigma +[07-31 00:55:30|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_only2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_only2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_only2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_only2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_only2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_high_motion_high_sigma +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_av_high_sigma +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_physical_ai_high_sigma +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_4k_cooldown_high_sigma +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_face_high_sigma2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt +[07-31 00:55:31|CRITICAL|cosmos_policy/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_policy.config.experiment +[07-31 00:55:31|INFO|cosmos_policy/_src/imaginaire/utils/checkpoint_db.py:927:get_checkpoint_by_hf] Downloading checkpoint from HuggingFace: nvidia/Cosmos-Predict2-2B-Video2World/model-480p-16fps.pt +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_crowded_high_sigma2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-1-Size-2B-Res-720-Fps-16-Note-rf_robotics_high_sigma2 +[07-31 00:55:31|INFO|cosmos_policy/_src/predict2/configs/video2world/experiment/specialized_model/SFT_2B_RF.py:853:] Storing Stage-c_pt_4-Index-2-Size-2B-Res-720-Fps-16-Note-rf_with_edm_ckpt +[07-31 00:55:31|CRITICAL|cosmos_policy/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_policy.config.experiment +[07-31 00:55:31|INFO|cosmos_policy/_src/imaginaire/utils/checkpoint_db.py:927:get_checkpoint_by_hf] Downloading checkpoint from HuggingFace: nvidia/Cosmos-Predict2-2B-Video2World/model-480p-16fps.pt +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_base_stage +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_base_stage_inference_only +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_cl_stage +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_cl_stage_inference_only +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero__inference_only +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_task6_ft +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_task6_regen_cl +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_robocasa_50_demos_per_task +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_robocasa_50_demos_per_task__inference +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80 +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__inference_only +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func__inference_only +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_base_stage +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_base_stage_inference_only +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_cl_stage +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_cl_stage_inference_only +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero__inference_only +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_task6_ft +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_libero_goal_task6_regen_cl +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_robocasa_50_demos_per_task +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_robocasa_50_demos_per_task__inference +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80 +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__inference_only +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func +[07-31 00:55:31|INFO|cosmos_policy/config/experiment/cosmos_policy_experiment_configs.py:942:register_configs] Registering experiment: cosmos_predict2_2b_480p_aloha_185_demos_4_tasks_mixture_foldshirt15_candiesinbowl45_candyinbag45_eggplantchickenonplate80__resumeFrom50K_648_rollouts_Vsprime_value_func__inference_only +[07-31 00:55:33|INFO|cosmos_policy/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: /home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model +[07-31 00:55:33|INFO|cosmos_policy/_src/imaginaire/utils/misc.py:152:set_random_seed] Using random seed 0. +[07-31 00:55:33|INFO|cosmos_policy/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from /home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model +[07-31 00:55:33|WARNING|cosmos_policy/_src/predict2/models/text2world_model.py:145:__init__] DiffusionModel: precision torch.bfloat16 +[07-31 00:55:33|INFO|cosmos_policy/_src/imaginaire/utils/checkpoint_db.py:927:get_checkpoint_by_hf] Downloading checkpoint from HuggingFace: nvidia/Cosmos-Predict2-2B-Video2World/tokenizer/tokenizer.pth +[07-31 00:55:33|INFO|cosmos_policy/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: /home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model +[07-31 00:55:33|INFO|cosmos_policy/_src/imaginaire/utils/misc.py:152:set_random_seed] Using random seed 0. +[07-31 00:55:33|INFO|cosmos_policy/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from /home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model +[07-31 00:55:33|WARNING|cosmos_policy/_src/predict2/models/text2world_model.py:145:__init__] DiffusionModel: precision torch.bfloat16 +[07-31 00:55:33|INFO|cosmos_policy/_src/imaginaire/utils/checkpoint_db.py:927:get_checkpoint_by_hf] Downloading checkpoint from HuggingFace: nvidia/Cosmos-Predict2-2B-Video2World/tokenizer/tokenizer.pth +[07-31 00:55:34|INFO|cosmos_policy/tokenizers/wan2pt1.py:165:_policy_video_vae] loading /home/azureuser/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2-2B-Video2World/snapshots/f50c09f5d8ab133a90cac3f4886a6471e9ba3f18/tokenizer/tokenizer.pth +[07-31 00:55:34|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on DiffusionModel: set_up_tokenizer: 0.91 s +[07-31 00:55:34|CRITICAL|cosmos_policy/_src/predict2/models/text2world_model.py:173:__init__] Using mean loss reduce with loss scale 10.0 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: [crossattn_emb] +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[07-31 00:55:34|INFO|cosmos_policy/tokenizers/wan2pt1.py:165:_policy_video_vae] loading /home/azureuser/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2-2B-Video2World/snapshots/f50c09f5d8ab133a90cac3f4886a6471e9ba3f18/tokenizer/tokenizer.pth +[07-31 00:55:34|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on DiffusionModel: set_up_tokenizer: 0.95 s +[07-31 00:55:34|CRITICAL|cosmos_policy/_src/predict2/models/text2world_model.py:173:__init__] Using mean loss reduce with loss scale 10.0 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: [crossattn_emb] +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1805:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720, for every 1 blocks. Total blocks: 28 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[07-31 00:55:34|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[07-31 00:55:34|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on meta to cuda and broadcast model states: 0.14 s +[07-31 00:55:34|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on Creating PyTorch model: 0.76 s +[07-31 00:55:34|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on Creating PyTorch model and ema if enabled: 0.77 s +[07-31 00:55:34|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on DiffusionModel: set_up_model: 0.77 s +[07-31 00:55:35|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on instantiate model: 1.70 s +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1805:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720, for every 1 blocks. Total blocks: 28 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/networks/minimal_v4_dit.py:1811:enable_selective_checkpoint] Enable selective checkpoint for block 27 +/home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/torch/distributed/checkpoint/state_dict_loader.py:153: UserWarning: torch.distributed is disabled, unavailable or uninitialized, assuming the intent is to load in a single process. + warnings.warn( +[07-31 00:55:35|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on meta to cuda and broadcast model states: 0.14 s +[07-31 00:55:35|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on Creating PyTorch model: 0.77 s +[07-31 00:55:35|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on Creating PyTorch model and ema if enabled: 0.78 s +[07-31 00:55:35|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on DiffusionModel: set_up_model: 0.78 s +[07-31 00:55:35|INFO|cosmos_policy/_src/imaginaire/utils/timer.py:143:_log] Time spent on instantiate model: 1.75 s +[07-31 00:55:35|INFO|cosmos_policy/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /home/azureuser/REGEN-main/old/cosmos_policy/cosmos_v2_finetune/cosmos_predict2_2b_480p_libero_goal_task6_regen_cl_from40k_2gpu/best_action_l1/model +/home/azureuser/REGEN-main/.venv/lib/python3.10/site-packages/torch/distributed/checkpoint/state_dict_loader.py:153: UserWarning: torch.distributed is disabled, unavailable or uninitialized, assuming the intent is to load in a single process. + warnings.warn( +[07-31 00:55:37|CRITICAL|cosmos_policy/_src/predict2/models/text2world_model.py:839:load_state_dict] load model in non-strict mode +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:37|CRITICAL|cosmos_policy/_src/predict2/models/text2world_model.py:840:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=[], unexpected_keys=[], incorrect_shapes=[]) +INFO:cosmos_policy.experiments.robot.robot_utils:Logging to local log file: cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal_gpu1-cosmos-2026_07_31-00_55_25--CLbest_t15--gpu1.txt +INFO:cosmos_policy.experiments.robot.robot_utils:[worker gpu=1] tasks=[2, 4] CUDA_VISIBLE_DEVICES=1 +INFO:cosmos_policy.experiments.robot.robot_utils:[worker gpu=1] starting task_id=2 +INFO:cosmos_policy.experiments.robot.robot_utils:Using default initial states +[07-31 00:55:38|CRITICAL|cosmos_policy/_src/predict2/models/text2world_model.py:839:load_state_dict] load model in non-strict mode +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|WARNING|cosmos_policy/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[07-31 00:55:38|CRITICAL|cosmos_policy/_src/predict2/models/text2world_model.py:840:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=[], unexpected_keys=[], incorrect_shapes=[]) +INFO:cosmos_policy.experiments.robot.robot_utils:Logging to local log file: cosmos_policy/experiments/robot/libero/logs/libero_goal/ENV_EVAL-libero_goal_gpu0-cosmos-2026_07_31-00_55_25--CLbest_t15--gpu0.txt +INFO:cosmos_policy.experiments.robot.robot_utils:[worker gpu=0] tasks=[1, 3, 5] CUDA_VISIBLE_DEVICES=0 +INFO:cosmos_policy.experiments.robot.robot_utils:[worker gpu=0] starting task_id=1 +INFO:cosmos_policy.experiments.robot.robot_utils:Using default initial states +[worker gpu=1] tasks=[2, 4] CUDA_VISIBLE_DEVICES=1 +[info] using task orders [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] +[worker gpu=1] starting task_id=2 +Using default initial states +Local assets not found. Downloading from HuggingFace Hub... +Assets already downloaded at /home/azureuser/.cache/libero/assets + 0%| | 0/50 [00:00